Estimation and Provision of Differentiated Quality of Supply in Distribution Networks

Size: px
Start display at page:

Download "Estimation and Provision of Differentiated Quality of Supply in Distribution Networks"

Transcription

1 Estimation and Provision of Differentiated Quality of Supply in Distribution Networks A thesis submitted to The University of Manchester for the Degree of Doctor of Philosophy in the Faculty of Engineering and Physical Sciences 2016 Mr Sami Abdelrahman, B.Sc., M.Sc., School of Electrical and Electronic Engineering

2 2

3 Table of Contents List of Figures... 6 List of Tables... 9 List of Acronyms Abstract Declaration Copyright Statement Acknowledgement Introduction Definitions and terminology PQ renewed interest and concerns Why now? Problem formulation Literature review Harmonics modelling and analysis Power Quality evaluation Literature review summary Aim and objectives Contributions Thesis Overview Power Quality Categorisation PQ standards and guides PQ monitoring PQ estimation PQ Phenomena Harmonics Voltage sag Voltage unbalance Summary Power System Modelling for PQ studies Test systems Generic Distribution Network (GDN) Real test feeder PQ phenomena modelling

4 3.2.1 Modelling for harmonics Modelling of voltage sag Modelling of voltage unbalance Summary Harmonic Estimation in Distribution Networks Probabilistic Assessment of Harmonics Models and Methodologies Results of different case studies Probabilistic estimation of harmonics Harmonics in radial feeders Feeder harmonic analysis Harmonic estimation algorithms Harmonic propagation with capacitors connected to a feeder Summary Global PQ Evaluation Indices The need for unified PQ indices PQ reserve index (PQR) The Framework Planning levels Immunity levels (Thresholds) Calculating PQR PQR results and applications Compound bus PQ index (CBPQI) Analytic Hierarchy Process (AHP) Case study and CBPQI results Application of PQ indices in real measurements Measurement synthesizing PQ performance analysis Flexibility of the indices Results of the analysis Summary The concept of Provision of differentiated PQ General concept

5 6.2 Application of CBPQI for provision of differentiated PQ Optimal provision of differentiated PQ Summary Conclusions and Future Work Major conclusions Recommendations Future work Assessment of PQ requirements Network modelling for PQ studies Uncertainties PQ estimation/evaluation Optimizing PQ mitigation References Appendix A. Test Systems Appendix B. Wind and PV output profiles Appendix C. Transformers zero sequence circuits Appendix D. PQ Measurements Appendix E. Author s Thesis Based Publications E.1 International Journal Papers E.2 International Conference Papers E.3 Technical reports Word count: 63,417 5

6 List of Figures Figure 1-1: PQ disturbances survey results in Europe and the US [11] Figure 1-2: Percentile values from PDF and CDF plots Figure 1-3: PQ Contracts in Deregulated Market (adopted from [57]) Figure 1-4: PQ based Customer Classification (adopted from [6]) Figure 1-5: The Customised Damage Function of a plant, (adopted from [5]) Figure 1-6: Harmonic distortion vs DG penetration generic dependency (adopted from [61]) Figure 2-1: The planning and compatibility levels of a PQ phenomenon (adopted from [71]) Figure 2-2: PQ monitoring from DSOs survey (adopted from [73]) Figure 2-3: Survey results about PQ monitoring (adopted from [65]) Figure 2-4: Framework of harmonic state estimation (adopted from [65]) Figure 2-5: Overview of some PQ phenomena definitions (adopted from [78]) Figure 2-6: General PQ evaluation process (adopted from [10]) Figure 2-7: A general framework for obtaining system-wide sag indices (adopted from [89]) Figure 2-8: Probable regions of voltage sag events (adopted from [97]) Figure 3-1: Generic Distribution Network (GDN) single line diagram Figure 3-2: Segmented domestic load duration curve (LDC) of GDN network Figure 3-3: Output curves of PV and wind generators for January and May Figure 3-4: Real test feeder single line diagram Figure 3-5: Segmented total load duration curve (LDC) of real test feeder Figure 3-6: Daily loading curves for the real test feeder total load Figure 3-7: Bus 15 PQ measurements Figure 3-8: Norton equivalent of harmonic sources [131] Figure 3-9: Impedance frequency scan Figure 3-10: 3 rd harmonic currents flowing through both terminals of the test feeder lines93 Figure 3-11: Harmonic performance at the end of the feeder (capacitor connection point) 97 Figure 3-12: Voltage tolerance curves Figure 3-13: Heat map of SSI with different values for parameter a Figure 3-14: Ranges of sampled power factor for unbalance simulation [76] Figure 4-1: Random harmonics injections (lower uncertainty) Figure 4-2: Segmented domestic LDC Figure 4-3: Segmented DG accumulated output curve Figure 4-4: GDN harmonic performance (Case I) Figure 4-5: Heat Maps identifying the most affected areas before and after DG connections (Case I) Figure 4-6: Daily/Annual PDF THD results of phase A for bus (a) 196 (b) Figure 4-7: Annual THD CDF results of phase A for bus (a) 196 (b) Figure 4-8: GDN harmonic performance (Case II)

7 Figure 4-9: Heat Maps identifying the most affected areas before and after DG connections (Case II) Figure 4-10: Heat Maps for different hours during a day Figure 4-11: Heat Maps for the areas that violated the IEEE standard limits Figure 4-12: Impedance frequencies scan for selected GDN buses (resonance near 13 th harmonic) Figure 4-13: Harmonic voltages and THD performance for selected GDN buses (Case II) Figure 4-14: Heat maps for real test feeder harmonic performance (daily study) Figure 4-15: THD increases with the feeder length Figure 4-16: Characteristic buses locations on the test feeder Figure 4-17: THD values for superposition case study Figure 4-18: THD vs. total harmonic electric distance Superposition case study Figure 4-19: THD vs. total harmonic electric distance Generic case study Figure 4-20: Buses electric distance for different harmonic frequencies Figure 4-21: General case study results Figure 4-22: Impacts on harmonic performance of the feeder by increased harmonic injection at Bus Figure 4-23: Coarse estimation for the general case Figure 4-24: A case of 5 th harmonic currents not seen at the substation Figure 4-25: Flow chart of probabilistic estimation methodology Figure 4-26: True/Estimated harmonic performance for the general case study Figure 4-27: Estimation errors for the probabilistic estimation Figure 4-28: CDF of harmonic performance for buses 1, 15 and Figure 4-29: Comparison of 95th percentile THD values and 3 rd and 11 th harmonic voltage values and corresponding estimation errors Figure 4-30: LC circuit parameters Figure 4-31: LC circuit currents and voltages Figure 4-32: Simple 3-bus circuit resonance analysis Figure 4-33: Characteristic buses frequency scan with capacitors connected Figure 4-34: Estimated/true 95th percentiles for THD, V9, V11 and V13 six capacitors (100 kvar) connected Figure 4-35: Estimated/true 95th percentiles for THD, V9, V11 and V13 not monitored single capacitor (600 kvar) at Bus Figure 4-36: Estimated/true 95th percentiles for THD, V9, V11 and V13 monitored single capacitor (600 kvar) at Bus Figure 4-37: Estimated/true 95 th percentiles for THD, V9, V11 and V13 monitored single capacitor (600 kvar) at Bus 35 (resonance considered) Figure 5-1: Framework for the overall PQ evaluation using PQR Figure 5-2: Harmonics and unbalance sampled thresholds levels (THD red/dashed, VUF blue/solid) Figure 5-3: Sensitive areas Figure 5-4: Heat maps for overall and separate PQ GDN performance (Note: The heat bars have different ranges and colour for 0% reserve)

8 Figure 5-5: PQ sensitive and poor performing overlapped areas Figure 5-6: AHP model for manager selection example Figure 5-7: AHP model for calculating CBPQI Figure 5-8: Framework for the overall PQ evaluation using CBPQI Figure 5-9: GDN harmonic performance Figure 5-10: GDN unbalance performance Figure 5-11: GDN voltage sag performance Figure 5-12: Uniformly distributed weighting ranges for PQ phenomena Figure 5-13: Overall PQ performance based on CBPQI Figure 5-14: Worst performing buses (normalized) Figure 5-15: Site 1 PQ phenomena variation Figure 5-16: Site 1 PQ performance (normalized based on thresholds) Figure 5-17: Segmented normal distribution function for weighting factors calculations (adopted from [142]) Figure 5-18: Weighting factors of phenomena for each site Figure 5-19: 2013 vs 2014 PQ performances Figure 5-20: Seasonal PQ performances Figure 5-21: Site 5 performances Figure 5-22: Site 6 performances Figure 5-23: Average and weekly sites rank (original case) Figure 5-24: Average and weekly sites ranks (flexible cases) Figure 6-1: The concept of differentiated PQ provision to different types of loads Figure 6-2: PQ agreements in deregulated electricity market environment (modified from [57]) Figure 6-3: GDN different PQ zones Figure 6-4: Flowchart of the optimization methodology Figure 6-5: CBPQI, BPI, THD and VUF performances before and after mitigation Figure 6-6: Hourly PQ performance over a day before and after application of mitigation solution Figure 6-7: Peak hour PQ overall performance before and after application of mitigation solutions (CBPQI normalized based on worst bus) Figure B-1: PV output monthly profiles Figure B-2: Wind generation monthly profiles Figure C-1: Zero sequence circuits for the different transformers connections (adopted from [77]) Figure D-1: PQ measurements for harmonics, unbalance and flicker for 8 LV sites

9 List of Tables Table 2-1: Summary of relevant PQ standards (adopted from [65]) Table 2-2: Comparison of the harmonic voltage limits between different standards and guidelines (adopted from [55]) Table 2-3: Voltage sag severity with reference to SEMI F47 curves (adopted from [68]). 68 Table 3-1: GDN transformers modelling Table 3-2: Transformers connection impact on harmonic propagation Table 3-3: Capacitor different connections impact on the real test feeder harmonic performance Table 3-4: Faults rate and types for sag tables calculation [134] Table 3-5: Fault clearing time for sag tables calculations Table 4-1: Non-linear loads harmonic injections levels [86] Table 4-2: DG harmonic injections levels [63, 135] Table 4-3: The most affected buses annual harmonic performance (Case I) Table 4-4: The most affected buses annual harmonic performance (Case II) Table 4-5: Characteristic buses electric distance to Bus 01 (S/S) Table 4-6: Total and individual harmonic electric distances from substation (Bus 01) in Ohm Table 4-7: Simple 3-bus network parameters Table 5-1: Worst performing buses based on separate and PQR rankings of reserves (in p.u.) Table 5-2: Comparison of alternatives pairwise matrix Table 5-3: Comparison of priorities pairwise matrix Table 5-4: Comparison of sub-priorities pairwise matrix Table 5-5: Example priorities calculation Table 5-6: Example buses performances Table 5-7: Example score calculation Table 5-8: Example score weighting Table 5-9: Buses ranked based on CBPQI Table 5-10: Measurements Summary Table 5-11: Normal distribution segments for weightings calculations Table 5-12: Weighting factors of phenomena for each site Table 5-13: Importance of sites considered Table 5-14: PQ performance comparison based on the mode as a measure Table 5-15: PQ performance comparison based on the 95 th percentile as a measure Table 5-16: Ranks of site based on different cases of weighting phenomena and sites (best site on top) Table 5-17: Average site ranking (from worst site) for different case studies Table 6-1: Optimal solutions for different cases Table A-1: GDN lines details Table A-2: GDN transformers details Table A-3: GDN load details

10 Table A-4: Real test feeder lines details Table A-5: Real test feeder transformer details Table A-6: Real test feeder load details

11 List of Acronyms μ mean of a sample α standard deviation of a sample AHP Analytic Hierarchy Process ANSI American National Standards Institute BPI S Bus Performance Index for Sag CBPQI Compound Bus Power Quality Index CDF Cumulative Distribution Function CEER Council of European Energy Regulators DER Distributed Energy Resources DFIG Doubly Fed Induction Generator DG Distributed Generation DSI Duration Severity Index DSM Demand Side Management DSO Distribution System Operator DVR Dynamic Voltage Restorer EMC Electromagnetic Compatibility EV Electric Vehicle FACTS Flexible AC Transmission System GDN Generic Distribution Network GSP Grid Supply Point h harmonic frequency HGI Harmonics Gap Index h r resonance harmonic frequency HSE Harmonic State Estimation HV High Voltage I 1 fundamental current IEC International Electro-technical Commission IEEE Institute of Electrical and Electronics Engineers I h current at the h th frequency LDC Load Duration Curve LV Low Voltage MC Monte Carlo simulation MDSI Magnitude Duration Severity Index MSI Magnitude Severity Index MV Medium Voltage OHL Overhead Lines PCC Point of Common Coupling PDF Probability Density Function PF Passive Filter Plt long-term flicker index PQ Power Quality PQGI Power Quality Gap Index PQR Power Quality Reserve index Pst Short-term flicker index 11

12 p.u. PV QoS RES RMS dev S/S SCC SGI SSI STATCOM SVC THDc THDv/THD UGI UPQI UPS V 0 V 1 V 2 V h VPP VQ VSC VSD/ASD VUF WLS w x Z thd Per Unit Photovoltaic Quality of Supply or Quality of Service Renweable Energy Sources RMS voltage deviation from nominal Substation Short Circuit Capacity Sag Gap Index Sag Severity Index Static Compensator Static VAR Compensator current total harmonic distortion voltage total harmonic distortion Unbalance Gap Index Unified Power Quality Index Uninterrupted Power Supply zero sequence voltage positive sequence/fundamental voltage negative sequence voltage voltage at the h th frequency Virtual Power Plant Voltage Quality Voltage Source Converter Variable Speed Drive/Adjustable Speed Drive Voltage Unbalance Factor Weighted Least Square weighting factor of phenomenon x harmonic distortion electric distance 12

13 Abstract Title: Estimation and Provision of Differentiated Quality of Supply in Distribution Networks Sami Abdelrahman, The University of Manchester, July 2016 Power quality problems are now receiving much interest from the distribution system operators (DSO). This can be attributed to rising competition in today s electricity markets, new regulations and standards regarding the power quality, and also because newer types of load equipment require certain levels of power quality. Furthermore, the current trend towards increased penetration of distributed generation (DG) in the distribution network (in particular electronics interfaced and intermittent DG) has its impact on the power quality performance of the distribution networks, which also draw more attention of the DSOs to the power quality issues. The research presented in the thesis is divided into two main parts; the first part is to estimate and evaluate power quality indices in distribution networks with DG. By comparing the estimated PQ performance with the customers PQ requirements the weak areas of the network are identified, i.e., the areas requiring power quality improvements. Most of the PQ phenomena are well defined in international standards and appropriate thresholds for individual phenomena are set in these standards. The thesis presents simulation and evaluation methodologies for some of the main PQ phenomena. Particular attention is given to the longer terms study of harmonics in distribution network with limited monitoring. The second part deals with the optimum provision of differentiated power quality based on the temporal and spatial variations of the network state and the customers power quality requirements. Two new global PQ indices are developed to evaluate the overall PQ of a bus in compressed format. The proposed global indices are the Compound Bus PQ Index (CBPQI) which is based on Analytic Hierarchy Process (AHP) methodology and the PQ Reserve index (PQR) which is based on weighted average of performance of different phenomena. The CBPQI is then used as an objective function to optimize the selection of the PQ mitigation solutions. This involves utilising the available power quality mitigation solutions and considering the new smart technologies for network operation in order to provide the optimum level of power quality to different areas of the network at different times. The research outcomes contribute to the benchmarking of power quality performance, quantifying the impact of DG on the power quality in the networks, and to identifying the areas of distribution networks where the investments in power quality monitors and solutions should be made. The studies presented in the thesis are based on both simulations and real PQ measurements. All the simulations are performed in DigSilent PowerFactory and OpenDSS simulation software packages. The studies are performed on a generic distribution network and a real distribution feeder. 13

14 Declaration No portion of the work referred to in this thesis has been submitted in support of an application for another degree or qualification of this or any other university or institute of learning. 14

15 Copyright Statement i. The author of this thesis (including any appendices and/or schedules to this thesis) owns certain copyrights or related rights in it (the Copyright ) and she has given The University of Manchester certain rights to use such Copyright, including for administrative purposes. ii. Copies of this thesis, either in full or in extracts and whether in hard or electronic copy, may be made only in accordance with the Copyright, Designs and Patents Act 1988 (as amended) and regulations issued under it or, where appropriate, in accordance with licensing agreements which the University has from time to time. This page must form part of any such copies made. iii. The ownership of certain Copyright, patents, designs, trademarks and other intellectual property (the Intellectual Property Rights ) and any reproductions of copyright works in the thesis, for example graphs and tables ( Reproductions ), which may be described in this thesis, may not be owned by the author and may be owned by third parties. Such Intellectual Property Rights and Reproductions cannot and must not be made available for use without the prior written permission of the owner(s) of the relevant Intellectual Property and/or Reproductions. iv. Further information on the conditions under which disclosure, publication and commercialisation of this thesis, the Copyright and any Intellectual Property University IP Policy (see in any relevant Thesis restriction declarations deposited in the University Library, The University Library s regulations (see and in The University s policy on Presentation of Theses. 15

16 Acknowledgement I would like to express my gratitude and appreciation to my supervisor Prof. Jovica V. Milanović for the patience, guidance and trust he showed during the period of the research. I would like to thank the SuStainable project board and partners, for the sponsorship of the research, the continuous feedback, the provision of data and for the knowledge sharing regarding the industrial practices. My great appreciation goes to my colleagues in Power Quality and Power Systems Dynamics Team in the Electrical Energy and Power System Group of the School of Electrical and Electronic Engineering at The University of Manchester, in particular Dr. Huilian Liao and Dr Selma Awadallah for their continuous help and useful discussions. Special thanks go to Dr. Jan Meyer and his PQ group at the Technical University of Dresden, Germany, for sharing their data, knowledge and experience in PQ measurements during the six months of collaborative work. I would also like to thank the Sudanese community in Manchester for their support during stressful periods of the PhD, in particular my friends Abdelmagid, Ahmed, Amgad, Elamin, Fatin, Israa, M. Atta, M. Elzubair, Razan, Sara and the Tahas. 16

17 17 To my family

18 1 Introduction Nowadays the term Power Quality (PQ) is commonly used to group certain electromagnetic phenomena that take place in power systems. Any phenomenon that alters the power system s voltage waveform from a sinusoid with a constant amplitude and frequency can be considered a PQ phenomenon. The term itself is not new, according to [1] it was firstly used in a power system publication in the late 1960s. The term is used interchangeably with the term Quality of Supply or Quality of Service (QoS). Voltage Quality (VQ) has also been used to describe these phenomena. However, a shift from VQ to PQ usage and greater acceptance of the term Power Quality were recorded during the 1990s [1]. This is simply because some of the PQ phenomena cannot be described based on the voltage quality only, and the current quality is not separable from the evaluation of these phenomena. Phenomena like harmonics and unbalance are equally affected by the voltage and currents distortions and unbalance. Furthermore, some parameters like power factors and short circuit capacity are now routinely included in the PQ evaluation of a bus. 18

19 1.1 Definitions and terminology The PQ is defined in IEC [2] as characteristics of the electricity at a given point on an electrical system, evaluated against a set of reference technical parameters with additional note of These parameters might, in some cases, relate to the compatibility between electricity supplied on a network and the loads connected to that network. On the other hand the IEEE 1100 standard [3] defines PQ as The concept of powering and grounding electronic equipment in a manner that is suitable to the operation of that equipment and compatible with the premise wiring system and other connected equipment. Regardless of the discrepancy in the two main standardizing institutes in PQ definitions, both institutes indicate the importance of considering load and equipment when analysing the PQ issues. The referral to load and equipment in the definitions indicates the importance of considering the customer PQ requirements in the PQ analysis. The customer is considered to be any type of load (and recently embedded generators), that can range from a few kvas households to ten or more MVAs industrial plants, and connected to a utility Point of Common Coupling (PCC). The PCC, as defined in [4], is the point on a public power supply network, electrically nearest to a particular load, at which other loads are, or could be, connected. On the other hand, the PQ phenomenon or the PQ disturbance can have a more precise definition that is easier to grasp. For a three phase balanced power system any phenomenon, incidental or continuous, that changes the voltage and/or current waveforms from a pure, constant magnitude and constant frequency sinusoids with equal phase shifts can be considered a PQ disturbance. Nevertheless, to have such an ideal waveform is a hypothetical scenario and cannot be found in practice. Fortunately, equipment is designed to perform adequately under certain levels of variation in some of the aforementioned voltage and current parameters, the so-called electromagnetic compatibility (EMC). The 19

20 EMC is defined in [5] as The ability of an equipment or system to function satisfactorily in its electromagnetic environment without introducing intolerable electromagnetic disturbances to anything in that environment. The EMC is extensively studied and presented in the IEC standard family. The standard is provided in six main parts: Part 1 - General, Part 2 - Environment, Part 3 Limits, Part 4 Testing and measurement techniques, Part 5 Installation and mitigation guidelines and Part 6 Generic standards. Moreover, when analysing PQ disturbances in a network certain aspects are usually considered. Aspects like origin of disturbance, emission level, propagation and immunity are repeatedly mentioned in PQ standards and reports. Techno-Economic analysis has been also performed as a part of the PQ analysis [6-8]. The origin of a PQ disturbance is the location in the network where the disturbance initiated or first took place. The fault location is considered the origin of a voltage sag disturbance; similarly the connection point of an unbalanced load or single phase generator can be considered the origin of the unbalance phenomenon in the network. The emission level is the level of the electromagnetic energy that emanates from the disturbance source [5]. The emission level is quantified to determine how severe a certain disturbance is. Measuring the harmonic currents injected from a source is an example of quantifying the emission level of a PQ disturbance (harmonics). The propagation of a disturbance is a measure of how the disturbance is conducted to other buses and network components through the power system [1]. The immunity of the equipment is a measure of the level of the disturbance at the supply at which the equipment can still be electromagnetically compatible, i.e., function satisfactorily. The Techno-Economic analysis of a PQ disturbance, in very broad terms, is the evaluation of an existing overlap between the immunity of equipment operating under a certain disturbance level, how much it costs to operate in such an environment and how much it costs to eliminate this overlap. 20

21 1.2 PQ renewed interest and concerns Why now? As discussed above, PQ is not a new issue in power system networks albeit it was addressed under different names and fields in the power system analysis. Nevertheless, the interest and concerns about PQ have been renewed since the early 1990s. The deregulation of the electricity sector is widely applied now and the competition in the electricity market is increasing. This led to considering electricity as a product with a measurable quality. Another main reason for the renewed interest is the increasing deployment and acceptance of smart grids technologies in contemporary and future power networks. Some of these technologies are both sources and victims of PQ disturbances. The following list summarizes some of the PQ challenges and concerns facing contemporary utilities and regulators [1, 9-11]: - The new generation of load is based on microprocessors and power electronic devices which are more sensitive to PQ variation than the old electromechanical based loads. - The drive to increase the efficiency of the power system resulted in the increased deployment of devices such as high efficiency adjustable speed drives, conditioning devices (e.g. FACTS) and power factor correction capacitors to the network. These devices are highly sensitive to PQ phenomena and also some of the major contributors to reduced levels of PQ in the network. - The increased awareness of customers about the PQ disturbances and their impact on their equipment. Customers start to challenge the utilities about the PQ performance. - Regulatory bodies start to pressure the utilities to comply with the PQ standards and sometimes penalties and rewards are applied based on the PQ performance. In addition, the deregulation of the electricity market has made the PQ issues 21

22 even more complicated, where there is no direct control from the suppliers to the end users of the electricity flow. The regulatory bodies have to pay more attention to the PQ disturbances and the responsibilities of all the parties involved. - The increased interconnections and automation of processes in the industry lead to more significant consequences in case of the failure of equipment or process interruptions. - The increased penetration of the distributed generation (DG), especially the intermittent, power electronic interfaced generators which have a significant impact on the PQ performance. - The increased penetration of electrical vehicles (EV); non-linear loads with higher uncertainty of injections, locations and operation behaviour. - Many indices have been developed to describe the PQ performance of networks. Standards have been revised and published proposing new planning and compatibility levels, and providing recommendations for both utilities and equipment manufacturers regarding the PQ performance and compatibility. Under most of the current regulatory frameworks and codes, utilities are at least obliged to monitor and quantify some of the PQ disturbances in their networks. - Regardless of the level of competition in the electricity markets, some utilities want and are committed to deliver good power quality. This will not only improve the image of such utilities but will also lead to the more efficient operation of assets. Reduced losses, increased life expectancy of components and deferred investments come as bi-products of improved PQ in networks. - PQ has also attracted attention because it can now be measured. Advanced cost effective monitoring technology is now affordable for all stakeholders at all 22

23 voltage levels. Moreover, long terms measurements campaign results are now available for comparisons and more significant statistical analysis. However, the real challenge now is the handling and analysis of such a huge amount of data to produce relevant conclusions and practical solutions. Most of the identified PQ challenges, if not all, can be addressed with straightforward solutions. The two main solutions of PQ disturbances are increasing the immunity of sensitive devices in the manufacturing stage and the elimination of PQ disturbance sources by means of improving the existing components or adding extra devices to the power systems. However, the remaining questions in this matter must be addressed. Who is responsible for improving the PQ and who should pay for the extra cost of enhanced PQ? Is it feasible to have completely immune devices or completely disturbance free power systems, and how can the trade-off be decided? If not, how should the compatibility levels between accepted equipment design and normal network operations be determined? Addressing these questions is not trivial or straightforward. It requires high levels of collaboration and understanding from all stakeholders in the power system industry. Even though for such problems it is extremely difficult to have general models and a case-by-case analysis would still be required for certain customers and networks, general agreement between all parties regarding PQ issues is highly recommended and must be pushed forward. This can be performed by means of customers and utilities surveys, working groups that represent both research institutes and industries, the global standardization of manufacturing and the transparent publication of equipment tests and network disturbances records. Figure 1-1 (a) shows the survey results of the most common PQ disturbances in the US and (b) the results of the disturbances with the highest economic impacts in the EU-25 countries [12]. Based on this survey, voltage sags/swells are the most common PQ issue in the US (48% of respondents) and the transients are the 23

24 highest impacting phenomenon based on the European survey (29% of respondents). Nevertheless, voltage dips (voltage sag in the US terminology) are also considered highly impacting phenomenon based on the European survey (23.6% of respondents). Note that the colours in the figure are representing the rank, and not a certain phenomenon (green represents the highest ratio). (a) Common PQ issues in the US 1.3 Problem formulation (b) The highest impacting (economically) PQ disturbances in the EU-25 Figure 1-1: PQ disturbances survey results in Europe and the US [12] This thesis addresses the problem of developing a methodology for the provision of differentiated Power Quality. Different customers with different PQ requirements are 24

25 considered. Increased levels of distributed generation (DG) penetration, in particular intermittent power electronics interfaced DG, are also considered. The PQ phenomena under consideration include harmonics, voltage unbalance and voltage sags as those may vary depending, not only on activities of different industrial customers but also, and possibly significantly, on the presence and operation of stochastic and intermittent power electronic interfaced DG in the network. Simulations in power system software packages (DIgSILENT and OpenDSS) are used to generate sets of data representing variation in power quality indices, in addition to data obtained from real measurements, caused by different network operation conditions and different penetration levels of different types of DG. This is performed to establish the expected PQ levels in current and future distribution networks. The PQ levels data are established probabilistically and compressed in unified indices to cater for the increased levels of uncertainty in network performance and to facilitate the evaluation. Particular attention is paid to temporal and spatial variations of harmonics, influenced by connection time and location of different types of DG and different types of loads (e.g. electric vehicles) in the network. The work includes simulations with and without DG connected to the test networks, after developing appropriate models of DG, new types of loads and PQ mitigation devices. The work also included the evaluation of the harmonics in radial distribution networks with limited monitoring. 1.4 Literature review This section presents an overview of the most relevant literature dealing with topics covered in the thesis. The literature review is presented in two sub-sections relevant to the work presented in the two main chapters of the thesis Chapter 4 and Chapter 5, i.e., the harmonic analysis and the overall PQ evaluation. 25

26 1.4.1 Harmonics modelling and analysis Probabilistic modelling Different models were developed to perform the harmonic analysis of networks in the past. In general, the models used in a harmonic study depend on the purpose of the study, the amount and the period of the data collected and the accuracy level needed [13]. The most common model used involves modelling the non-linearity of the load by a current source for each harmonic frequency. This method yields about 10% accuracy in the calculated voltage distortion [14]. Probabilistic models that can take into account operating mode and multiple switching operation uncertainties were developed in [15]. In the stochastic harmonic load models developed in [16], the uncertainties considered were loading conditions, load compositions and aggregate harmonic load parameters. Statistical measurements of harmonics have been performed and practically applied since 1985 [17]. The authors of [17] emphasized the importance of the statistical analysis of harmonics; they indicated that harmonics are a stochastic phenomenon which requires measurements throughout a sufficient period, normally seven days covering weekends and weekdays. For the power quality assessment for a certain location, it is only logical to perform a statistical comparison between the network performance and the predefined limits or compatibility levels [17]. The IEC , harmonic limit standard for MV and LV system recommends that the evaluation of the emission should be performed statistically, to take into account the time variation of the phenomenon [18]. It is a common practice to take harmonic measurements for at least one week, and to compare the 95 th percentile of the measured THD with the planning levels specified in national codes or international standards. The 95 th percentile is the value below which 95 % of the measurements lie. It is one of the location statistical measures which also includes the information on observation frequency. 26

27 The p th percentile π p of a sample can be found from the probability density function f(x) or the cumulative distribution function F(x) of the sample as shown by equation (1.1). π p p = f(x)dx = F(π p ) (1.1) Alternatively, the percentiles can be readily found from the PDF and CDF plots of a sample, as shown in Figure 1-2 (a) and (b) [19, 20]. p F(x) Density p f(x) Cumulative probability x p p p p x (a) PDF of a sample (b) CDF of the sample Figure 1-2: Percentile values from PDF and CDF plots Based on the harmonic performance benchmark, the utility could get penalized if the limits are exceeded [21]. However, a survey performed by ERGEG with different types of respondents (utilities, academia, research centres) stated that most of the respondents agreed that a 95%-of-time clause should be avoided and the limits should be applied for 100% of the time, to have more efficient and transparent voltage quality limits. The report also indicates that 5% of a week is 8.4 hours, which is a long time to have voltage quality phenomena that exceed standards limits [22]. As per the IEEE 519 standard [14] if the harmonics are fluctuating with time, i.e., the harmonic sources are time dependent, the harmonic analysis must be done over a period of time. The period of study (or data collection) T D must be divided into intervals m and a number of measurements k must be taken in each interval T, (T D =mt) the mean, mean square and standard deviation of the harmonic currents can be expressed as in (1.2), (1.3) and (1.4) respectively [14]; 27

28 k I h,mean = I kh k k I h,mean square = I kh 2 k 1 1 (1.2) (1.3) 2 2 I h = I hmax I hmin (1.4) where I kh is the h th harmonic current at the measurement k and I hmax and I hmin are the maximum and minimum harmonic currents throughout the interval m. In [16] the day was divided into two intervals, day and night time, taking 15 measurements from different types of loads in each interval and representing the results (THD C and THD V ) statistically on histograms and PDFs. The PDFs are useful representations that can show the minimum and maximum values of the results. Another study was performed in [23] taking measurements from two sites for a 6 hour period with 1 minute intervals and again the results (THD C and THD V ) were presented statistically on histograms and CDFs. The advantage of CDFs is that they showed the percentage of the time (from the study period) the THD will be below a certain value; however, the probability of the maximum THD values cannot be shown. It is worth mentioning that the recommended methods for the presentation of the results of harmonic studies/measurements, as per the IEEE standards, are harmonic spectrum tables, harmonic spectrum bar charts, PDF and CDF [14]. All the methods discussed above are effective in performing harmonic analysis in the networks with non-linear loads that can be represented by a daily loading curve or for the study of a certain load with a repetitive cycle and known harmonic spectra. Based on [15], the non-linearity of the loads in the network can vary depending on the rating and the operating conditions. Some loads have steady harmonic performance 28

29 (fluorescent lamps), some loads can vary deterministically (battery chargers), and some can perform randomly (arc welders). Four categories of non-linear loads can be considered in the probabilistic study of harmonics [15]; - constant number of known injections loads; - random number of known injections loads; - constant number of random injections loads; - random number of random injections loads Harmonic estimation Standard limit levels regarding harmonic voltages and total distortion are defined in international standards [14, 18]; however exact pain levels vary between different equipment types. Increased losses, premature aging, communication interference and equipment damage under resonance conditions are some of the well-known impacts of excess harmonics; however most of these effects are difficult to model and many uncertainties are involved in the estimation of customers and utilities losses due to harmonics. Moreover, only a few regulators worldwide enforce harmonic limits or penalize DSOs based on harmonic performance, and the limits are enforced based on connection agreements between the customer (load or generation) and the utility [24]. Thus, the monitoring and the observability of harmonics levels throughout LV and MV distribution feeders are still very limited in most of today s distribution networks. Harmonic State Estimation HSE is the process of establishing the levels of the harmonic voltages in all network buses based on a limited number of harmonic readings. For transmission systems it is a well-developed process [25-28]. Taking advantage of the high number of monitored buses, symmetry of the network, the meshed topology and the a priori knowledge of most probable harmonic sources location, the HSE for transmission 29

30 systems can give accurate estimates of harmonic levels and can pinpoint the harmonic sources by applying proven estimation mathematical models like the Weighted Least Squares WLS technique [29]. On the other hand, the HSE for distribution networks is still developing. The limited monitoring infra-structure, in addition to distribution network characteristic (e.g. high R/X) lead to harmonic estimation convergence issues [25]. Recent references have discussed the HSE problem for distribution networks and have suggested solutions like monitoring certain harmonic frequency or utilising synchronised measurements from phasor measurement units PMUs [30-32]. References [11, 33] present a technique to estimate the harmonic levels and to make an educated guess about the most probable location of harmonic sources. This algorithm requires two or three monitors located at nodes that divide the feeder into sections. Based on the fact that harmonic currents flow towards the sub-station [11], the most probable location of non-linear sources can be associated to certain sections of the feeder. Good estimates of overall harmonic levels can be provided. However, in case of large capacitors connected to the network or of a resonance situation, the estimation accuracy needs to be improved [33]. It was also shown in [34] that a high level of knowledge about radial distribution feeders harmonic performance can be achieved based on substation readings, while extra monitors and field measurements can help pinpoint sources and increase the accuracy; they are not always permanently located at distribution feeders [34]. There are well developed techniques for the harmonic analysis of radial distribution feeders to overcome the problem of limited monitoring. Usually an analysis for expected low harmonic frequencies and potential resonance at the feeders are adequate for identifying any unaccepted performance, where further analysis with more accurate 30

31 modelling will be required [35]. The THD capacity estimation is used in distribution planning, as in [36] for calculating the remaining load growth permissible before reaching harmonic limits, or as in [37] for determining the permissible penetration level of distributed generation Power Quality evaluation Overall PQ evaluation Since the early 2000s, several publications have addressed the issue of unified PQ assessment, adopting various mathematical models. Reference [38] is one of the earliest publication discussing the global PQ index. The methodology presented there is based on the calculation of the missing RMS voltage throughout a study period, i.e. comparing the sampled voltage to the ideal voltage and the RMS error (RMSE) was used as an index to compare different PQ solutions. Although this methodology can be computationally extensive, as it involves time domain analysis, it considers both event-type and continuoustype PQ disturbances, like harmonics, voltage sags/swells and transients. Due to the increased concerns about network performance in the future, utilities have started to deploy more PQ (and other) monitors in their networks which will lead to a huge amount of recorded data. This abundance of data, unless properly structured and processed to yield useful information about network performance, will not be useful to the user [39]. Reference [40] suggested a data mining method for structuring and classifying the recorded data, before calculating a global PQ index. References [41, 42] also suggest a multi-level structured framework for PQ data analysis and compression. The framework has two stages: time compression and space compression. In the time compression stage, the raw recorded data for the considered period are classified into several disturbances, then the disturbances are characterised using a suitable number of indices. These indices 31

32 are normalised based on predefined thresholds. The normalised indices are then consolidated into one number for each disturbance. The unification of the consolidated indices (Unified Power Quality Index UPQI) is performed based on the exceedance of the thresholds. The index has a value of unity or below if all disturbances are at or below the limits. The space compression is performed by combining the unified index probabilistically for all sites, feeders, substation or networks. The UPQI can be adopted for different applications including site and network benchmarking, characterisation, compliance verification and development of PQ contracts [41]. The UPQI in [41, 42] is applied to combine continuous-type phenomena only. Several indices for event-type disturbances were also proposed in [43] but without explicitly suggesting a method to apply them to the unified index. These modifications of UPQI were suggested in [44]. The index proposed in [45] is based on the available reserve for a PQ phenomenon at a certain bus, i.e., by applying equation (1.5) below, r = 1 m g (1.5) where r is the remaining reserve of a certain PQ phenomenon, m is the actual level of performance, and g is the threshold. The consolidation in this methodology was performed by taking the minimum reserve as the bus overall score in case of no PQ limits violation, or taking the sum of negative reserves in case of PQ limits violations. Examples of recent applications of similar, global PQ indices using real life PQ monitoring data can be found in [46, 47]. An illustration of distribution networks PQ performance using global PQ index is given in [46] while [47] presents a commercial PQ management system installed in a German industrial park. The PQ index is continuously reported to the system operator. Many suggested global PQ indices are based on the application of fuzzy logic [48-51]. In [48], the overall index of PQ is calculated based on the analytic hierarchy process (AHP) to measure the relative importance of phenomena, then a fuzzy system was applied 32

33 to identify the overall assessment based on the different levels of different phenomena. Reference [49] proposes a fuzzy-wavelet-based index, i.e., all the considered phenomena are redefined in the time-frequency domain before performing the overall assessment. Reference [50] combines the fuzzy systems with the artificial neural networks (ANN), to calculate the different levels of PQ disturbances, continuous or event-type, with the fuzzy systems being responsible for giving different weights to different phenomena based on the consequential cost of the disturbance to end users. In [51], a fuzzy total power quality index is proposed by adopting a fuzzy expert system model. The knowledge base of the model is based on standards, expert opinion and field measurements, i.e., it considers the qualitative and quantitative data in the assessment. References [52-54] presented two methodologies to evaluate the overall PQ performance, considering the costs of the disturbances in the evaluation. The first methodology [52] is a value-based evaluation of PQ. The events costs were classified into three classes, mis-operation, aging and losses. By assigning comprehensive cost values for all PQ disturbances, the methodology will avoid subjective weightings of the different PQ phenomena in the overall evaluation. The suggested main application for the index is in the network planning, as the cost included in the evaluation can give an indication of the feasibility of the suggested PQ reinforcement plans. The second methodology is based on the analytical hierarchy process (AHP) model [53, 54], where the alternatives for the model were the possible, actual and ideal states of the PQ, and the weighting criteria for calculating the global index were power supply inconvenience criterion (sags/interruptions), clear sinusoidal inconvenience criterion (harmonics/voltage deviation) and the disturbance cost criterion (aging/losses). Reference [55] introduced a new application of the PQ global indices to study the variation in PQ due to the connection of distributed generation. 33

34 Differentiated PQ and PQ contracts As a result of the various requirements of different customers to PQ level, the area of providing differentiated PQ and classifying the customers based on their requirement is an on-going research area. As stated in [56] the optimum provision of the PQ is one of the system operators tasks, the regulators should ensure the cost efficiency of the solutions which will be paid by the customers. This is a complex matter, as the customers with similar requirements will not always be in confined geographical zones. These different requirements, and different levels of acceptance of higher costs of power were the reasons behind the adoption of the premium PQ contracts [57]. Sometimes it is not optimal to the DSO to improve the overall system PQ performance, especially when there are few customers with higher requirements and willing to pay extra; the optimal solution then is to address the issue locally and provide this kind of premium PQ contracts. Utilities like EDF (France) and Detroit Edison Company (USA) have developed and adopted these types of contracts since the mid-nineties [57]. The requirements for establishing PQ contracts and agreements depend on the system characteristics and the parties involved in the agreement. Reference [58] summarizes some of the typical areas that should be addressed in PQ contracts in the following points: i) the PQ phenomena and reliability measures to be considered, ii) the indices to be used, iii) the baseline of the performance, iv) penalties/incentives for performance worse/better than the agreed baseline, v) measurements and calculations for performance evaluation, vi) the responsibilities of each party in disturbing the performance and vii) the responsibilities of each party in enhancing the performance. Figure 1-3 shows the arrangements of PQ contracts in deregulated markets (adopted from [58]). 34

35 Distribution system operator Transmission system operator Figure 1-3: PQ Contracts in Deregulated Market (adopted from [58]) The general framework of the differentiated provision of PQ involves two stages: i) Customers are classified based on their requirements to the PQ and ii) the potential consequences of PQ disturbances to their activities are analysed. This refers to the technoeconomic nature of the problem and how customers will be grouped [7]. A methodology was developed in [7] for classifying customers based on their sensitivity to the PQ disturbances correlated with the expected economic losses due to inadequate PQ incidents. For example, if the customer s equipment is very sensitive to PQ disturbances, and the expected economic losses due to the disturbance are high, this customer will be classified as a sensitive customer. Taking NACE (Statistical classification of economic activities in the European Community) as a reference, and using the proposed classification methodology, reference [7] has classified the customers into four categories; sensitive, essential, important and desirable. In each category, further grouping was performed based on the similarity in the sensitive equipment used by different customers, to facilitate the selection of mitigation solutions [7]. Figure 1-4 shows examples of the classification and grouping presented in [7]. Independent power producers Retail marketers/ Energy service companies 35

36 (a) PQ classification of customers based on equipment sensitivity and economic impact (b) Grouping based on similarity in characteristics of sensitive equipment and process Figure 1-4: PQ based Customer Classification (adopted from [7]) It is important to include the economic losses in the classifications of customers, as the feasibility of solutions will be based on the customers willingness to pay to have adequate power supply. The problem that arises here is the inconsistency in the reported losses per the PQ disturbance from different customers, even from the same types of customers. In [6], a tool was developed to estimate the cost per interruption duration based on several reported losses, from different types, sizes and locations of industry plants. Basically, this tool matches the economic activity under study with the various databases available in the tool. Similarities in the activity, location or size will give different weights to the surveyed reported losses and damage functions and as a result a customized damage function to the activity under study will be produced. The weighting factors can also be controlled to make the tool more flexible. For example, for the tool to consider similar size plants losses as the main reference or similar location plants losses as the main reference, 36

37 some control variables are introduced. Figure 1-5 shows the customized damage functions for a plant. As shown in the figure, by taking two extreme weighting factors, i.e., size focused (a=1 b=0) or location focused (a=0 b=1), two customized damage functions confining an area of uncertainty were produced [6]. Figure 1-5: The Customised Damage Function of a plant, (adopted from [6]) Reference [6] also investigates the area of high quality power provision to certain customers. The author summarized the advantages in the following points: - Distributing the mitigation costs between the customers in the same group or PQ zone. - Enabling the low losses plants to share mitigation solutions, which are not economically feasible for individual plants. - Increasing the efficiency of the solution by affording more expensive, but efficient solutions. - Allowing optimal loading management between the grouped customers (e.g. shifting the peak load of different plants) will lead to additional savings in mitigation costs by deploying smaller size mitigation devices. 37

38 The evaluation of the high quality zones values was performed by studying eight customers in a 158 bus UK distribution network (132 kv, 33 kv, 11kV and 6.6 kv). The evaluation was performed by studying the costs of PQ inadequacy and mitigation solutions for individual customers and for the grouped customers. The optimisation problem has the following objective function, S i = min f = min[npv 10 (C mitigation + C sag )] (1.6) S base case = N S i i=1 (1.7) where S i NPV 10 = optimum solution for the assessed plant i = net present value for the ten year assessment period C mitigation = total cost of mitigation, including initial, operation and maintenance cost C sag N = remaining financial loss after mitigation = total number of plants S base case = total financial loss of all plants (applying individual solutions) The base case is the selection of mitigation for individual customers without considering the grouping in a PQ zone. The same optimisation problem as in (1.6) was used to calculate the cost after grouping, but the C mitigation is the cost of the mitigation solution for the zone rather than the individual solution. There were three mitigation strategies considered in the study: i) mitigation devices, i.e. power injection devices; ii) alternative supply feeder and iii) a combination of mitigation devices and alternative supply. The relocation of customers was not considered an option, neither geographically or electrically (supplying from a far substation) as the relocation cost will not make the quality zones solution feasible. Instead, the high PQ zone was selected with an optimisation technique, i.e. to ensure that all the power system constraints were met (voltage limits, flow limits, ) and these eight customers were selected for the high PQ 38

39 zone. The mitigation solutions in addition to optimum load management were applied, and the value of creating a high PQ zone was calculated. The conclusions of the study can be summarised in the following points; i) the high PQ zone was economically feasible; however, some plants will not gain significant benefit in the optimum solution; ii) the optimum loading management will increase the value of high PQ zone significantly; and iii) the mitigation solution of redundant supply will have a clear effect in mitigating the sag events that propagate from the transmission level, which will increase the value of the high PQ zone for those customers [6]. The concept of differentiated PQ provision is relatively new and has not been applied in a system scale yet. Therefore, the current market and regulation models might not fully accommodate the concept and some challenges may arise. Demarcating the PQ responsibilities (disturbing or enhancing) of different parties is a complex problem. Moreover, finding optimum PQ enhancement solutions with a positive cost-benefit analysis is not always possible. References [58, 59] discuss some of the challenges of PQ provision in deregulated markets. On the other hand, most of the current regulatory frameworks have some constraints and challenges in terms of the application of the differentiated PQ provision. From the Council of European Energy Regulators (CEER) report [10] the following relevant statements can be extracted: a- The ultimate aim of voltage quality regulation is to ensure that the functioning of equipment is not impacted by voltage disturbances coming from the network. b- At European level, the 3 rd Package Directive 2009/72/EC [27], which had to be transposed by Member States by 3 March 2011, states that the regulatory authority shall have the duty of setting or approving standards and requirements for quality 39

40 of supply or contributing thereto together with other competent authorities (Article 37(1h)). c- In a liberalised electricity market, the customer concludes either a single contract with the supplier (SP) or separate contracts with the supplier and the distribution system operator (DSO), according to the existing national regulations. As per statement a, the PQ is a customer oriented issue. This is the main reason for studying and modelling the PQ requirements of different customers. According to statement b, it is the regulator s responsibility to set the limits of different phenomena. From the same report and a survey performed by CEER and ERGEG [22], it is a common practice to adopt the international recognized standards (IEC/IEEE) in addition to the well-established EN [60], to define compatibility levels for different phenomena. This should ensure the immunity of the typical customer s equipment, which is mainly defined based on the mentioned standards as well. Some regulators set higher standards, and also some DSOs set higher planning levels to attract more sensitive customers. As per statement c, the customers and DSOs are allowed to adopt contractual agreements of certain levels PQ, commonly referred to as Premium Power Quality contracts. However, in the countries where the national regulators prohibit a direct contact between DSOs and customers, the role of aggregators, ancillary service providers, or technical virtual power plants (TVPP) owners can be investigated to perform the provision of premium PQ. From the above review, the main concerns about regulatory frameworks and market constraints can be summarized as follows: - The conflicts of PQ requirements: when some customers do not require a high quality but are geographically located in a premium zone. The regulators may not allow the enforcements/solutions costs to be passed on to these customers. 40

41 - The share of responsibilities in causing PQ disturbances: should the regulators allow the penalties to be passed on to the customers responsible for the disturbances? There is no commonly accepted and accurate way for deciding these responsibilities in cases of limited PQ observability which is often the case in practice. - A standard way to represent the PQ: a standard way to evaluate the PQ and estimate the PQ losses must be agreed upon between regulators, utilities and customers Distributed Generation (DG) impact on PQ The increased DG penetration in today s distribution networks has different impacts on the networks. Some questions may arise here [11], will the DG connection have a positive or negative impact? Which PQ phenomena are impacted? What penetration level is significant? These, and other unanswered questions, increase the importance of studying the overlap between the PQ performance and the DG presence. The recent increased penetration of DG, especially the renewable type, is due to the various benefits that different parties in the power industry can gain. Regulators incentivise the connection of renewable generators, with schemes such as feed-in-tariff, to achieve clean energy generation targets. Customers benefit financially from selling the energy to the utility and also from the increased reliability of having the DG as back up sources. Utilities also benefit from the relief on their assets as a result of several feeding points instead of the traditional source-to-drain flow of power, which sometimes helps to defer investment in the network reinforcement plans. Regardless of all these benefits, many concerns are under investigation regarding the increased penetration of DG. The main concern is the impact on the PQ performance. Phenomena like voltage regulation, voltage 41

42 unbalance and harmonics will be affected by the connection of DG, effects that were not considered during the planning stage of distribution networks. Some concerns of the utilities in the case of high DG penetration are substation reverse injections, the need for new protection schemes and the hosting capacity of the assets. The concerns of the regulators are the price of electricity, the favouritism towards renewable generations, and the concerns of the owners are the operation and maintenance costs of DG units. The PQ and other concerns must be investigated and adequately addressed before wide deployment of DG [11]. Different types of DG can have different impacts on PQ phenomena. The intermittent photovoltaic (PV) and wind generators can lead to voltage fluctuations. Connection interfaces like power electronic converters can increase the harmonics level in the network. The DG can cause overvoltage during light load periods but also can support the voltage during sag events. Different penetration levels and locations have different impacts. In general, a penetration level up to 15% of the feeder capacity will not lead to significant negative impacts. This ratio of capacity drops to 5% in long rural feeders, and can go up to 30% of the feeder capacity when the DG connections are close to the substation [11]. The DG penetration level effects, on the sag only, were studied in [61] where the close correlation of the location and level of penetration of DG to the sag performance was demonstrated. These results were illustrated in a single radial feeder and in large scale distribution network (287 bus network) [61]. The DG penetration level impact on the harmonic performance is studied in [62]. The authors introduced the concept of hosting capacity based on PQ performance. The hosting capacity was defined as the maximum amount of DG at a specific point that can be 42

43 supported by the network without PQ limits violations. Reference [62] presented generic equations correlating the level of DG penetration with the expected increase in the harmonic distortion, as presented in Figure 1-6. The impact of the DG on the harmonic performance was also studied in [63] focusing on the interactions between DG invertor emissions and the grid background harmonic pollutions. Similarly, [64] investigated the impact of wind turbines on the harmonic performance. Particular attention was given to the cases where the wind generators are connected with power factor correction capacitors and cause resonance problems at the point of connection. The impact of the PV and storages connections on a distribution feeder was studied in [65]. The phenomena of voltage regulation, unbalance and harmonics were analysed under different levels of DG penetration. Figure 1-6: Harmonic distortion vs DG penetration generic dependency (adopted from [62]) Literature review summary The modelling and analysis of the harmonics are well discussed and presented in the literature [13, 15, 23]. The available probabilistic models and statistical representation of the harmonic indices are well established and recommended by the international standards [14, 18, 60]. Nevertheless, the introduction of smart grid technologies, new types of loads and DG will add greater uncertainty to the evaluation and will require longer periods of study with more robust probabilistic models. The variability in DG output (in 43

44 particular the stochastic intermittent types like PV and wind) will increase the networks harmonic performance variation. The shift from the common weekly assessment towards long period evaluations is inevitable. The harmonic estimation based on a limited number of monitors is still an ongoing research area, especially for distribution networks [25, 66]. The well-developed techniques for the radial feeders harmonic analysis with the aid of a few monitor readings can provide practical solutions for the estimation problem [11, 33, 35]. Depending on the application of the algorithms, different levels of accuracy in the harmonic estimation can be accepted. The main goal regarding PQ can be the identification of the remaining THD capacity in a feeder to host higher DG levels and disturbing load growth [36, 37]. Since 2000, a significant number of publications has addressed the unification of the performances different PQ phenomena in one index [35-52]. These global indices have been developed for various applications. However, there is still no standard or widely accepted method between stakeholders to evaluate the PQ comprehensively, for a bus or a feeder. Very recently though, some utilities have already adopted a global index in their network operation evaluation, and another global PQ index is commercially developed [46, 47]. It can therefore be concluded that the use and application of global PQ indices are now acceptable in the industry, and the pursuit of further research in the area can be encouraged. The concept of customer classification based on PQ requirements is thoroughly investigated in [6, 7]. The PQ contracts and agreements have been applied since the midnineties [1, 11, 57]. However, these types of contracts have remained largely one-to-one contracts and have not been investigated at the system level when considering different PQ requirements by different groups of customers. Some of the barriers and limitations in the 44

45 current regulatory frameworks and energy market models have to be addressed for the true potential of the concept to be realized. 1.5 Aim and objectives The PQ levels can vary significantly, temporally and spatially, in power networks. Moreover, the impacts of these PQ levels also vary between buses, based on the types and equipment of load connected. The overall aim of the research presented in this thesis is to investigate the concept of the differentiated provision of PQ in MV distribution networks. This requires establishing the levels of considered PQ phenomena based on limited monitoring and high operational uncertainties. The suggested PQ solutions must be directly correlated to the different levels of PQ requirement at different locations. This aim can be achieved by performing the following objectives: 1- Collecting the information on the currently available approaches and technology for the PQ monitoring and measurements in the MV distribution networks through surveys and from the literature. 2- Building detailed distribution network models, including the models of the relevant network components for the PQ studies. Increased levels of DG penetration and new types of loads (e.g. electric vehicles EV) must be considered in the modelling. 3- Evaluating the PQ performance statistically, assuming both full and limited observability of the system, for short (e.g. 1 day) and long (e.g. 1 year) study periods. 4- Presenting the results statistically to quantify the uncertainty in the evaluated performance (PDFs and CDFs are adopted). 45

46 5- Improving visualization of the results so that PQ levels in networks can be captured more easily and efficiently (heat maps are adopted in the thesis). 6- Compressing the amount of simulated (or measured) PQ data into fewer indices that can be handled more efficiently. 7- Investigating the concept of providing differentiated PQ based on variable PQ thresholds of different load types. The economic impact on end users and the zonal effects of PQ mitigation solutions must be considered. 1.6 Contributions This research contributes to the area of the PQ evaluation, estimation and provision. The following points summarize the main contributions of the thesis with the numbers given in parentheses corresponding to the relevant publication where these results are presented (Appendix E provides the full list of thesis based publications): 1- A comprehensive evaluation of PQ in distribution networks with renewable energy sources (RES) was performed. For that purpose, two distribution network models (a generic and a real network) were constructed in two simulation software environments to validate the results. A number of case studies with different levels of DG and power-electronic interfaced loads are produced. Real UK and Portugal weather data for one year were incorporated in the models. PQ visualization of the networks was applied with the aid of heat maps for the easy identification of both poor performing and PQ sensitive areas [E4, E5, E11]. 2- New probabilistic sag indices were developed to evaluate the networks sag performance. The new indices consider the uncertainty in equipment trip due to different sag severities. The sag performance of a bus is evaluated based on the 46

47 expected number of sags, the severity of each sag and the probability of main protection failure [E1, E2]. 3- A framework for probabilistic harmonic analysis was developed based on studies performed over long study periods to evaluate the annual harmonic performance of distribution network, especially under increased DG and EV deployment. A number of uncertainties was considered in the probabilistic analysis to cater for the expected increasing temporal and spatial variations in the phenomenon [E4, E5]. 4- A simple and computationally efficient harmonic estimation methodology for radial distribution feeders was developed. The methodology is based on substation only monitors and basic knowledge of network parameters [E10]. 5- Two new PQ global indices were proposed to compress the amount of PQ data recorded for separate phenomena into one index. Different levels of flexibility were examined in the overall evaluation of PQ, such as different level of importance of phenomena or/and buses. A comparison between the performance of proposed global indices and their flexibility for describing overall PQ performance of the network is performed and illustrated using actual, long term PQ monitoring data from real distribution network [E6, E8, E9, E10]. 6- Based on the global PQ index, an optimization technique was applied to place appropriate PQ mitigation devices in the test network, in order to improve the overall PQ performance of the network, taking into account the different levels of PQ requirement at different locations in the network [E3, E7, E12]. 47

48 1.7 Thesis Overview This thesis is organised in seven chapters. In addition to this chapter, Chapter 1- Introduction, the contents of the remaining six chapters are summarized below: Chapter 2 Power Quality Categorization In this chapter, further discussions about PQ in general and specific phenomena of interest to this research are presented. The three considered PQ phenomena are voltage sag, harmonics and voltage unbalance. The phenomena are discussed in terms of relevant standards, impacts and mitigation solutions. Chapter 3 Power System Modelling for PQ Studies In this chapter, the modelling of two distribution networks and a number of case studies is presented. The adopted indices, models and simulation scenarios for the different PQ phenomena are presented. Some simplified characteristic studies for PQ performance are presented in this chapter as well. Chapter 4 Harmonic Estimation in Distribution Networks In this chapter, the harmonics phenomenon is further discussed. Probabilistic analysis and estimation of harmonics are also presented. The results of the harmonic performance evaluation using newly proposed methodology are presented and illustrated using different visualization techniques. Chapter 5 Global PQ Evaluation Indices This chapter presents the two newly developed global PQ indices. The methodologies and mathematical background about the Compound Bus PQ Index (CBPQI) and the PQ Reserve (PQR) index are provided. The applications and comparisons of the two indices are illustrated using simulation results for the MV networks and real measurements data for LV sites. 48

49 Chapter 6 Optimum Provision of Differentiated PQ In this chapter, the CBPQI, developed in Chapter 5, is applied for optimizing PQ mitigation solutions. The mathematical model of the optimization problem is briefly discussed. The results of the optimization based on a global PQ index are validated based on the combined and individual evaluations of the phenomena performances before and after mitigation. Chapter 7 Conclusions and Future Work Major conclusions of the research are presented in the first part of the chapter. Finally, the future work and areas for potential improvements as well as recommendations regarding the applicability and scalability of the research results are given in the second part of this chapter. 49

50 2 Power Quality Categorisation As discussed in Chapter 1, there is no single standard definition of the term Power Quality (PQ). Different points of views and understanding of the PQ issues have led to different definitions. Some utilities define it as the reliability of the system, so as long as the system is highly reliable they are providing high PQ to their customers. On the other hand, manufacturers or end-users of equipment define PQ as the power characteristics of the supply that are required to ensure that the equipment works properly and efficiently [11]. In general, PQ is an issue driven by the customers requirements, and could be very broadly defined as Any power problem manifested in voltage, current, or frequency deviations that results in failure or misoperation of customer equipment [11]. The Council of European Energy Regulators (CEER) considers the assessment of the impact on customers as the main reason for monitoring and regulating PQ, The ultimate aim of voltage quality regulation is to ensure that the functioning of equipment is not impacted by voltage disturbances coming from the network as it was stated in [10]. Therefore, when 50

51 analysing PQ issues the main focus should be on compatibility between the customers equipment and different PQ phenomena, e.g., voltage sags, unbalance, harmonics, flicker, transients, overvoltages etc. In addition, an agreement upon the causes and responsibilities between the parties is yet to be reached. The results of customers and utilities surveys about the causes of PQ are shown in [11]. The results showed similarities between customers and utilities who believed that natural causes are the main cause of PQ problems; however, the results showed different opinions when it came to who is more responsible, the utility or the customers, for causing PQ problems [11]. 2.1 PQ standards and guides The increased concern about PQ has led to the revision of the existing standards and the publishing of new standards. The EN Voltage Characteristic of Electricity supplied by public distribution systems [60] is the main standard used by European utilities to benchmark their PQ performance, although it is more related to the voltage characteristics than general PQ performance. Similarly, the IEC Testing and Measurement Techniques - Power Quality Measurements Method standard [2], is more specialised in the PQ phenomena measurements while the IEEE P1159:2009 Recommended Practice for Monitoring Electric Power Quality standard specifies the terms and definitions used by different technical parties to describe electromagnetic phenomena affecting the PQ performance, which yield better information exchange from different sites monitoring and networks parties. There are also separate standards which discuss the network compatibility level to different phenomena, for example, harmonic performance planning and compatibility limits [14, 18], the voltage sag sources, effects and method of measurements [67], the 51

52 immunity tests for equipment against the voltage sags and short interruptions [68], methods for calculating voltage sag indices and characteristics of the sag events [69], voltage sag compatibility between equipment and power systems with the emphasis on financial and technical analysis for the sag phenomenon [70] and assessment of the emission limits from three-phase unbalanced connections (generators or loads) [71]. The IEEE 1547 standard, in its different parts and guides, discusses the connection of distributed energy resources (DER) with emphasis, for example, on monitoring, information exchange and control of distributed resources (in IEEE ) and on conducting distribution impact studies for DER interconnection that include the impact on some PQ phenomena (in IEEE ). Also, the IEC Photovoltaic (PV) systems Characteristics of the utility interface discusses the harmonic requirements for PV when connected to the networks. Table 2-1 summarizes the relevant PQ standards for the PQ phenomena considered in this thesis. Table 2-1: Summary of relevant PQ standards (adopted from [66]) Standard IEC IEEE/ANSI EN PQ Monitoring IEC IEEE Std 1159 IEC IEC Voltage sag IEC IEEE Std 1564 EN IEEE Std 1159 IEEE Std 1250 IEEE Std 1346 Harmonics IEC IEEE Std 519 EN IEC IEC IEC IEC Unbalance IEC ANSI Std C84.1 EN IEEE Std 241 Flicker IEC IEC IEC IEC IEEE Std 519 IEEE Std 141 EN EN

53 Two levels of PQ performance are usually determined and compared to standards; the planning levels and the compatibility levels [18]. The planning levels, as the name implies, are usually for planning purposes; the utility set these limits as internal targets and objectives considering the supply and consumers emissions (for a certain phenomenon). The planning levels are lower or equal to the compatibility levels, and they are difficult to determine as they vary from one network to another based on the structure, location, importance, etc. [18]. On the other hand, the compatibility levels are the reference values to set the immunity of the system equipment (network and users) to ensure the electromagnetic compatibility between the emission levels and equipment connected [18]. The compatibility levels are usually assessed for the whole system statistically (95 % probability levels) considering a distribution that covers the temporal and spatial variations of the phenomenon [18]. Figure 2-1 (adopted from [72]) illustrates the concept of the planning and compatibility levels of the PQ events for a site. The same presentation of results is recommended by [18], with inclusion of all the sites data (or 95 % of the sites) of the system under study in one results sample. Figure 2-1: The planning and compatibility levels of a PQ phenomenon (adopted from [72]) 53

54 The most recent work in PQ standardization is concerned with the increased penetration of DER and power electronics equipment in the network. For example, an amendment 1547a-2014 to the IEEE standard is published to respond to widely expressed requested changes in voltage, frequency and voltage regulation requirements for DER connection. On the other hand, the guide : IEEE Guide for Application of Power Electronics for Power Quality Improvement on Distribution Systems Rated 1 kv Through 38 kv presents the technology of customer power devices to improve the PQ performance of distribution networks. Devices like dynamic voltage restorer (DVR), static voltage regulator (SVR) and transfer switches (TS) are investigated for sag, swell and interruption mitigation. Devices like distribution static compensator (distribution STATCOM) and distribution static var compensator (SVC) are investigated for harmonics and reactive compensation [73]. 2.2 PQ monitoring The question related to the number of measurements, types and locations of monitors for accurate PQ estimation is a complex multi-parameter optimisation problem (affected by different topologies, monitor types, monitored phenomena and reasons for PQ monitoring). In terms of best practice for PQ monitoring a Technical Brochure 596 recently (2014) published by CIGRE/CIRED Joint Working Group C4.112 Guidelines for Power quality monitoring measurement locations, processing and presentation of data [66] provides the most recent and relevant reference to the current industrial practice of both distribution and transmission utilities. The technical brochure (TB) [66] identifies six PQ monitoring objectives: Compliance Verification, Performance Analysis/Benchmarking, Site Characterisation, Troubleshooting, Advanced applications and studies and Active PQ management. For each of these objectives it provides recommendations and guidelines for choosing monitoring locations and the number of monitors needed to get sufficiently 54

55 accurate information on PQ. The trade-off between the costs of monitoring and the amount of information provided is discussed and PQ monitoring needs, research results and trends for reliable estimation of relevant PQ indices at non-monitored locations are identified. It then develops a set of recommendations on the parameters to be recorded, how it should be undertaken and for how long the monitoring should last. Recommendations are also made with respect to sampling rate, averaging window, telecommunication and data handling infrastructure and types of instrument transformers to be used. Finally the TB discusses reporting methodologies for PQ data and methods of producing indices which reduce the large volumes of data collected by PQ monitoring to a limited number of indicators that can be used to characterise either a single site or multiple sites [66]. The current trend in Europe and worldwide is to deploy more PQ monitors in the networks. The IEC [2] describes the standard methodologies of monitoring different PQ phenomena, the classes of monitors, the transducers and other related factors to standardize the process of PQ monitoring between different stakeholders. The increased pressure from regulators and customers on DSOs to provide PQ information, also the development in the enabling technology for monitoring (low cost monitors, developed communications and data storages etc) make it easier to deploy more monitors. Two recent survey results (2013/2014) about the current infra-structure of PQ monitors are shown here. The first survey covers 15 DSOs in Europe, and was published in Sustainable D2.1: Survey and reporting of state of the art technologies and services used by DSOs in Europe [74], Figure 2-2 shows the relevant results about the PQ monitoring. As shown in the figure, out of the 15 surveyed European DSOs, ten have deployed or are planning to deploy PQ monitoring in their networks. The other survey results were adopted from [66]. The survey covered 114 participants among transmission system operators (TSO) and DSOs from all over the world. Figure 2-3 shows the relevant results of the monitored PQ 55

56 phenomena in DSOs networks. As shown in the figure, out of the 73 DSO respondents, more than 70 % do monitor main PQ phenomena like voltage sag, unbalance and harmonics. Figure 2-2: PQ monitoring from DSOs survey (adopted from [74]) Figure 2-3: Survey results about PQ monitoring (adopted from [66]) PQ estimation As discussed in [2, 11, 66, 75] the PQ monitoring requires certain types of transducers with different classes for different applications. The PQ monitoring also requires the continuous monitoring of the current and voltage signals to capture events and continuous type PQ phenomena. Furthermore, monitoring some of the PQ phenomena s 56

57 parameters requires monitors with special capabilities, e.g., high sampling rates (windows as small as ½ a cycle), higher frequency capabilities and voltage and current angle information. Therefore, although the cost has dropped and technology of monitoring systems improved significantly, it is still neither feasible nor practical to monitor every bus in the network. The adopted solution to the limited monitoring of the distribution network is the PQ state estimation, where mathematical models and knowledge of network parameters are utilized alongside a number of available measurements to estimate the performance in the rest of the network. To present examples of the PQ estimation techniques, the voltage sag, unbalance and harmonic estimations are briefly discussed here. Estimating the voltage sag (caused by faults) for the non-monitored bus can be divided into two main parts. First, the fault events from the recorded measurements should be identified and the fault based on the different monitors readings should then be located. Secondly, the voltage sag caused by that fault on the remaining buses should be calculated, i.e. the area of impact of the fault. Mathematical models for fault localisation are proposed in many research studies, yet fault localisation is not an easy task to perform in practical networks. For example, if the system is not fully observable, the mathematical models will not reach a unique solution and so more than one fault location could be proposed by the models for the same set of measurements [72, 76]. For the voltage unbalance estimation, the models can be based on the traditional distribution networks state estimation. By applying a three phase state estimation and utilising more pseudo measurements, to overcome the limited monitoring problems in distribution networks, an acceptable level of accuracy in the unbalance estimation can be achieved. A number of mathematical models is developed to solve the state estimation problem; for example, the Weighted Least Squares WLS is one of the most widely used 57

58 models. The WLS is an optimization problem where the error between the readings and estimated values is the objective function to be minimized, based on calculated voltages and angles. The monitor readings in the mathematical model are weighted based on their noise level [72, 77, 78]. For the harmonic phenomenon, the main aim of the harmonic state estimation is to establish the levels of voltages at different frequencies at non-monitored buses and if possible to identify the sources of the harmonic currents. Figure 2-4 [66] shows the general framework of harmonic state estimation, where information like network topology (especially capacitors switching states) and the system frequency models are critical for accurate harmonic levels estimation. 2.3 PQ Phenomena Figure 2-4: Framework of harmonic state estimation (adopted from [66]) Similar to the PQ definitions, an agreement upon the categories of the PQ phenomena is yet to be reached. As mentioned before, some utilities consider the reliability of supply and voltage regulation as the only PQ phenomena to be measured and improved. Some regulators also incentivise and penalize utilities based on the reliability and interruptions of customers only. However, a common trend now is to monitor and measure 58

59 different electromagnetic phenomena under the umbrella of PQ, and some regulators have started to apply the PQ standards. CEER report [10] and Cigre brochure [66] showed the increasing adoption of monitoring different PQ phenomena in European countries, though the CEER report distinguishes between the continuity of supply and voltage quality. The PQ phenomena variation can be divided into two major categories; disturbances and steady state variation [11]. The disturbances include phenomena like the transients, voltage sags and swells or interruption of supply. These phenomena cannot be presented in a continuous manner as they occur only momentarily during long periods of time. Basically, the PQ performance of such phenomena is based on the count of occurrences over a relatively long period of time. On the other hand, the steady state variation category includes phenomena like voltage regulation, harmonics distortion and voltage flicker. These types of phenomena could persist for a long time; therefore continuous monitoring is needed. The system or bus PQ performance with respect to these phenomena can be measured and benchmarked for very short periods of time (seconds). As shown in [11] the PQ phenomena can be also divided into more categories; i) The transients category, which includes the impulsive and the oscillatory phenomena; ii) The short-duration variations category which includes the interruptions, sag and swells. This category can be further divided based on duration; i.e. instantaneous (cycles), momentary (seconds) or temporary (up to 1 minuet). iii) The long-duration variations which includes sustained interruptions, undervoltages and overvoltages. iv) The steady state phenomena including the voltage unbalance and harmonics. 59

60 Sometimes additional characteristics are needed to classify the measurements and fully describe electromagnetic phenomena that lead to PQ problems [11]. Attributes like amplitude, frequency, spectrum, and source impedance are required to describe steadystate phenomena, and attributes like rate of rise, duration, rate of occurrence are needed to describe non-steady-state phenomena. Figure 2-5 [79] shows an overview of some of the PQ phenomena definitions, illustrated in a voltage RMS plot. Figure 2-5: Overview of some PQ phenomena definitions (adopted from [79]) Regardless of the variety of PQ phenomena, general steps can be applied in any PQ evaluation process. Based on the available information and measurements, identifying the causes of disturbances and available solutions are the two steps in common to all PQ evaluation methodologies. Interaction between customers and utilities is critical in this PQ evaluation process. The general evaluation process can be summarised in Figure 2-6 (adopted from [11]). The considered PQ phenomena in the thesis are the voltage sag, harmonics and unbalance. They are further discussed in the following sub-sections, and the 60

61 general PQ evaluation process is adopted to different extents, as presented in Chapter 4 and Chapter 5, for the evaluation of these phenomena. POWER QUALITY PROBLEM EVALUATIONS IDENTIFY PROBLEM CATEGORY Voltage Regulation/ Unbalance Voltage Sags/ interruptions Flicker Transients Harmonic Distortion PROBLEM CHARACTERIZATION Measurements / Data Collection Causes Characteristic Equipment Impacts IDENTIFY RANGE OF SOLUTIONS Utility Transmission System Utility Distribution System End-Use Customer Interface End-Use Customer System Equipment Design/ Specification s EVALUATE SOLUTIONS Modelling/ Analysis Procedures Evaluate Technical Alternatives OPTIMUM SOLUTION Evaluate Economics of Possible Solutions Figure 2-6: General PQ evaluation process (adopted from [11]) Harmonics Harmonics are sinusoidal voltages or currents with frequencies which are integer multiples of the fundamental frequency in the network [80]. The main causes of harmonics are: i) Saturable devices (due to: physical characteristics of the iron core): transformers, rotating machines, non-linear reactors. ii) Arcing devices (due to: physical characteristics of the electric arc) (They can produce harmonic currents 20% of their rating): furnaces, welders (also cause transients and phase imbalance), fluorescent lighting (about 50% of a modern building s load). 61

62 iii) Power electronics (due to: semiconductor device switching which occurs within a single cycle of the power system fundamental frequency.) (They can produce harmonic currents 20% of their rating): VSD, DC motor drives, electronic power supplies, rectifiers, inverters, SVCs, HVDC transmission. The first two reasons are the conventional sources of harmonic injections; they are well studied, modelled and mitigated in power systems. However the third source can be considered the main reason for renewed interest in the harmonic phenomenon and the need for new models and mitigation techniques. The power electronic interfaced loads and generators involve high level of uncertainties in both general performance and harmonic emission. Output level, location (EV and intermittent DG) and switching frequency are examples of variable factors that might affect the harmonic performance of networks and the selection of mitigation techniques Harmonic evaluation and indices The general methodology for calculating the system harmonics indices can be divided into three steps: calculating the different spectra of the voltage and currents over a window of time, calculating the required indices from the spectra for different sites, and then calculating the total system indices from the sites indices. Several indices are developed to describe the harmonics phenomenon; the most common indices are the Total Harmonic Distortion (THD) for the voltages and currents and can be calculated by (2.1) and (2.2), THD V = 2 h=2 V h V 1 (2.1) THD I = 2 h=2 I h I 1 (2.2) 62

63 where h is the harmonic number, V h and I h are the harmonics voltages and currents, and V 1 and I 1 are the fundamental voltage and current. Other harmonic indices are developed for more specific applications; the Total Demand Distortion TDD was developed to describe the harmonic performance in the case of low fundamental current, where the THD could be misleading. The telephone interference factor was also developed to describe the harmonic performance when it affects the audio and communication system (high harmonic orders). The K-Factor index in the USA (similar to Factor-K in Europe) to describe the de-rating of transformers under the harmonic presence, and also the number of zero crossing per fundamental cycle to describe the impact of harmonics on the equipment that works on the concept of waveform changing from positive to negative or vice versa (e.g. contactors and electronic clocks). The main steps in the harmonic evaluation process are the following [56]: - Calculating the spectra over a 10 cycle (200 ms) period (for 50 Hz system), with consideration for synchronizing the window with the actual frequency. - The RMS values of the obtained spectra are then aggregated over a 3-second period (15 samples) to create the very short time index (V h,vs ). - The 3-second RMS values are then aggregated into 10-minute values (200 samples), which gives the short time index (V h,sh ). - The short time and very short time indices are evaluated over longer periods of study, typically a day or a week. The calculated indices are described statistically; i.e., the 95% or 99% of the sample or the maximum value can be used to describe the site performance with comparison to standards. Table 2-2 gives a comparison between the limits of the harmonic voltage indices between different standards and between the compatibility and planning level in the IEC standards (adopted from [56]). 63

64 Table 2-2: Comparison of the harmonic voltage limits between different standards and guidelines (adopted from [56]) Harmonic order (h) Comp. levels (IEC ) (EN 50160[60]) 64 Planning levels (IEC :1996 [18]) Voltage limits (IEEE 519:1992 [81]) Odd non mult. 2.27(17/h) n/a 0.2+(25/h) 3 Of 3 > THD Consequences of inadequate harmonic levels The ultimate consequences of persisting harmonics in the network are reduced life time (accelerated aging effectively) of equipment and increased system losses. The consequences of harmonics can be categorized as: a) thermal stress (caused by an increase in copper, iron and dielectric losses), b) insulation stress (caused by an increase in peak voltage - voltage crest factor), c) load disruption (e.g., due to mal-operation of protection systems, contactors and numerically controlled process relying on number of waveform zero-crossings). The thermal and insulation stresses could lead to significant damage or

65 even complete destruction of equipment under resonance situations, particularly the capacitors. In [82] the complete destruction of 13.8 kv switchgear in a paper mill is documented. Switching a capacitor bank under resonance conditions led to high switching transient and breaker failure which in turn caused fire at the switchgear. To mitigate the problem, detuning reactors were added to the capacitor bank and a protection relay capable of sensing harmonic currents was added to the switchgear. This increased the total investment, operation and maintenance cost of the power factor correction system. Another case was documented in [83], describing insulation failure of a newly commissioned backto-back converter tie. Multiple costly measures like changing cables and terminations were performed due to a number of insulation failures during the commissioning period. The problem was identified as a resonance that had been excited near the switching transients frequency. It was solved by utilizing cable terminations that are frequency independent up to Megahertz level, and by installing a passive filter tuned at the switching frequency. Mal-operation and derating of equipment are also considered as costly consequences of harmonic presence in the network. The zero-crossing operation based equipment (e.g. contactors and line commutated switches) and the solid state relays and controllers can malfunction or trip due to high harmonic distortions [84]. Furthermore, the derating of standard transformers or installing K-rated transformer when supplying distorting load can be an example of the extra cost when operating in a harmonic rich environment [80, 85]. Premature aging, inconsistency in energy and power factor meters (induction disk based) readings and high neutral currents are also examples of sub-optimal operation of distribution networks due to harmonic presence [86, 87]. 65

66 Harmonic mitigation The mitigation of harmonics can be based on network or device mitigation solutions. For the network solutions, the harmonics can be mitigated by shifting the resonance frequencies to safe bands. This can be performed by changing capacitor locations or/and sizes or providing higher short circuit capacity at the disturbing load connection node. Equation (2.3) [80] shows the relationship between the short circuit level and the size of connected capacitor at a bus. h r = SCC pu Q Cpu (2.3) As it can be seen from the equation, the increased short circuit capacity pushes the resonance frequency to a higher frequency band, at which in a normal situation, there is lower level of harmonic injections. Contrary, connecting large capacitors leads to reduction in resonance frequency, potentially leading to harmonic resonance problems. The device based solutions can be divided into two main techniques: installing devices that are able to block or trap harmonic currents (shunt and series filters) or reconfiguring the existing devices to have improved immunity and performance in terms of harmonics. The first category of installing filters is a well proven solution if properly designed and located. The second category involves increasing neutral size (reduce zero sequence harmonic losses), utilizing transformer connections (delta and zigzag connections) to block harmonic currents flow, improving converter harmonic injections (by increasing the number of pulses and adopting PWM techniques for harmonic cancellation), installing smoothing reactors ( chokes ) and tuning and/or detuning of capacitor banks [11, 80, 87, 88]. 66

67 2.3.2 Voltage sag Voltage sag is defined as a decrease in the voltage RMS value between 0.9 and 0.1 p.u typically lasts between half a cycle (10 ms in 50 Hz systems) and several seconds (see Figure 2.5). The main cause of voltage sags are the faults in the network, but they can also occur as a result of connecting large loads, motor starting or transformer energising. Voltage sag is the most disruptive and costly PQ phenomenon, as it often leads to industrial/commercial process interruptions and subsequent significant financial losses to end-use customers. Voltage sags are typically described in international standards by magnitude and duration only. Other sag characteristics include, phase angle jump (shift), point on the wave of sag initiation and voltage recovery and symmetry (i.e., single phase sags, two phase sags, three phase sags) [1] Sag evaluation and indices The standard IEC [2] defines the basic measurement of sags as the V rms(1/2) which is the value of the RMS voltage for one cycle refreshed every half a cycle. Voltage sags are usually characterised by sag magnitude (remaining voltage) and sag duration. Sometimes the voltage depth or voltage drop are used instead of magnitude. The following definitions typically apply [56], - The remaining voltage is the lowest V rms (1/2) measured during the sag. - The depth/drop is the difference between the remaining voltage and the reference voltage. - The duration is the time difference between the start of the sag (voltage below a certain threshold) and the end of the sag (voltage is equal to the threshold). Typically, the start is detected when the RMS voltage in one of the three phases drops below the threshold and the end when the voltage of the last phase recovers. 67

68 The voltage sag evaluation and indices, presented in [69], can be described based on single event characteristics, site indices and system indices. An example of event characteristics is voltage sag severity which can be calculated by equation (2.4) below according to certain reference voltage tolerance curves. SEMI F47 reference curve [89] is adopted in [69] to provide an algorithm to calculate sag severity as shown in Table 2-3. S e = 1 V 1 V curve (d) (2.4) where, V d V curve (d) is the voltage sag magnitude is the event duration is the magnitude of the reference curve for duration d Table 2-3: Voltage sag severity with reference to SEMI F47 curves (adopted from [69]) Duration range Calculation of voltage sag severity d < 20 ms S e =1-V 20 ms < d < 200 ms S e =2(1-V) 200 ms < d < 500 ms S e =3.3(1-V) 500 ms < d < 10 s S e =5(1-V) d >10 s S e =10(1-V) The site indices are calculated, for one site, based on the recorded sag events characteristics. Indices like System Average RMS Variation Frequency Index (SARFI) and others are presented in [69] to provide a measure of the voltage variation performance for a site or a bus by giving a count of voltage sag, swells and interruption events. The system indices for voltage sag are then calculated from the site indices. The aggregation of site indices into system indices can be performed by either calculating a weighted average of the monitored sites, or by determining the value of the site indices not exceeded by 95 % of the sites (the 95 th percentiles of the sites indices) [69]. 68

69 In general, the IEEE 1564 recommends a five-step procedure to measure the performance of power system regarding the sag, as shown in Figure 2-7. The five steps can be summarised as follows: - Sample the voltage with a certain sample rate and resolution. - Calculate the voltage characteristics from the sampled voltage for a period of time. - Identify the sag events, and calculate each single event characteristics. - Calculate the site indices from the characteristics of single events detected for the site. - Calculate the system indices from the number of monitored sites indices. Figure 2-7: A general framework for obtaining system-wide sag indices (adopted from [90]) In addition to the standard sag indices suggested in the IEEE 1564 guide [69], a number of sag severity indices have been proposed in the literature [91-94]. In [6], two indices Magnitude Severity Index (MSI) and Duration Severity Index (DSI) were proposed to represent the magnitude and duration severity respectively, and served as input parameters for different assessment approaches. Magnitude-Duration Severity Index (MDSI) uses single numerical value, by combining MSI and DSI, to represent the failure risk of equipment when subjected to voltage sags. Apart from these numerical indices, voltage sag indices based on fuzzy logic approaches were also proposed [95-97]. 69

70 One of the informative sag performance indices is the Magnitude Duration Severity Index (MDSI) equation (2.7) [6]. It combines the Magnitude Severity Index (MSI), equation (2.5), that describes the voltage sag based on the magnitude of the remaining voltage, and the Duration Severity Index (DSI), equation (2.6), that describes the voltage sag based on the duration of the sag. Figure 2-8 shows the probable regions for sag events, where in the yellow region (P) the sag event will not cause trips (not severe), in the red region (R) the sag event will definitely cause trips (most severe) and in the blue region (Q) the trips are uncertain (variable severity) [6, 98]. Figure 2-8: Probable regions of voltage sag events (adopted from [98]) The values V max, V min, t max and t min from Figure 2-8 are used to calculate the MDSI using the following equations, MSI = DSI = 0, m > V max 100 (V max m) ( ), V V max V min m V max (2.5) min { 100, m < V min 0, d < t min 100 (d t min ) ( ), t t max t min d t max (2.6) min { 100, d > t max MDSI = MSI DSI 100 (2.7) The MDSI ranges from 0 (in the boundary between the safe (P) and uncertainty area (Q) in Figure 2-8) up to 100 (in the boundary between the uncertainty area (Q) and the definite trips area (R) in the Figure 2-8). The MDSI can be efficient for the following 70

71 applications; to quantify the impact of different sag events on certain equipment, to make a comparison between the severities of a certain sag event on different equipment (different probable regions in Figure 2-8), identification of the equipment most prone to sag in certain industrial processes and to rank different sag events based on their impacts on certain equipment [6] Consequences of inadequate voltage sag level Voltage sag is considered to be one of the most prominent PQ problems. This is mainly due to the relatively large number of occurrences throughout a typical transmission and distribution network, the increasing sensitivity of network components and customer equipment to voltage sags, and the high costs of lost productivity due to equipment trips or the trips of sensitive power plants (power electronically interfaced generation) [99]. Many reports and surveys have presented the high cost of the sag induced process interruptions. The main consequences of voltage sags reported in [100] are the reduction of network energy transfer capability during a sag event, which might also reduce the required fault clearing time for stability limits. Voltage sag may also lead to tripping of power electronic devices (generation interfacing or network conditioning devices) [100]. These can be network related impacts. Moreover for the customer sensitive devices like computers, controllers and adjustable speed drives, mal-operation is highly expected for voltages below 85% for more than 40 ms. Normally the immunity of such sensitive devices to voltage sag incidents can be fully modelled based on the sag magnitude and duration. However, devices like AC contactors, induction motors and DC adjustable speed drives can be affected by other characteristics of a voltage sag event, namely point-on-wave and phase angle jump of events [76, 100]. 71

72 As far as customers are concerned, particularly industries, the voltage sag phenomenon has a drastic impact when it causes an industrial process interruption due to critical equipment tripping [101]. Costs including wasted materials, work progress interruption, idle workers, equipment damage, delayed delivery of products and reputation of the industry can be all associated with inadequate PQ in terms of voltage sag [72]. Although, the estimated costs of sag disturbances vary significantly between surveyed customers in [6] based on e.g., size, type, location of customers; sag disturbances and momentary interruptions are still considered the most costly PQ disturbance. Industries like semiconductor manufacturing, pharmaceutical industry, paper industry, steal manufacturing, food processing, textile industry and the information technology sector are reporting the highest losses due to voltage sag [6]. Report [102] showed that for 12 lowtechnology sites, ranging between 5 and 30 MVA over a ten month period, the average financial loss per event reached 14,300 and for high technology manufacturing, e.g. semiconductor production, the typical financial loss per event could reach 3,800, Voltage sag mitigation The mitigation of voltage sags can be network based or device based [6, 8]. The network based solutions are mainly concerned with improving the network performance regarding short-circuit faults, the main cause of sags. This can be performed by taking measures to reduce the number of faults in the network by undergrounding of lines, insulating overhead lines, trimming trees, installing animal guards and improving the information and signs systems of the underground cable locations to prevent dig-in faults. Furthermore, limiting the faults severity and area of impact can improve the sag performance of the network. This can be performed by having improved protection systems so that faults can be cleared faster, installing fault current limiters to reduce shortcircuit currents and installing surge arrester to limit the impact of lightening surges. 72

73 The device based mitigation solutions of voltage sag can be performed in two ways: by either installing new devices that are capable of negating the effect of voltage sag disturbance or by improving the immunity of sensitive equipment to sag disturbances. Devices like online Uninterrupted Power Supply (UPS) and Flexible AC Transmission System (FACTS) devices like Static VAR Compensator (SVC) and Dynamic Voltage Restorer (DVR) have proved very efficient in mitigating voltage sag by injecting the required reactive power to support the voltage during disturbances; however high cost and sophisticated control of such devices hinder the wide deployment of such devices [8]. On the other hand, increasing the immunity of devices to voltage sag can be performed in the manufacturing stage by improving the ride-through sag capabilities of equipment which is also a trade-off between cost and performance. A combination of network based and device based solutions is discussed in [6], where a redundant supply with a very fast switch (static transfer switch STS) is applied to improve sag performance for a sensitive load. In this arrangement the STS is capable of switching the load supply from a faulty feeder to the healthy one in ¼ to ½ cycle. This ensures that the sag disturbance remains at the load bus for a very short period and most of the equipment devices will be able to ride-through. This arrangement requires that the two supplies are from two electrically distant buses to ensure both supplies are not affected by the same faults and disturbances Voltage unbalance Voltage unbalance describes the condition when the three phase voltages are of different magnitudes and/or do not have an equal phase shift of 120 with respect to each other. The voltage unbalance can originate at the distribution levels due to several reasons; the three main origins can be unbalanced injections (single phase power conversion equipment), malfunctioning equipment (non-standard performance) and potential 73

74 connection problems. Long untransposed lines are also sources of unbalance, though these are not expected to be found in distribution networks. The unbalanced injections can occur due to large single phase loads and generators connected at three phase nodes, e.g. single phase traction railways systems [103, 104]. Examples of non-standard operation conditions and equipment which lead to unbalance problems are unequal phases impedances or transformer tap positions, persistent unsymmetrical faults and single phase blown fuses (frequent problem for capacitor banks) [ ]. Connection problems can cause unbalance when long lines are not transposed and when terminations and connections are not properly performed (loose terminations, non-identical insulators, etc.) [105, 108]. As the power supplied from transmission levels is usually regulated to be balanced, the unbalance in distribution networks is typically not emitted from upstream infrastructure under normal operation conditions. Voltage unbalance yields extra costs in operation, maintenance or replacement for both network operators and end customers; it is an essential topic for the future designs of smart grids where the presence of single phase load and generation in particular will be noticeably increased [77]. In recent years, unbalance studies have attracted more attention due to the increasing complexity of the distribution network. The large-scale integrations of single-phase load, dual-phase load and storage amplify the unpredictability of unbalance in the network. Therefore, the initial network construction plan may not provide balanced working conditions for customers with such temporally and spatially varying disturbance sources. The distributed generation (DG), such as single-phase renewable energy sources (e.g. PVs), may either mitigate unbalance or even aggravate it. 74

75 Unbalance evaluation and indices Voltage unbalance is typically defined using Voltage Unbalance Factor (VUF). The VUF is defined as the ratio of the negative sequence (sometimes zero sequence [109]) voltage component to the positive sequence voltage component, equation (2.8) [110]. VUF = V 2 V % (2.8) The alternative approach given by NEMA (National Electrical Manufacturers Association) defines unbalance as the ratio of maximum voltage deviation from the average line voltage over the average line voltage equation (2.9) [111]. VUF = Max( Va V, Vb V, Vc V ) V (2.9) where V a, V b, and V c are line voltages of phase a, b, and c respectively, V is the average value of the RMS voltage of V a, V b, and V c. It is important to mention that the IEC definition is mathematically more rigorous compared to the definition given by NEMA. So, while calculating VUF using different definitions may give different results, the standard defined by NEMA (National Equipment Manufacturer s Association) is relevant to the line voltage unbalance rate which only involves voltage magnitude and not the phase angles, while the IEC standard calculates VUF considering both voltage magnitude and phase angles. The unbalance evaluation methodology can be summarized as follows [56]: - RMS value of VUF aggregated over 3 seconds: very short time VUF vs. - RMS value of VUF aggregated over 10 minutes: short time VUF sh. - 99% and 95% probability of VUF vs and VUF sh should not exceed determined planning levels for daily and weekly values respectively. 75

76 At the MV level, i.e. 1 kv < V n < 36 kv, most of the international and regional standards and guidelines recommend a VUF limit up to 2%; some documents show more relaxed limits for single phase dominant distribution networks (up to 3%), where most of the documents have more strict values for the HV and EHV (1% - 2%). According to the EN [60] for a period of one week, the 95 th percentile 10-minute mean RMS values of the voltage unbalance should not be more than 2 % under normal operating conditions. EN defines the limit in terms of negative to positive sequence voltage components (equation (2.8)) and it restricts the voltage unbalance to 2% only, for all voltage levels [60]. In case of unbalanced supply voltage other IEC standards are concerned with low voltage characteristics and also set the limit to 2% only [2, 112]. IEEE Recommended Practice for Electric Power Systems in Commercial Buildings allows voltage unbalance of 2 or 2.5% for some electronic equipment [113]. ANSI C84.1 suggests a maximum voltage unbalance up to 3.0% under no-load conditions [114]. Considering that EN and IEC are the most relevant standards for European DSOs the defined index and limits in these standards are adopted in the thesis Consequences of inadequate unbalance levels In general, unbalanced voltage causes overheating in the three phase equipment of both DSOs and customers, such as transformers and motors, contributing to accelerated thermal ageing and therefore a reduction in the equipment s lifetime. With unexpected negative sequence power flowing in the same path with positive sequence power, the capacities of online equipment are reduced, resulting in a reduction in efficiency. In addition, generators and induction motors must be de-rated due to equipment safety considerations, as will be discussed below. All the above will eventually lead to increased costs of investment, operation and maintenance of distribution networks. 76

77 Voltage unbalance has more significant impacts on certain equipment types, mostly 3 phase equipment. Induction motors are one of the equipment devices most susceptible to voltage unbalance [111]. The presence of the negative sequence component in the supply voltage produces inverse rotating magnetic fields in the motor. This will lead to negative mechanical torque in the shaft, which subsequently leads to forced deceleration, increased mechanical losses, reduced efficiency and increased noise. The presence of the negative sequence torque in the opposite direction of the working torque will also prevent the machine from rotating at full nominal speed and torque. The unbalance in the supply voltage also leads to an even higher current (also unbalanced) consumed by the motor, leading to higher operating temperatures and faster aging [10, 77, ]. Power electronic converters are also significantly impacted with voltage unbalance. Voltage unbalance will lead to non-characteristic harmonic injections (e.g. triplens) from the converters which are normally not expected and therefore unmitigated. Increased reactive power demands (lower power factor) are also expected under unbalance conditions. For line commutated converters, asymmetric conducting of switches leads to higher input current, sometimes with fewer pulses (higher harmonic distortion) which might lead to protection tripping of the whole converter system [107, 118, 119]. For power transformers, unbalance can lead to reduced capacity due to the negative and zero sequence components presences. Some winding arrangements prevent the zero sequence from propagating to the upstream network (improving the unbalance performance); however the zero sequence currents circulating in the transformer windings might lead to excess heating in the transformer [120]. It was estimated that there were 2.4 GWh and $134,000 additional annual losses due to the presence of unbalance based on 17,600 transformers in Brazil (about $7 per transformer per year) [121]. 77

78 Unbalance mitigation As unbalanced loads are one of the main causes of voltage unbalance, the redistribution of single phase loads among the three phases can be considered one of the straight forward mitigation solutions [122, 123]. The reconfiguration of the system can be performed in the planning stage of new connections, providing alternative supply points or reconnection of existing loads to different phases via manual or automatic switching. The reconfiguration of the system was proved efficient in rebalancing the system and reducing losses [103]. The modern power electronic and FACTS devices are also investigated as an unbalance mitigation solution. In [124] an active power filter based on PWM converter was utilized to rebalance loads, by injecting the appropriate sequence component to cancel the measured negative sequence at the load. Dynamic Voltage Restorer DVR, Static VAr Compensator SVC and active line conditioner are also investigated to inject/absorb reactive power, shift power between phases and changing the affective impedance of lines as unbalance solutions [125, 126]. The recent European project SuStainable [127] has also investigated the feasibility of demand side management DSM as a solution to unbalance by utilizing production and consumption of available mobile storage devices, e.g. electric vehicles. 2.4 Summary This chapter first discussed the PQ in general and then focused on the main PQ issues with regards to distribution system utilities and customer concerns. The categorization of PQ and key phenomena of interest for this research in terms of definitions, standards and monitoring were also presented. These key phenomena - harmonics, voltage sag and voltage unbalance- were discussed in terms of their categorization, evaluating indices, consequences and mitigation solutions. 78

79 3 Power System Modelling for PQ studies Two distribution systems are modelled as case studies for the PQ analysis research presented in this thesis. The first system is the 295-bus Generic Distribution Network (GDN), a generic system with different levels of voltages with parameters based on typical UK systems. The second system is an existing real test feeder which consists of 35 buses at the MV level (15 kv). This chapter presents detailed description of both networks with methodologies and assumption adopted in the modelling, the systems parameters are provided in Appendix A. This chapter also presents the simulation models of the considered PQ phenomena. The considered PQ disturbances are modelled probabilistically to accommodate the expected uncertainties and variations in the PQ performance. All models and simulations presented are performed in DIgSILENT and OpenDSS software packages. 79

80 3.1 Test systems Generic Distribution Network (GDN) The GDN comprises four 400/275 kv transmission in-feeds, 132 kv and 33 kv subtransmission networks (which are mostly meshed) and a distribution network which predominantly operates at 11 kv, but also has some small sections at 0.4 kv (and is mostly radial). This GDN was used in different types of studies in the past [128, 129]. Figure 3-1 shows the single line diagram of the network. The PQ studies and evaluations performed are focused in the MV level, i.e. the 11 kv buses supplied by the S/S transformers C, D, H, I, J, K and L as shown in the figure. The following sub-sections give further details about the network components Modelling of Buses/Lines/Transformers The GDN comprises of 295 buses, of which 150 are loaded buses at the subtransmission, HV distribution and MV distribution level. There are 5 transmission buses at the 400 kv and 4 buses at the 275 kv level; 23 buses at the sub-transmission 132 kv level, of which 3 buses with connected loads; 25 buses at the HV distribution 33 kv level, of which 6 buses with connected loads; 234 buses at the MV distribution 11 kv level, of which 140 buses with connected loads; and 4 buses at the LV distribution 0.4 kv level (1 bus load is modelled, negligible value though compared to the aggregated 11 kv loads). The main interest of the PQ studies presented in this thesis is the PQ performance at the MV buses and feeders. Therefore, the modelling at the LV network is not detailed. The main interest was in modelling limited number of components at this level in order to study the propagation of disturbances to the MV level, rather than the performances of the LV buses themselves. Similarly, the transmission levels buses (400 kv and 275 kv) are modelled to represent a number of different locations of Grid Supply Points (GSP), and the PQ performance at these levels is not evaluated. 80

81 Figure 3-1: Generic Distribution Network (GDN) single line diagram 81

82 The branches in the GDN are a combination of different sizes of overhead lines (OHL) and cables, total of 278 branches of which 52% OHL. The lines ratings and faults rates are assumed equal at the same voltage level and type of branch. The branches details can be summarized in the following; ; 4 (132 kv) OHL [X/R=1]; 22 (132 kv) cables [X/R range between with average of 4.7]; 5 (33 kv) OHL [X/R=1]; 18 (33 kv) cables [X/R range between with average of 3.0]; 136 (11 kv) OHL [X/R range with average of 0.7]; 93 (11 kv) cables [X/R range between with average of 0.5]. A neutral is added to all the 11 kv lines. The neutral is connected to the star point in the 11 kv side of all substation transformers that feed the 11 kv, this was necessary to facilitate the convergence of the unbalance load flow under extreme unbalance simulations. All 39 transformers in the GDN are two winding transformers and of the same rating (99 MVA). Table 3-1 presents transformers details. The four 400/400 kv/kv are connected to an infinite bus and represent the GSPs (see Figure 3-1). They are modelled to control the power flow and angles at the different locations of GSPs. The rest of the transformers are modelled as typical power transformers that connect the transmission network to the sub-transmission, the HV distribution and the MV distribution network. The main focus in the selection of parameters was how these models affect the propagation of PQ disturbances from HV levels to the MV level. Table 3-1: GDN transformers modelling Number of Transformers Voltages ratios (kv/kv) Winding connections* Short Circuit voltage (%) X/R ratio (pu) 4 400/400 Yy /275 Yy /132 Yy /132 Yy /33 Yd /11 Yyn /11 Dyn /0.4 Yd /0.4 Yd *Y=Star, D=delta, n=connected to neutral 11=-30 phase shift 82

83 Load (p.u.) Loads All the loads are three phase balanced loads. The total peak load of the GDN is MW and MVAr. The loads were divided into three types, industrial (1.6 MVA), commercial (76.7 MVA) and domestic (306.6 MVA). 147 buses have a combination of 80% domestic and 20% commercial load, and 3 buses are dedicated as industrial loads only. Annual hourly loading curves are extracted from year 2010 survey of different types of loads and applied to each load in the network. For each type of load, the corresponding annual load duration curve (LDC) is used, three in total. The LDCs are used to select representative hours of year to evaluate the PQ performance. Figure 3-2 shows the stepwise LDC segmented into 11 operating points having different durations. The maximum hour demand during the year is MW and MVAr; note that these are different from the maximum load of network due to different peak time for different types of load. The average power factor is 0.98 and the total real power losses are around 1%. The conservative values of the power factor and losses are chosen to compare deterioration in operation performance due to PQ disturbances in an ideal network Hours Figure 3-2: Segmented domestic load duration curve (LDC) of GDN network 83

84 Distributed generation The DG units are connected at different buses at the 11 kv level. The buses are chosen to cover all different locations which might have effects on the PQ performance of the network, i.e., at the start of a feeder (close to the MV/LV sub-station), at the middle of a feeder and at the end of a feeder. Three generator units types of DGs are considered, wind generators, photovoltaic and fuel cells. The maximum DG penetration in the year occurs at a point when the DG generation covers 24.45% of the real load (31.3 MW out of 128 MW total load), all DGs work at unity power factor with no voltage controllers modelled. The wind generators are modelled as doubly fed induction generators DFIG. The fuel cell and photovoltaic generators are modelled as full converter connected generators. The wind and photovoltaic generators have an annual hourly output curves, extracted from realistic outputs data based on the UK weather [130, 131]. The maximum output of these two types of DGs occurs at different times of the year. The fuel cells are assumed to have a constant output throughout the year. The fuel cells and PV DG units are mostly single phase with few relatively larger size (>2 MVA) units modelled as three phase. Two scenarios with the same total level of DG penetration (max 30% penetration of the feeders load) are adopted. The first scenario is the connection of fewer (12) larger size (2-5 MW) DGs. The second scenario is the connection of more (25) units of smaller size. Figure 3-3 shows the output curves of the (a) wind and (b) PV for two selected months (normalized based on the maximum output), Appendix B shows the whole year output curves of the wind and PV generation. 84

85 Output (p.u.) Output (p.u.) Output (p.u.) Output (p.u.) 1 January Hour of the month May Hour of the month (a) Wind Gen output curve January Hour of the month May Hour of the month (b) PV output curve Figure 3-3: Output curves of PV and wind generators for January and May 85

86 3.1.2 Real test feeder The real test feeder shown in Figure 3-4 is a 35-bus medium voltage (15 kv) distribution feeder connected to the external grid through 60/15 kv Yy transformer. It is modelled in DIgSILENT PowerFactory and OpenDSS simulation software packages. Thirty buses are load buses with both commercial and domestic types of load connected. The total peak load is 3.1 MW and 1.1 MVAr. The displacement power factor (power factor for non-distorted sinusoidal voltages and currents) is assumed constant and equal to 0.94 at all load buses. The load distribution along the feeder is based on each load 15/0.4 kv service transformer size. The loads (38 in total) have service transformers size range between 200 and 630 kva (for brevity service transformers are not shown in Figure 3-4). Based on the feeder total current recorded at the substation, the service transformer size of each load, and the assumption of 0.94 constant power factor throughout the feeder, the load values are assigned at each hour of the study periods External Grid /15 kv Yy S/S transformer Figure 3-4: Real test feeder single line diagram Different types of cables are installed, with X/R ratio ranging between 0.2 and The total length of feeder branches is around 8 km. To accommodate studies at harmonic 86

87 Current (A) triplens, zero sequence path is provided by connecting sub-station transformer windings and loads as grounded star. Different scenarios of connection of untuned capacitor banks are applied. A total of 600 kvar capacitors, connected as grounded star, is applied, compensating around 50% of the feeder peak reactive load, and improving the average displacement power factor to approximately Similar to the GDN, the LDC of the feeder load (recorded at the substation) is plotted in Figure 3-5 and segmented into 10 segments (in addition to maximum and minimum load points) to represent testing hours for annual evaluation. The daily loading curves of 4 days (Thursday-Sunday) with 15-minute resolution are used in simulations Hours Figure 3-5: Segmented total load duration curve (LDC) of real test feeder The mixture of commercial/domestic loading curves led to the approximately flat peak in weekdays, as shown in Figure 3-6. The actual PQ measurements at some buses are also used to estimate the feeder PQ performance, as will be discussed in Chapter 4 and Chapter 5. Figure 3-7 shows the harmonic and unbalance performances based on measurements at the 0.4 kv side of Bus 15 service transformer over the period of four days with 10 minute resolution. The unbalance performance in Figure 3-7 (b) is illustrated based on both definitions of VUF (see Chapter 2), i.e., the ratio of negative sequence (blue line) and zero sequence (red line) to the positive sequence voltage. 87

88 VUF (%) Harmonic voltages (V)/ THD (%) Current (A) Time slot (15 min) Figure 3-6: Daily loading curves for the real test feeder total load har 3 har 5 har 7 har 9 har 11 har Time slot (10 min) (a) Harmonic performance THD (%) Time slot (10 min) (b) Voltage unbalance performance Figure 3-7: Bus 15 PQ measurements V2/V1 V0/V1 88

89 3.2 PQ phenomena modelling Modelling for harmonics In distribution networks, the main continuous sources of harmonics are the traditional non-linear loads such as saturated transformers, rotating machines and arc furnaces. More recent types of non-linear loads are the electronic devices supplied by switched-mode power supplies and the fluorescent lighting. However, the strongest impact on the distribution network harmonic performance can be expected from the increasing number of power electronics interfaced loads and generators [80]. For all studies presented in the thesis, the harmonic sources are represented by the Norton equivalent shown in Figure 3-8 consisting of a primitive admittance matrix, Y prim (h), and a current source in parallel. The current source injects different level of harmonic currents. The net harmonic injections in the network by the power conversion element (generator/ load) is then calculated based on the corresponding Y prim (h) and existing background harmonic voltage. Y prim (h) I h Figure 3-8: Norton equivalent of harmonic sources [132] Impedances of network components Assuming that the inductance (L) and the capacitance (C) of the network components are independent of the frequency, the change in the reactance will be according to equations (3.1), (3.2) [80]: X L (h) = hx L (3.1) 89

90 Z (ohm) - Angle (deg) X C (h) = X C h (3.2) The interaction between X L and X C determines the dominant reactance of the network at different frequencies, i.e., if inductive or capacitive current is flowing. The interchange points between inductive and capacitive network (and vice versa) in typical frequency scans of a node usually correspond to parallel and series resonance frequencies (hills and valleys in the scan) as shown in Figure 3-9 (a). The real (resistive) part of the network impedance at different frequencies determines the height of the hills in impedance plots, therefore a network with high resistive loads can reduce harmonic resonant voltage [87]. Figure 3-9 (b) shows the R and X scan of the same bus. Note that the change in the real (resistive) part of the network impedance with frequency is not due to the skin effect, which is not considered in the thesis, it is rather due to the change of components impedances and angles at different frequencies. As shown in Figure 3-9, the bus has two parallel resonance frequencies in the scanned bandwidth, at around the 11 th and around the 19 th harmonics. The change in the reactance sign in Figure 3-9 (b) indicates the change in the dominant reactance, from inductive (X increases with the frequency) to capacitive (X decreases with the frequency). The resonance phenomenon in distribution feeders is further discussed in Chapter Z angle Harmonic number (a) Z and angle frequency scan 90

91 Ohm Harmonic number (c) R and X frequency scan Figure 3-9: Impedance frequency scan R X Transformer connections Under balanced conditions, the transformer connections can block the propagation of zero sequence harmonics, i.e., triplen harmonics. The delta and isolated star connection at either high or low voltage side of transformer will block the propagation of the harmonic currents. A study performed in the real test feeder by injecting equally zero (3 rd harmonics), negative (5 th harmonics) and positive (7 th harmonics) sequence harmonics was performed to analyse the impact of different connections. A single balanced source of 100% injections with 0 angle shift is connected in the middle of the feeder. Table 3-2 (a) shows the three phase current injections of the harmonic source. The harmonic currents and voltages are recorded at the head (substation) and at the end of the feeder. Table 3-2 (b, c and d) shows the harmonic voltages and currents recorded at the head and at the end of the feeder for phase A for three case studies. Case 1 is when the zero sequence path is available through Yn/yn connected service and substation transformers, the recorded harmonic currents and voltages are shown in Table 3-2 (b). Case 2 is when the zero sequence path is disconnected due to delta connected transformers, the recorded harmonic currents and voltages are shown in Table 3-2 (c). Case 3 is when zero sequence currents are injected into the feeder through Yn/yn connected service transformer, while the rest of the network is isolated through delta connections, the recorded harmonic currents and voltages are shown in Table 3-2 (d). As expected, for both the positive and negative 91

92 Table 3-2: Transformers connection impact on harmonic propagation (a) Harmonic injections at the middle of the feeder Harmonic I1 (A) IAngle1 ( ) I2 (A) IAngle2 ( ) I3 (A) IAngle3 ( ) (b) Case 1 (connected zero sequence path) harmonic currents and voltages Head of feeder End of feeder Harmonic V1 (V) VAngle1 ( ) I1 (A) IAngle1 ( ) V1 (V) VAngle1 ( ) I1 (A) IAngle1 ( ) (c) Case 2 (disconnected zero sequence path) harmonic currents and voltages Head of feeder End of feeder Harmonic V1 (V) V Ang1 ( ) I1 (A) I Ang1 ( ) V1 (V) V Ang1 ( ) I1 (A) I Ang1 ( ) E E (d) Case 3 (partially connected zero sequence path) harmonic currents and voltages Head of feeder End of feeder Harmonic V1 (V) VAngle1 ( ) I1 (A) IAngle1 ( ) V1 (V) VAngle1 ( ) I1 (A) IAngle1 ( ) E sequences harmonics, currents flow from the source toward the substation (high short circuit capacity bus), therefore the head of the feeder always records higher currents and lower harmonic voltages while the opposite is true at the end of the feeder (this is discussed in details in Chapter 4). The third case, i.e. the flow of triplen harmonics in case of partially connected zero sequence path, deserves particular attention, as in this case the whole feeder will record high zero sequence harmonic voltages with very low currents (which mainly leak through the capacitance of the lines as there is no other path to the ground). Figure 3-10 shows the impact of different length of branches (and subsequently 92

93 Line 01 Line 02 Line 03 Line 04 Line 05 Line 06 Line 07 Line 08 Line 09 Line 10 Line 11 Line 12 Line 13 Line 14 Line 15 Line 16 Line 17 Line 18 Line 19 Line 20 Line 21 Line 22 Line 23 Line 24 Line 25 Line 26 Line 27 Line 28 Line 29 Line 30 Line 31 Line 32 Line 33 Line 34 Line 35 Current (A) capacitance) in Case 3 by comparing the third harmonic current flowing through both ends of each line of the real test feeder. Terminal i current (blue line in Figure 3-10) is recorded at the start of the branch and terminal j current (red line in Figure 3-10) is recorded at the end of the branch current at terminal i current at terminal j Figure 3-10: 3 rd harmonic currents flowing through both terminals of the test feeder lines Shunt Capacitor connections The power factor correction capacitors are usually expected to be connected at the LV networks, i.e., at the customers plants. In some cases, however, DSO can connect capacitors at the MV level to support the voltage and to improve the overall power factor of the feeder. The capacitors size and connection will significantly influence the flow of the harmonic currents, by providing additional, lower impedance, path than the GSP. Also, although capacitors are not intrinsically harmonic sources, they can significantly affect the harmonic performance of the network by magnifying currents and voltages under resonance condition, and by changing the overall network impedance at different frequency. Capacitor banks can be found in delta or star connections, grounded or isolated, and as tuned, detuned or untuned banks [11, 80]. The different delta/star connection, and grounded or isolated capacitor banks have different applications, advantages and disadvantages [133]. The tuning of the capacitor is mainly performed to prevent harmonic problems in the plant and to protect the capacitor itself. Untuned capacitor banks are 93

94 connected for the sole purpose of reactive power support, and they are not expected to cause resonance or to operate in rich harmonics environment. The capacitor banks for the purpose of power factor improvement are normally sized based on the reactive power requirement of the plant. If capacitor bank connection leads to problematic resonances, some measures can be adopted to avoid aggravating harmonic problems. If it is possible to resize, relocate or locally deploy (i.e., VAr compensation for equipment rather than for the plant) the capacitor banks, the adverse effects of the interaction between capacitor banks and harmonic sources can be avoided. In rich harmonic environment, the connected capacitor banks are usually detuned or tuned to avoid harmonic resonance problems. The detuned capacitors are also for the sole purpose of reactive power support. They however, based on the size and connection location, are expected to cause resonance which is most probably to be excited by the connected load injections. In this case the capacitor banks can be detuned, hence the name, to alter the resonant frequency by connecting reactors to the capacitors and pushing the peak in the impedance scan to a safe frequency. Note that by detuning existing capacitors, the operation current of the capacitors is not expected to increase significantly, as the detuned capacitor will still not provide zero impedance path for any harmonic current. Nevertheless, the operating voltage of the capacitors is expected to increase due to the introduction of reactance between the supply point and capacitors. Tuning the capacitor banks is also performed by connecting a reactor in series with the capacitor banks. In this case though, the reactor is selected to create low impedance path when combined with existing capacitors (i.e. series resonance), at a certain problematic harmonic frequency. The conversion of power factor correction capacitors into harmonic filters requires revaluation of the voltage and current ratings of the capacitors. As in such a case, higher currents and peak voltages are expected in the new power factor correction set. This is due 94

95 to the additional harmonic duty of the set (i.e. filtering harmonic current) on top of the existing fundamental duty (i.e. reactive current injection) [11]. Tuning and detuning capacitor banks are not always practically possible, especially if problematic harmonics are low order harmonics or if resonance occurs at low frequencies. For example, to detune a 100 kvar, 400 V, 50 Hz capacitor bank due to the 3 rd harmonic presence, the reactor must shift the resonance to a frequency below the 3 rd harmonic, typical practice is to tune the set to the 2.7 th harmonic. A 14% relative impedance (relative to the capacitor impedance) detuning reactor is required, i.e., mh reactor with physical dimensions of 35 cm x 22 cm x 35 cm and weighting 55 kg [134]. A simple study was performed at the real test feeder to examine the effects of connected capacitor bank at the end of the feeder. A relatively good harmonic performance was simulated (all buses THD < 2%). Harmonic currents were injected from six sources, injecting 3 rd, 5 th, 7 th and 9 th harmonic currents, with dominant 5 th harmonic injections. A 100 kvar three phase capacitor was connected with different connection arrangements (delta, wye and grounded wye) and tuned to different harmonic frequencies (3 rd and 5 th harmonics). Table 3-3 (a) and (b) show the recorded currents and voltages at both the head and end of the feeder based on different connections arrangement of the capacitor bank. Figure 3-11 (a) and (b) also show the end of the feeder harmonic currents and voltages, note that the voltages in Figure 3-11 (b) are shown as a ratio to the fundamental voltage. As it can be seen from the figure and the tables the delta connected capacitor has the minimum impacts on the harmonic performance compared to the no-capacitor case. The Yg capacitor bank can affect all harmonics. Different capacitor connections however did not significantly affect the total current and voltage distortion of the feeder. Noticeable performance improvement was observed though for the tuned filters with proper connections, i.e., Yg for 3 rd harmonic filtering and Y or Yg for 5 th harmonic filtering. The 95

96 Current (A) Table 3-3: Capacitor different connections impact on the real test feeder harmonic performance (a) Harmonic currents performance Head of the feeder currents (A) End of the feeder currents (A) Total current (A) h no capacitor untuned tuned to 3rd tuned to 5th D Y Yg D Y Yg D Y Yg (b) Harmonic voltages performance Head of the feeder voltages (V) THD End of the feeder voltages (V) h (%) no capacitor untuned tuned to 3rd tuned to 5th D Y Yg D Y Yg D Y Yg THD (%) h=3 h=5 h=7 h=9 Capacitor connection (a) Harmonic currents at the end of the feeder 96

97 Voltage & THD (%) rd 5th 7th 9th THD Capacitor connection (b) Harmonic voltages and total distortion at the end of the feeder Figure 3-11: Harmonic performance at the end of the feeder (capacitor connection point) selected capacitor size and location do not cause resonance near the injected harmonics. In such a case, as shown in Table 3-3 and Figure 3-11, regardless the connection arrangement the capacitor will not degrade the harmonic performance. Slight increase in the THD was recorded though from 1.36% in the no-capacitor case to 1.4% in the Yg untuned capacitor case. The filtering effects at the capacitor connection bus (end of the feeder) are also illustrated in Figure 3-11 by the significant increase in the harmonic current at the tuned frequency and the significant reduction in the harmonic voltage at the tuned frequency. The adopted model of capacitor banks in the rest of the thesis is model of untuned Yg connected capacitor bank Modelling of voltage sag The assessment of voltage sags at a bus is usually based on the assessment of all recorded sag events at that bus over certain period of time. The number and severity of voltage sags are the measures of sag performance of that bus. This section presents a new single-event characteristics index, sag severity index (SSI), based on generally accepted and widely used voltage sag tolerance curves of equipment. The proposed SSI represents 97

98 the severity of voltage sags in terms of magnitude and duration based on equipment sensitivity. Voltage sag magnitude and duration are obtained from conventional shortcircuit simulations carried out in the DigSILENT/PowerFactory software package. Then, a new index, Bus Performance Index for Sag (BPI S ) is proposed to take into account sag frequency, magnitude, and duration, and form a single numerical value to represent the sag performance of a bus Sag Severity Index (SSI) The IEEE standard 1564 [69] recommends the ITIC and SEMI F47 [89] voltage tolerance curves as a reference for the severity of voltage sag events. Figure 3-12 (a) shows the SEMI F47 tolerance curve (the red lines), with the susceptible area to sag shaded with grey. The single voltage tolerance curve (deterministic) cannot fully represent the susceptibility of equipment. For the same type of equipment the sensitivity to voltage sag can vary between different manufacturers, different ratings and other factors [6]. Therefore, the probabilistic tolerance curves shown in Figure 3-12 (b) are adopted for the calculation of SSI. The probabilistic modelling is based on a number of voltage tolerance curves distributed normally around the original tolerance curve. Magnitude v (p.u.) Duration t (s) v3 v2 v1 Magnitude v (p.u.) Duration t (s) t1 t2 t3 T 1 T 2 T 3 (a) Deterministic Figure 3-12: Voltage tolerance curves (b) Probabilistic 98

99 For every recorded sag event, the sag magnitude (remaining voltage) and duration are used to determine the severity of the event. If the recorded sag is above the tolerance curve the SSI=0, otherwise the SSI is calculated by the following equation (3.3), y 1 C SSI w Bij = ( ax V max(t x ) v Bij ) + ay V max(t y ) v Bij V max (T x ) V min (T x ) V max (T y ) V min (T y ) x=1 b t Bij T min (T y ) (y 1) ( T max (T y ) T min (T y ) ) (3.3) Equation 3.3 calculates the severity of the j th sag event at bus Bi with respect to the randomly selected C w curve from the normally distributed probabilistic tolerance curves (green shaded area in Figure 3-12 (b)). The recorded sag event characteristics are the sag magnitude v Bij and the sag duration t Bij. The parameters T min, T max, V min and V max are the time and voltage boundaries of the region from the susceptible area where the sag event falls (Figure 3-12). The subscript y denotes the region where the sag event falls in, where the subscript x denotes all the regions before region y in the duration axis. T x and T y denote regions x and y duration ranges. The constant a is used to smooth the boundaries between regions and improve the continuity of the index, as shown in Figure 3-13, i.e. the constant a describes how the severity changes between regions. The constant b describes the increase in severity with increased duration within the region of the sag. To give an example, a recorded voltage sag with the characteristics v=0.3 and duration t=0.7 sec, will fall in the region with V max = 0.8 p.u., V min =0.1 p.u., T max =1 sec and T min =0.5 sec (from the original tolerance curve). The region where the sag falls is T y = T 3, and the regions below (T x ) are T 1 and T 2. By changing the reference curve C w the value of the SSI will change due to the new values of T min, T max, V min and V max. By running a large number of Monte Carlo simulations and by changing the reference curve in each simulation, the distribution of the SSI value for every sag event can be calculated. 99

100 High Colour contour map Magnitude v (p.u.) Magnitude v (p.u.) High Colour contour map Low Duration t (s) Low Duration t (s) (a) a = 2 (b) a=3 Figure 3-13: Heat map of SSI with different values for parameter a Bus Performance Index for Sag (BPIS) The bus sag performance is described based on two main aspects; the frequency of sags, and how severe each sag is. This can be performed by calculating severity of each sag based on the distance to adopted tolerance curves (SSI) and summing up all calculated severities. Due to the unavailability of sag records when studying the PQ performance for planning purposes, statistical historical records of faults are used to evaluate the bus performance probabilistically. As the faults in networks are the main cause of voltage sag [6], the performance index is calculated based on this type of sags only. To assess the sag performance of a bus, all the sags that can happen at the bus (due to faults in the neighbouring network components) are recorded by simulating faults in all possible locations, i.e. calculating sag tables. The type and frequency of faults vary between different components and different voltage levels. In this thesis, they are based on historical data, as shown in Table 3-4 [135]. The duration of sags predominantly depends on the primary protection operating times. In case of primary protection failure, the duration will be based on the backup protection operating time. Table 3-5 shows the main and backup protection average operating times adopted in this study and the probability of failure of the main protection. The BPIS is calculated based on the severity of every recorded sag (SSI) multiplied by the expected annual frequency of that sag, equation (3.4), to give annual bus performance in terms of sags. 100

101 BPI S Bi = f Bij SSI Bij (3.4) j BPI S Bi = sag performance index of bus Bi j f Bij SSI Bij = sag events counter = the frequency of occurance of sag event j at bus Bi = sag sevirty index of sag event j at bus Bi Table 3-4: Faults rate and types for sag tables calculation [135] Buses OHL Cables Fault rate (/year) Single phase to ground (%) Double phase to ground (%) Phase to phase fault (%) Three phase fault (%) Table 3-5: Fault clearing time for sag tables calculations Components Relays Mean (ms) Std (ms) Fault probability (%) Buses Primary % Back-up N/A Lines Primary % Back-up N/A Modelling of voltage unbalance The unbalanced sources in the network are the single phase connected DG units (PV and fuel cells) and a number of three phase loads with unequal reactive power injections per phase. The unbalanced loads are modelled assuming three ranges of power factor samples with average values of 0.8, 0.95 and 1 [77]. Normal distribution ranges of power factors are adopted, as shown in Figure Figure 3-14: Ranges of sampled power factor for unbalance simulation [77] 101

102 The distributions are selected to have the range of µ + 3σ covered by + 20% of the selected average, where µ is the mean and σ is the standard deviation of the normal distribution. In the cases of average power factor equal to 1 and 0.95, the values beyond the unity power factor are discarded, as shown in the figure. The phases of the unbalanced loads were assigned the normally distributed power factor ranges, with the maximum level of unbalance occurs when assigning different average value for each phase. The unbalance in the network, originated from the loads, was simulated by sampling the power factor ranges for each phase and based on the sampled power factor the reactive power of that phase is calculated. The real power of the unbalanced loads is kept constant. The unbalanced reactive power consumptions mainly affect the voltage magnitude unbalance. On the other hand, the real power unbalance in the network was originated from the single phase DG units connected at different phases in different locations. The unbalanced real power injections mainly affect the voltage angle unbalance. A probabilistic three phase load flow is the adopted analysis for the phenomenon, and the propagation of the sequence voltages and currents is measured. The voltage unbalance factor (VUF) is adopted as the main unbalance measure. Both definitions of VUF, based on zero or negative sequence, are considered (see Chapter 2). The VUF zero sequence definition needs to be considered only where zero sequence currents have available paths. The transformer winding connection can affect the propagation of the unbalance predominantly by altering the paths of zero sequence currents (zero sequence circuits of different transformer connections are shown in Appendix C). If the network is isolated from ground at the transformer, high neutral voltages might be recorded under unbalance conditions. To analyse the propagation of zero sequence components (due to both unbalance and harmonics) and to facilitate faster convergence of the three phase load flow under severe unbalance conditions, a zero sequence path was modelled at the medium 102

103 voltage level of the test networks (11 kv for the GDN, 15 kv for the real test feeder) by modelling a neutral fourth wire grounded at the star point of the substation transformer. The neutral fourth wire was modelled with the same conductor type and size of the three phases in the corresponding branch. 3.3 Summary This chapter presented the models and test systems used in all the studies in the thesis. Two test networks are presented. The GDN, a generic 295-bus network with different voltage levels modelled, and a real 35-bus 15 kv radial distribution feeder. The simulation methodologies and models adopted for simulating the considered phenomena were also presented. The harmonic models of network components that might affect the flow of harmonic currents are discussed based on characteristic case studies. For modelling voltage sags, two new indices are discussed. The sag severity index (SSI) is a probabilistic single-event characteristics index which describes the severity of sags based on the sag magnitude and duration. The second index, the bus performance index for sag (BPI S ) is a site index that describes the sag performance of a bus based on the severity of the recorded number of sags. The voltage unbalance is evaluated based on the propagation of the negative and zero sequence components from the source of unbalance to the network. The unbalance was measured by the voltage unbalance factor (VUF) calculated at all buses based on probabilistic load flows. The sources of the unbalance in the network were single phase DG units and three phase loads with unequal reactive power consumption per phase. This chapter presented two original contributions of the thesis: Contribution 1 the development of the simulation platforms for the PQ studies in networks with stochastic DGs; Contribution 2 the development of new probabilistic sag indices for characterization of sag phenomenon at network buses. 103

104 4 Harmonic Estimation in Distribution Networks In the contemporary and future distribution networks, the number and level of uncertainties are increasing, especially under the paradigm shift towards active distribution networks and networks with increased level of distributed generation (DG) and electric vehicles (EV). Uncertainties like the power output of these types of intermittent and stochastic generation, and uncertainties regarding the locations, owners behaviour and charging patterns and levels of EV, have a significant impact on the evaluation of PQ at a bus, feeder or a whole network. Locations of harmonic sources and harmonic injections are affected by those uncertainties, and it is very difficult and unrealistic to try to quantify them deterministically. Therefore, probabilistic evaluation of the harmonics phenomenon is inevitable. This chapter discusses the evaluation of the harmonics in two main parts; the first part is the probabilistic assessment of harmonics based on long term studies, using the 104

105 historically available knowledge of harmonic sources and expected levels of injections. The second part is the probabilistic harmonic estimation based on limited monitoring. 4.1 Probabilistic Assessment of Harmonics The general aim of harmonic studies is to analyse the network performance under different frequencies. This includes identification of the extent of the presence of higher harmonics in supply voltage, and potential harmonic resonance frequencies. The uncertainty of the output of the DG (and consequently their potentially different harmonic contribution) as well as the possibility of variable locations of harmonic sources (both DG and electric vehicles) must be included in the assessment of harmonic performance of a network as they may vary significantly during the year. This section presents a probabilistic methodology to model and study harmonic sources and harmonic propagation through the power network over specified time period. The uncertainties considered include the harmonic injections from diverse, but fixed location sources (renewable generation, i.e., wind, PV and other converter connected generation) and from variable location and injecting sources, e.g., random non-linear loads and EVs. Furthermore, these two are combined with known harmonic generation at known location by non-linear loads considering their daily and annual variation in harmonic output. The total voltage harmonic distortion (THD) is used as the main performance measure. The IEC [18] recommends that the evaluation of the emission should be performed statistically, to take into account the time variation of the phenomenon. It is a common practice to take harmonic measurements for at least one week, and to compare the 95 th percentile of the measured THD with the planning values specified in relevant standard. Although the common practice of day/week evaluation of harmonics may not be feasible any more for long term planning applications. This is mainly due to the high spatial (EV) and temporal (DG) variations in harmonic sources, with expected significant seasonal changes in the bus 105

106 harmonic performance. Therefore, annual variation of harmonic performance needs to be considered, which with the aid of classification techniques, representative hours selection and statistical measures can give good estimate of the annual performance without relying on brute force or time consuming simulation techniques. All the simulations in section 4.1 are performed in DIgSILENT/PowerFactory simulation package Models and Methodologies Load models The non-linear loads are modelled as current sources at different frequencies. For the studies performed in the GDN (150 load buses), 30 non-linear loads are selected, 10 at fixed locations while the remaining 20 are selected randomly at different buses in every testing hour. Similarly, for the real distribution feeder (35 buses), out of 17 harmonic sources, 7 are selected randomly every test hour while the other 10 are at fixed locations. The randomly selected sources at different hours are representing the uncertainty in the EV charging locations. Two classes of non-linear loads are adopted, the first is dominated by the 3 rd, 5 th, 7 th and 9 th harmonic currents, e.g., personal computers, TV sets, fluorescent lamps etc. This type of harmonic spectra is used for domestic and commercial non-linear loads. The second load class is adopted for the industrial loads, e.g., three-phase adjustable speed drives, the harmonic spectra are mainly dominated by the 3 rd, 5 th, 7 th and 11 th harmonic. Table 4-1 shows the values of current injections for individual harmonics for each class of load [87]. The random injections from loads are modelled in two ways; firstly as uniformly distributed ranges from 0 to the values in Table 4-1 (higher uncertainty), and secondly as normally distributed functions with + 20 % 3σ range around the mean values given in Table 4-1 (lower uncertainty), Figure 4-1 shows an example of 1 Amp load for Type 1 3 rd, 5 th and 7 th injection samples. These probabilistic representations are implemented to cater for the uncertainties involved in the harmonics injections of 106

107 composite loads, such as different operating conditions, different loading levels, ratio of non-linear load etc. Table 4-1: Non-linear loads harmonic injections levels [87] Harmonic order Type 1 (Domestic and Commercial) Type 2 (Industrial) 1 100% 100% 3 69% 4.7% 5 48% 32% 7 28% 16% 9 27% 0% 11 0% 6.5% Figure 4-1: Random harmonics injections (lower uncertainty) DG models Two different ways of modelling of DG for harmonic studies in the GDN network are adopted; the first one is to model concentrated penetration of DG, i.e. larger sizes (2-5 MW) in fewer locations with higher uncertainty in the harmonic injections (uniformly distributed ranges) (see Figure b). The second one is to model higher number, smaller sizes ( MW), more dispersed and less uncertain harmonic injections (normally distributed ranges) DG units, (see Figure b). In both cases, the DG penetration does not exceed 30% of the total feeder load at all hours of the year. Three types of DG are considered, wind generators, photovoltaic and fuel cells. The wind generators are modelled as three phase asynchronous generators of DFIG type. The fuel cells and small 107

108 photovoltaic generators (< 0.2 MW) are modelled as single phase static generators connected via converters. The annual hourly output curves for the wind and photovoltaic generators are extracted from realistic output data based on the UK weather (see Chapter 3) while the fuel cells are assumed to have a constant output throughout the year. Assuming same geographical conditions for all DG locations e.g. all PV units are assumed to have the same p.u. output during the same hours at all locations. The fuel cells are assumed to be available at full capacity during the whole year. All DG units work at unity power factor, with no voltage controllers applied. Similar to the non-linear loads, the harmonic injections of the DG are modelled as uniformly/normally distributed functions up to/around the values showed in Table 4-2 [64, 136]. Table 4-2: DG harmonic injections levels [64, 136] Harmonic order Wind Gen. PV/Fuel Cells 1 100% 100% 5 1.9% 0.16% 7 0.4% 0.18% % 0.12% % 0.11% Study periods The assessments of harmonics are performed for two study periods, annual and daily evaluation. The daily evaluations are performed by simulating 24 hours for a selected day, and the annual evaluations are performed by selecting a number of representative hours throughout the year. The selection of the representative hours for the annual evaluations is based on two criteria. The first criterion is by segmenting the load duration curves into 11 segments. The segmentation is performed for the three different load type duration curves, namely industrial, commercial and domestic, due to the distinct levels and injections of different types of loads. Corresponding to each type of load annual load duration curve (LDC) is produced. The LDCs are divided into 11 segments; the 1% peak period of the year (88 hours), then 9% segment (788 hours), and the remaining 9 segments 108

109 Load (p.u.) with 10% of the year each (876 hours). The median of each segment is taken as representative loading for the whole segment, and the corresponding hour is taken as a test point for simulation (33 testing points; 11 per each LDC). Figure 4-2 shows the domestic LDC segments. For example the median of the first segment of the curve is the 44 th hour, which corresponds to domestic loading of p.u. (this loading actually occurs in the 5635 th hour of the test year, which is at 18:00 on a Monday). At that hour the commercial and industrial loads are at p.u. and p.u. of their peak, respectively. Also for the Commercial LDC the peak segment is represented by the 44 th hour which has a p.u. value of of the peak (5510 th hour of the year, i.e., 13:00 on a Wednesday) which corresponds to domestic loads at 0.53 p.u. and industrial load of p.u. of the peak loading. The same testing hour selection is applied to the industrial LDC. This testing hour selection from separate LDCs is adopted to account for the effect of different types of loads on the harmonics injections. In selecting the test points from the LDCs, it must be ensured that they include weekdays and weekends, days and nights Hours Figure 4-2: Segmented domestic LDC The second criterion for selecting the testing hours for the annual evaluation is by segmenting the total distributed generation output curve into 10 segments to study the impact of DG levels on the harmonics levels. The selected testing hours are selected to cover the seasonality of the DG to a large extent. In order to isolate and measure the impact 109

110 DG output (p.u.) of the DG on harmonics the different load types are kept around the same values at all the testing hours. As shown in Figure 4-3, the DG output curve is segmented equally into ten segments. The maximum point of each segment is taken as a testing hour, for example, the first segment of the maximum DG penetration during the study year occurred at the 2773 rd hour of the year (Monday 1 p.m. in April), in this hour the output of the PV is 0.95 p.u., the output of the wind generators is 0.98 p.u. and the output of the fuel cells, as it is constant during the year, is 1 p.u. The minimum penetration of the DG during the year (Thursday 8 p.m. in November) is about 0.58 p.u. and it is contributed to by the fuel cells only Hours Figure 4-3: Segmented DG accumulated output curve Evaluation methodologies The harmonic probabilistic simulation is performed using DIgSILENT PowerFactory (v15.0.1). In the simulations all the sources, DG units and selected loads, are modelled as independent harmonic current sources. Each harmonic current injection (magnitude and angle), for each phase, are varied randomly by sampling the uniform/normal distributions within the given ranges using Monte Carlo (MC) simulations technique. The ranges are determined based on the documented harmonic performance of different types of DG and different types of non-linear load (see Table 4-1 and Table 4-2). The angles range is (0-180) degrees for all harmonic sources. Thus, a single harmonic simulation is performed by injecting random values of harmonic currents from fixed locations (DG/loads) and a number of randomly selected harmonic sources in the network 110

111 (loads). In the case of the segmented LDCs, fifty iterations per 10% segments, five per the 1% segments and 45 per the 9% segments are performed for the three different LDC per load type. This yields a sample size of 1500 simulation results, i.e., 3x5 (for the 1 st peaks) plus 3x45 (for the 2 nd peaks) plus 3x9x50 (for the remaining segments). In the case of segmented DG output curve, fewer testing hours are adopted to represent the annual performance, and only variation in the DG level is adopted. Therefore, the total annual sample size is doubled (3000 simulations, 300 at each testing hour) to compensate for the fewer testing hours. The results of the annual performance are samples of THD and harmonic voltages fitted into standard PDFs to facilitate comparisons, as shown in the next section. The annual evaluations are simulated on the GDN network only and the simulations are performed for the cases with and without DG connected to the network. The daily harmonic assessment method used is an hour-by-hour harmonic simulations for a single day with the same models of DG and non-linear loads as in the annual evaluation method. A summer day (high PV output) is selected and the simulations are run for the 24 hours. This assessment method is used in order to compare the annual performance and the daily performance of a bus and the effect of seasonality on the bus harmonic performance. Depending on the location, the load type and the impact of DG at the bus, some buses showed the same harmonic performance for annual and daily load variation. Based on these simulations the average values, 95 th percentiles and standard deviation of THD for each bus are calculated throughout the study day. The daily evaluations are performed for both GDN and real test networks Results of different case studies Two levels of evaluation can be performed to assess the harmonic performance for long periods, network (area) level evaluation and bus level evaluation. It is usually more practical to represent performances of areas and feeders first, and then identify smaller 111

112 zones or buses to perform more detailed analysis. In the next sub-sections the two case studies of the probabilistic harmonic evaluation results are presented. The areas performance is presented with the aid of heat maps and line charts, then based on the worst performing buses further analysis is presented. Analysis such as frequency scans and individual harmonic voltages comparison can be a computational burden, therefore only a selection of buses based on the area analysis can undergo such analysis without impacting the efficiency of the overall harmonic evaluation in general Case study I (concentrated DG penetration) For the network level evaluation, Figure 4-4 (a) shows the overall GDN network performance, i.e. all buses average and 95 th percentile THD combined results, where Figure 4-4 (b) shows all the buses performance separately (Note that the curves do not represent a continuous variable rather than facilitating the view of the THD values of all buses in one graph). Figure 4-4 (b) shows the average and 95 th percentiles of THD values for Phase B (the worst performing phase) for the 11 kv buses. Four curves are plotted to show the annual variation average before and after the connection of the DG and the annual variation 95 th percentiles before and after the connection of the DG. The negative impact of the DG connection is more pronounced for the group of buses ( ) where a jump in the 95 th percentiles from around 3% up to more than 13% is recorded, also the average THD increased to exceed 5%. On the other hand some buses showed better performance following the connection of the DG. Bus groups (1-11), (29-36) and (55-73) THD 95 th percentiles values are higher than 3% prior the connection of DG which reduced to around 1.7% following the connection of the DG, also for the same group of buses a slight improvement on the average THD is noted. 112

113 (a) Network harmonic performance (b) ) Buses harmonic performance Figure 4-4: GDN harmonic performance (Case I) Figure 4-5 (a), (b) and (c) show the most affected areas using heat maps. Figure 4-5 (a) shows the affected areas of the network before the connection of the DG based on average THD (all three phases). Figure 4-5 (b) and 6 (c) show the affected areas following the connection of the DG for the best performing phase (Phase C) and the worst performing phase (Phase B), respectively. Note that the scale for colour coding in these three figures is different due to different THD values ranges. It can be seen from Figure 4-5 (a), before the connection of the DG, that all the 11 kv buses have very similar performance, the average THD ranged between 0.5% and 1%, yet Substation L (the utmost right substation at the top of Figure 4-5 to which all the busses in the right hand side of the figure are connected) has the most affected buses. This area contains the three industrial loads and a high number of commercial and domestic load 113

114 (a): Harmonic performance without DG (max THD = 1%) (b): Harmonic performance of the least affected phase following DG connection (max THD = 3.7%) (c): Harmonic performance of the most affected phase following DG connections (max THD = 5.7%) Figure 4-5: Heat Maps identifying the most affected areas before and after DG connections (Case I) buses. Following the connection of the DG the area connected to Substation L is still the most affected area, the THD in this case however, ranged from 2.8% up to 3.7% for Phase C (Figure 4-5 (b)), while for the rest of the network buses the maximum average THD is 0.7%. Similar performance can be seen for the most affected phase in this case, phase B (Figure 4-5 (c)) though the THD in this case is higher and ranged from 3.5% to 5.7% while the maximum average THD for buses connected to other substations is 0.7%. 114

115 As far as the bus level evaluation is concerned, Table 4-3 shows annual harmonic performance of the three phases of the most affected buses in terms of the average and 95 th percentiles of THD after the connection of the DG. Table 4-3: The most affected buses annual harmonic performance (Case I) A B C bus Average (%) 95 th (%) bus Average (%) 95 th (%) bus Average (%) 95 th (%) From the tables another negative impact of the DG connection (especially the single phase DG) on the harmonic performance can be noticed, i.e., different harmonic distortion in different phases. The worst affected buses, in each phase, following connection of DG are 196, 136 and 225, for phases A, B and C respectively. The average values of THD for ten most affected buses range from 3.5% to 5.68%, and the 95 th percentiles range from 7.3% to 13.5%. There is a noticeable increase in the values of THD compared to the network without DG. The poorest performing bus (bus 196) has domestic and commercial loads connected and the latter two have only industrial loads connected. Single phase DG units are connected to all three buses, but interestingly not to the worst performing phase. Bus 196 has the largest harmonic distortion in phase A, yet the DG is connected to phase C. This demonstrates that harmonics performance will not depend only on the proximity to the sources but also on harmonic propagation from distant sources depending on the impedances of the network at different frequencies. 115

116 From the MC simulations, four samples of THD results are collected for each phase of each bus, i.e., annual harmonic performance with and without DG, and daily harmonic performance with and without DG. For the results of most of the buses in case of no DG connected it can be noted that the best fit PDF is skewed to the right, which match documented measurements of harmonic performance [17]. The histograms of the results with DG for majority of the buses are more symmetrical around the mean value, i.e. normally distributed. Weibull distributions are, therefore, used to fit the data as they have the flexibility of representing both, the skewed and symmetrical data samples. Figure 4-6(a) shows the PDFs of the THD for phase A of bus 196 for the four scenarios. The impacts of DG on harmonic performance of the bus can be clearly seen. Figure 4-6(b) shows PDFs of THD of phase A of bus 229 where the impact of the DG is negligible. Figure 4-7 shows CDFs of THD of the same buses (only annual variation showed). It can be seen from Figure 4-7 that the 95 th percentile of the samples for bus 196 exceeds the standard limits (5% for IEEE and 8% for IEC). However, the percentile measures are usually applied for shorter period of study with higher resolution of measurements (e.g. for 1 week with 3 minutes intervals of measurements [137]). Therefore, the comparison among the buses is performed taking the mean value of harmonics as a measure instead. (a) Bus

117 (b) Bus 229 Figure 4-6: Daily/Annual PDF THD results of phase A for bus (a) 196 (b) 229 (a) Bus 196 (b) Bus 229 Figure 4-7: Annual THD CDF results of phase A for bus (a) 196 (b) Case study II (dispersed DG penetration) The overall system annual performance in terms of THD, before and after the connection of DG, is shown in Figure 4-8 (a). As it can be seen, the overall impact of the DG can be considered insignificant in this case, the average of THD average values for all buses has increased from 1.5% to 2%, and the average of the 95 th percentile values of all buses has increased from 3.45% to 3.84%. However, the overall impact can be misguiding in judging the network performance as the increase in the THD is not evenly distributed between buses. Some buses are affected more than the rest of the network after the 117

118 THD (%) connection of the DG, showing an increase of the average THD up to 1.5% at some buses, and increase of the 95 th percentile values of THD up to almost 6%, Figure 4-8 (b) show these comparisons in terms of average and 95 th percentile THD. From Figure 4-8 (b) it can be seen that all the buses with exception of buses follow the same trend in the THD values before and after the connection of DG. For Buses the increase in the percentile values is much higher than the increase in the average value of THD. This is due to the fact that the percentile measure is more affected by the extreme values obtained during the MC process than the average measure. It can be also seen from Figure 4-8 (b) that some buses showed better performance after the connection of the DG. This could be mainly due to harmonic cancellation, as a result of adding higher number of sources IEC Limit = 8% IEEE Limit = 5% (a) Network harmonic performance average after DG 95th after DG Bus Number (b) Buses harmonic performance Figure 4-8: GDN harmonic performance (Case II) average before DG 95th before DG 118

119 In the annual and daily evaluation, the heat maps are used to present the overall network performance. It is worth mentioning the MC results show different buses rank between average and 95 th percentiles measures, and the percentiles in this case study is selected as measures. The heat maps in Figure 4-9 (a) and (b) show the percentiles values of THD for all the 11 kv buses before and after the connection of DG. (a) Performance without DG connected (max THD=8.3%) (b) Performance with dispersed DG (max THD=7.1%) Figure 4-9: Heat Maps identifying the most affected areas before and after DG connections (Case II) It can be seen that the feeder from substation L (the utmost right substation), in this case, shows similar performance before and after the connection of DG, while the rest of the network shows various level of increased THD. Also, to study the effects of the hourly variation results are presented at different hours of the study day. This method helps 119

120 tracking the load/dg types that pollute the network at different hours of the day, due to different usage of equipment and output level. For example, Figure 4-10 (a) shows the GDN network performance at 3:00 a.m., when the commercial and industrial loads are expected to be at the minimum with no PV output, and Figure 4-10 (b) shows the performance at 9:00 a.m. when the different types of loads are increasing with higher DG output is expected. (Note: The scale for colour coding in these figures is different due to different THD ranges.) (a) Harmonics Performance at 03:00 a.m. (b) Harmonics Performance at 09:00 a.m. Figure 4-10: Heat Maps for different hours during a day 120

121 Heat maps are also used to identify the areas where violations of the standard limits occurred, showing only the buses where IEEE standard [81] limits were violated (3% individual harmonic voltages and 5% THD). Figure 4-11(a) shows the areas where the 5 th harmonic voltage exceeded the limit, while Figure 4-11(b) shows the areas where the THD exceeded the limit. This way of representing the performance of the network facilitates easy identification of the critical harmonics and locations requiring attention. (a) V 5 above 3% (b) THD above 5% Figure 4-11: Heat Maps for the areas that violated the IEEE standard limits In terms of individual bus evaluation, Table 4-4 (a) and (b) show the worst ten performing buses at the 11 kv level, in terms of average and 95 th percentile THD, respectively. In this case study the rankings based on average and 95 th percentiles THD 121

122 values is different. The annual average THD is more consistent between phases while the ranking based on the 95 th THD percentile shows discrepancy between phases. Table 4-4: The most affected buses annual harmonic performance (Case II) (a) Average THD THD A B C bus % bus % bus % (b) 95th Percentile THD THD A B C bus % bus % bus % Analysis of case studies For long period harmonic evaluation, under the numerous uncertainties involved in the phenomenon, the probabilistic studies are the most suitable type of analysis. MC techniques, albeit computationally expensive, provide effective solution. With measures like reducing the study hours, analysing high and low uncertainties parameters and minimising the areas for detailed analysis, as presented in the two cases above, the MC simulations can be used more efficiently and the results can be used for benchmarking and developing mitigation solutions. Care must be taken though in selecting the appropriate statistical measures as the results may show discrepancy in rankings of the most affected buses (Table 4-4). 122

123 % Z (Ohm) From the overall harmonic evaluations, some poor performing areas may require detailed evaluation, for example areas with high penetration of power factor correction capacitors or cables might suffer from resonance problems as illustrated in Figure For developing mitigation solutions the problematic harmonic voltages must be identified first in order to identify the most probable sources of harmonics and appropriate filter design, Figure Bus 136 Bus 196 Bus Harmonic number Figure 4-12: Impedance frequencies scan for selected GDN buses (resonance near 13 th harmonic) (A) 137 (B) 137 (C) 0 V3 V5 V7 V9 Harmonics Orders V11 V13 THD Figure 4-13: Harmonic voltages and THD performance for selected GDN buses (Case II) 123

124 4.2 Probabilistic estimation of harmonics Harmonic State Estimation (HSE) is the process of establishing the levels of the harmonic voltages at all network buses based on a limited number of harmonic readings. The HSE for distribution networks is still at early development stage. The limited monitoring infra-structure, the limited knowledge of the status of some components (especially capacitors), the high variance and uncertainty in harmonics injections and locations, the high R/X ratio of branches and the radial or weekly meshed topology complicate the problem of harmonic estimation. The radial topology and limited number of monitors in addition to high variance of harmonics through long periods of study complicate the problem of optimum positioning of monitors to have full harmonic observability of the network. These aspects of distribution feeders usually lead to difficulties in getting independent readings, which in turn leads to ill-conditioned matrices for both estimation and monitors locating problems and causing, in general, solutions convergence issues. DSOs rely on the best possible estimate to identify if there are serious harmonic problems requiring more investigation, in such cases deploying more monitors and further analysis in the problematic areas are required. Another difficulty is the fact that some of the harmonic monitors are located at customer s premises (customer s side of the transformer), from which the readings will be impacted by the transformers impedance and connection, especially for high frequencies and zero-sequence frequencies evaluation [33]. However, on the positive side, the simple topology of distribution feeders lends itself to simplified calculations and the possibility of direct application of Kirchoff s Laws to gain useful information. High level of knowledge about radial distribution feeders harmonic performance can be achieved based on substation readings. While extra monitors and field measurements can help pinpointing sources and increase the accuracy, they are expensive and not always permanently located at distribution feeders [34]. 124

125 The next sections present two efficient methodologies for harmonic analysis and prediction in distribution feeders. Based only on the general knowledge of load types, network parameters and single monitor at the sub-station; available harmonic distortion capacity along the feeder can be estimated using developed look-up curves. Furthermore, the proposed probabilistic methodology can provide the THD and individual harmonic voltage levels in terms of standard specified indices over a given time period at all buses along the feeder with acceptable accuracy for network planning studies. The 35-bus real distribution feeder is used to illustrate the developed approaches. All simulations presented in section 4.2 are performed in OpenDSS and MATLAB environments Harmonics in radial feeders Based on the fact that the harmonic currents flow toward the substation; a rough estimate of THD levels can be produced based on the knowledge of the feeder parameters and the harmonic readings at the substation. This estimate of feeder harmonic performance can provide a look-up figure indicating the remaining THD capacity of the feeder based on the THD recorded at the sub-station. The THD capacity can be used in distribution planning considering PQ performance. As discussed in [36] for calculating the remaining load growth permissible before reaching harmonic limits, or in more recent studies as in [37, 62] to determine the level of permissible penetration of distributed generation. For capacitors free feeder with the assumption of no harmonic currents cancellation, the minimum THD of the feeder is always at the substation bus [138]. Figure 4-14 (a) and (b) show two hours of the daily evaluation of the test feeder, and it can be seen the worst performance buses always at the end of the feeder. Therefore, the THD along the feeder usually increases with the electric distance from the sub-station, i.e., with higher system impedance or lower short-circuit capacity, Figure The current practice for evaluating new connections under stages 2 and 3 of [24] is to calculate the THD 125

126 headroom at the connecting point and neighbouring buses. However with the adopted assumptions, listed below, for radial distribution feeders the THD vs. feeder length increases non-linearly with the electric distance from the substation, as illustrated in Figure (a) Harmonic performance at 12:00 (max THD=1.88%) (b) Harmonic performance at 21:00 (max THD=1.4%) Figure 4-14: Heat maps for real test feeder harmonic performance (daily study) Planning Level Remaining Feeder THD capacity THD room at the connection point THD Substation Feeder Length Connection point Figure 4-15: THD increases with the feeder length 126

127 Therefore the increase of THD at certain point will shift the whole curve up and most probably bringing the nodes further form the sub-station closer to the planning level. The following assumptions are adopted for the analysis: - The feeder is homogeneous; - It consists of similar load types; - The branches are of similar type; - The feeder nodes are relatively evenly distributed, i.e., no concentration of load at certain location; - There is no total cancellation of harmonics at any frequency Feeder harmonic analysis Characteristic buses The feeder is a real 35-bus medium voltage (15 kv) distribution feeder connected to the external grid through 60/15 kv Yy transformer (Chapter 3). For the one-day harmonic analysis presented in this section, it is assumed that all capacitors and filters are disconnected and the zero sequence path is available to ground by connecting sub-station transformer windings and loads as grounded star, to analyse the flow of triplen harmonics. Six characteristic buses are adopted for analysis; Bus 01, the MV bus of the substation (monitored), Bus 06, Bus 15 and Bus 28 at different locations on the main path, Bus 19 at the end of a small branch and Bus 35 at the end of the main path, see Figure Table 4-5 shows the fundamental frequency electric distance of the selected buses from the monitored bus, Bus 01. Table 4-5: Characteristic buses electric distance to Bus 01 (S/S) Bus Electrical distance (Ω) Bus Electrical distance (Ω) Bus j0.302 Bus j0.975 Bus j0.572 Bus j1.732 Bus j

128 Bus S/S Bus 01 Bus 06 Bus Bus 28 Bus 35 Figure 4-16: Characteristic buses locations on the test feeder Superposition case study The superposition case study is performed by equal harmonic injections at the characteristic buses (except Bus 01) one bus at a time, and the propagation of the harmonic distortion is studied. From Figure 4-17 it can be seen that the injected harmonics resulted in voltage THD value at the injecting bus that propagates un-attenuated towards the end of the feeder. This is simply because at the injection point the path towards the substation provides the low impedance path while the opposite direction is effectively an open circuit for harmonic voltages. The harmonic voltages drop in the direction of the substation is almost equal for all five cases and directly related to the network harmonic impedance. Therefore, the THD at the substation is the same for all the studied injection buses. Note the special case of harmonic injection at Bus 19 (red line-dot markers) as the bus is at the end of a branch (Figure 4-16) and not in the main feeder. After subtracting the THD drop in the branch the THD values at buses 15, 28 and 35 are equal. 128

129 Figure 4-17: THD values for superposition case study To study the propagation of the THD rather than separate harmonic voltages, the electric distance at different frequencies are aggregated. The total harmonic electric distance is produced by the weighted sum of electric distances at N considered harmonic frequencies and i buses. The h th harmonic impedance is weighted by the ratio 1/h, based on the expected harmonic currents flowing at that harmonic frequency, equation (4.1). The Z THD is normalized based on the maximum harmonic electric distance calculated for a bus, which is normally indicating the physically furthest connected bus from the substation for radial feeders. N Z THD,i = Z h,i h h=2 (4.1) By applying the concept of the total harmonic electrical distance for the combined results (average THD at the bus when injection is at the same bus) the linear relationship between Z THD and the THD propagation can be noticed in Figure The feeder electrical length is normalized based on the furthest bus, Bus 35, i.e., Z THD, 35 =11.13 Ω, see Table 4-6. Table 4-6: Total and individual harmonic electric distances from substation (Bus 01) in Ohm Z 1 Z 3 Z 5 Z 7 Z 9 Z 11 Z 13 Z THD Bus Bus Bus Bus Bus

130 Figure 4-18: THD vs. total harmonic electric distance Superposition case study Generic case study The generic case study is simulated to analyse the propagation of harmonics for known distorting loads. This is performed by adopting equal in-phase harmonic spectra for all loads using the theoretical values of convertors harmonic currents [80], equation (4.2) I h = I 1 h (4.2) Then by calculating the harmonic distortion for the characteristic buses for one day, considering the loading levels as the only varying parameter, the THD vs. total harmonic electric distance curve is produced. Figure 4-19 shows the minimum, average and maximum THD value recorded during the day plotted against harmonic electric distance Z THD. The relationship between THD and Z THD is not linear in this case due to interaction between different sources of harmonic currents, i.e., certain level of harmonic cancellation. In spite of injection at 0 phase shift from fundamental currents at each node, the harmonic phase angles at different summation points differ due to different harmonic impedances between the nodes. As shown in the figure, the THD tends to increase with the distance from the substation, also the variation range at the further buses is higher. This can be explained by the faster increase of impedance at the further buses with frequency as shown in Figure

131 Figure 4-19: THD vs. total harmonic electric distance Generic case study Figure 4-20: Buses electric distance for different harmonic frequencies General case study A general case study is produced for one day harmonic performance of the network, and used as a benchmark for the estimation. Two loading curves, for commercial and domestic loads, are adopted with a resolution of half an hour. At every half an hour three harmonic voltages and currents readings (representing a reading every ten minutes) are recorded, by sampling injections at all loading nodes from predefined average values of harmonic currents at each given node, and harmonic load flow is run. Every six hours the average values of harmonic injections and harmonic frequencies are varied representing new composition of non-linear loads connected at the bus. The first six odd harmonic frequencies are considered, i.e. 3 rd, 5 th, 7 th, 9 th, 11 th and 13 th, with injection levels varying between 0% and 90% of fundamental at different times and at different nodes. The THD 131

132 THD (%) has a range (for all buses) between 0.14% and 6.34% throughout the study day. Figure 4-21 shows the minimum, average, 95 th percentile and the maximum values of THD of the characteristic buses. Same as in generic case study the variation in the recorded THD increases with the distance from the substation. 7 6 Minimum Average 95th Percentile Maximum Bus 01 Bus 06 Bus 15 Bus 19 Bus 28 Bus 35 Figure 4-21: General case study results Harmonic estimation algorithms Coarse analysis The analysis is performed based on the generic case study presented in The curves of THD vs. harmonic electric distance from Figure 4-19 are smoothed and used to perform two types of studies: - The impact of increased injection at one node on the rest of the feeder. - A rough estimate of THD along the feeder based on a substation harmonic reading. The impact of the increased injections is performed in the generic case study. At Bus 06 the injection is increased to 2, 3 and 4 times the original injections, while the rest of the injecting loads are kept at the original values used in the generic case study. The one day harmonic evaluation is performed and the minimum, average and maximum THD values are plotted for the characteristic buses. The results of the impact of increasing the injection at certain point of the feeder on the whole feeder are shown in Figure As it 132

133 Figure 4-22: Impacts on harmonic performance of the feeder by increased harmonic injection at Bus 06 can be seen from the figure the increase of harmonic injection at one location has effects on the whole feeder with more pronounced impact at the buses further from the substation. The impact on the point of increased injections (Bus 06) is an increase in the maximum THD from 3.11% in the generic case study to 3.17% when the injection is doubled, 3.3% when the injection is tripled and 3.41% when the injection is fourfold the original injection, i.e., almost negligible considering the level of increase of harmonic injections. Comparing the impact at the substation and at the furthest bus down the feeder, shows the more pronounced impact at the weaker buses, although they are electrically further from the point of harmonic injection. Bus 01 shows an increase in the maximum THD from 2.03% to 2.22% between the original and the fourfold increase in harmonic injection while the furthest bus, Bus 35, shows an increase of the maximum THD from 5.09% to 5.52%. This shows almost the same ratio of increase, i.e., 10% of the original values at all buses. The second study is to perform rough estimate of the general case study performance based on the generic average curve of THD vs. electric distance. This is done by rescaling the THD curves from the generic case by a correction factor based on the THD values recorded for the general case study. Also, the variation level (measured by the coefficient of variation σ/µ) measured at the substation for the general case is used to estimate other indices, i.e. maximum and 95th percentiles of THD. The coefficient of 133

134 variation calculated at the substation (from the day measurements records) is assumed constant along the feeder. Based on the corrected THD values and the constant coefficient of variation the generic average curve is rescaled and the rough estimate is produced. Figure 4-23 shows the results of the coarse estimation (smoothed blue curves) compared to the true general case THD values for different considered indices. It can be seen that the coarse estimation results in very accurate estimation of the 95th percentile values. The accuracy of coarse estimation of maximum THD values is lower though, and in particular for buses further away from the monitoring point. The estimated maximum THD tends to be lower than the actual max THD. Figure 4-23: Coarse estimation for the general case Probabilistic estimation The harmonic levels prediction is performed based on the methodology described in [11, 33]. However, the main aim here is to estimate the levels of harmonic distortion rather than pinpointing the harmonic sources themselves. Based on the fact that all harmonic currents flow towards the sub-station [11] in normal conditions, one monitor readings at the sub-station should give a basic knowledge about the frequencies and the levels of harmonic currents flow throughout the feeder, assuming that there is no high level of harmonic cancellation or harmonics diverted to ground through capacitors or filters, Figure 4-24 shows an example where 5 th harmonic current is present at some of the feeder 134

135 branches and buses but might not be recorded by the substation monitor due to capacitor installed along the feeder. S/S I h Harmonic monitor I h I h I 5 I h I 5 I h1 I h2 Figure 4-24: A case of 5 th harmonic currents not seen at the substation The developed algorithm is based on Monte Carlo (MC) simulations of the injection levels at all nodes. For the MC simulation the sampling range and the number of simulations should be selected based on the expected variation of harmonic performance during the study period. A number of uncertainties are involved in simulation of harmonic levels. Uncertainties like loading levels, injection levels and angles affect the shape and the range of the samples of results of harmonic voltages and THD. For example, considering the generic case study (changing only load levels with all remaining parameters fixed) harmonic daily performance the THD variation during the day is approximately equal at the characteristic buses with a coefficient of variation σ/µ =0.35. This indicates daily variation range of more than + 100% of average value for 3σ covering 99.7% of observations. To perform MC based on such high range a high number of simulations is required. Therefore, to exclude loading variation from the estimation, predefined loading curves based on the knowledge of the day type (weekend/weekday) and the load type (domestic/commercial) are applied at different nodes, and the estimation function is modified to include time-stamped readings. By estimating a reading at a time, a range of + 30% of average with only 100 harmonic simulations yields estimation of 95 th percentiles of THD values with an average error of 6% for the characteristic buses. The flow chart of the developed methodology is shown in Figure The recorded spectra at the substation 135

136 are taken as the initial guess of the injections at all nodes, i.e., the ratio of harmonic currents to fundamental recorded at the substation is assumed to be relatively contributed to by all loads based on their size. These values are adopted as the mean values of the injection samples. Then, by randomly sampling normally distributed harmonic injections around the mean values with angles based on the monitored values, a number of harmonic load flows are run and the V h and THD values at all nodes are calculated. Start Load flow at each measurement to calculate load fundamental currents - Network model and parameters - Loading curves Identify the frequencies to be solved and injection nodes - Monitor readings - Suspicious nodes* set f=h first Sampling I h= f=h next Harmonic load flow compare to V h readings No error < ϵ or itr= max Yes I est h ={I h min error} - Estimated levels of injections at different locations No All frequencies? Yes HLF based on estimated injections - V h est at all nodes End *All loading buses are considered suspicious nodes Figure 4-25: Flow chart of probabilistic estimation methodology 136

137 The final estimate is produced by comparing the substation (monitor location) harmonic voltages V h obtained by every simulation with the recorded monitored values. The sampled harmonic injections resulting in the minimum absolute error at all frequencies are taken as the final estimated spectra. A final harmonic load flow is performed based on the final estimated spectra to establish the estimated harmonic levels at all nodes. The results of the one day probabilistic estimation are presented using cumulative distribution functions (CDF) and box plots for the 144 reading samples (6 readings every hour for 24 hours). Figure 4-26 shows the true (blue) and estimated (red) day performance box plots of the characteristic buses. As shown in the figure, the THD values at Bus 28 and Bus 35 are underestimated considering both adopted performance indices (average, percentiles and maximum values) and variation ranges (percentile range and total range). This is due to premature convergence of algorithm solution, Figure 4-25, i.e. error threshold is met at the substation before reaching best estimate at all buses, especially more distant buses. Figure 4-26: True/Estimated harmonic performance for the general case study The error of the estimate is expressed with respect to the true values. Figure 4-27 shows the estimation error samples, i.e., each reading error during the day is recorded and the sample is plotted as box plots. As it is shown in the figure, although average error for the 137

138 worst performing bus, Bus 35 is less than 12%, individual readings errors could reach up to 40%. This is mainly due to harmonic current cancellation, which leads to low current and voltage readings at the substation while higher levels of harmonics are present in the feeder. The harmonic performance CDFs of three characteristic buses are shown in Figure 4-28 for both true and estimated daily harmonic performance. The discrepancy between true (solid) and estimated (dashed) is the most evident (though still small) for buses 15 and 35 while for Bus 01 the CDFs are identical. Figure 4-29 gives closer look into 95 th percentiles values for THD and the 3 rd and 11 th harmonic voltages and corresponding estimation errors. Figure 4-27: Estimation errors for the probabilistic estimation Figure 4-28: CDF of harmonic performance for buses 1, 15 and

139 Figure 4-29: Comparison of 95 th percentile THD values and 3 rd and 11 th harmonic voltage values and corresponding estimation errors The main limitation of the proposed probabilistic estimation methodology based on substation readings is in cases where harmonic currents do not reach the substation. This can happen mainly when the lower impedance path is available through filters and power factor correction capacitors or when out of phase currents are injected at different nodes which leads to high level of cancellation. The latter is less expected to happen in radial feeders as suggested in [34], and also as suggested by the practical summation exponent (α) range of 1 to 2 in equation (4.3) [18] for summation of harmonic voltages. The expected phase angle between harmonic sources ranges between 0 (linear summation) and 90 (quadratic summation). α V h = (V hi In case of large capacitors or filters connected at the MV buses, that bus should and is usually monitored. The readings of the tuned harmonic at the capacitor bus in such cases can be distributed between all load buses based on their fundamental currents, similar to the methodology presented in Figure In other words, if the buses where significant harmonic currents are injected in the feeder are monitored, the same level of the accuracy 139 i α ) (4.3)

140 in harmonic estimation, as presented in Figure 4-29, can be achieved. This is discussed in detail in the next section. The accuracy also depends on the adequacy of modelling nonlinear sources as independent current sources which are not affected by the existing background harmonic voltages at the injection points, which is commonly acceptable for the THD levels in the range of 10% or less. The proposed methodology also may not be as accurate in estimation of propagation of triplen harmonics. In cases where a connected harmonic source (Yn connected load) injects triplen harmonics while the propagation of zero sequence currents along the feeder is blocked by delta connected service and substation transformers but the triplen voltages still propagate (see Chapter 3). This affects the methodology to probabilistically estimate the voltages based on monitoring currents only, as the monitor at the substation may show high level of triplen harmonic voltages with almost zero triplen harmonic currents Harmonic propagation with capacitors connected to a feeder LC circuits analysis In the case of large capacitors connected at distribution feeders, the inductance (L) of lines and transformers interacts with the capacitance (C) of the capacitors altering the total network impedance and behaviour at different frequencies. The interaction between (L) and (C) of the feeder might lead to cases of parallel and series resonance, depending on the components connections and topologies [80]. The simple LC circuit shown in Figure 4-30 is studied to present some of the expected behaviour of harmonic currents and voltages under parallel and series resonance. The circuit has constant voltage supply at all frequencies V s = 10 0 V, Z s = j 0.1 Ω, X C = 100 Ω and X L = 4 Ω. Equations (4.4), (4.5) and (4.6) represent the total, inductance and capacitance currents respectively. 140

141 Z S = j Ω I t V R I L I C V S = 10 /0 all f h X L = 4 Ω X C = 100 Ω Figure 4-30: LC circuit parameters jv s (hx L X C h) I t (h) = X L X C + X C X s h 2 X L X s + j(hr s X L R s X c h) (4.4) I L (h) = I t I C (h) = I t X C h 2 X L X C (4.5) h2 X L h 2 X L X C (4.6) In the parallel LC circuit the total current is either lagging the voltage source when the circuit is dominant by the L reactance or leading the voltage source when the circuit is dominant by the C reactance. In case of parallel resonance the total current is equal to zero when the C reactance and L reactance cancel each other, i.e. equal in magnitude but out of phase, acting as open circuit. In case of series resonance (the capacitance resonates with the supply impedance connected in series) high current with lower voltage will occur due to, the effectively, short circuit at the source terminals. In all cases the total current, in magnitude, is less than dominant L or C current in the circuit as shown in the Figure The resonance frequencies h r can be calculated by equation (4.7), to have the parallel frequency at h=5 (the capacitor interacts with the parallel reactance) and the series resonance frequency at h 32 (the capacitor interacts with the source series reactance), as shown in Figure X h r = c XL (4.7) 141

142 (a) Results for h between 1-20 parallel resonance at the 5 th harmonic (b) Results for h between series resonance near the 32 nd harmonic Figure 4-31: LC circuit currents and voltages To illustrate the impact of the capacitance under resonance and other frequencies on harmonic currents, a simple 3-bus network, shown in Figure 4-32 (a), is examined. The network has high short circuit external grid ( MVA), with a typical 33/11 kv substation transformer, two cables, one load bus and one capacitor bus. The network parameters are shown in Table 4-7. Table 4-7: Simple 3-bus network parameters Transformer rating (MVA) SC voltage (%) X/R connection S/S-B sub Yg/yg Lines rating (A) Z (ohm) B (us) neutral B sub B load j yes B load B cap j yes 142

143 Loads P (MW) Q (MVAr) connection Load Yn Capacitors Q (MVAr) loss factor connection Cap D The load injects two harmonic currents 5 th and 11 th, bus B load has resonance frequency near the 11 th, as shown in Figure 4-32 (b). The injected harmonic currents range from 10% to 50% of fundamental, the results shown in Figure 4-32 (c) and (d) are the harmonic currents injected by load (I_load), flowing into S/S (I_sub) and flowing into the capacitor (I_cap). Although the injected currents are equal (I_load is equal at both frequencies), due to excited resonance, magnification of the 11 th harmonic currents is noticed. The excited resonance does not only increase the value of the currents seen at the substation and capacitor node, but also faster increase in I_cap and I_sub with higher injection is noticed, as shown by the higher slope of Figure 4-32 (d) curves Capacitors impact on harmonic estimation To study the effects of large capacitors and resonance situations in the test feeder, capacitors are connected at different locations at the feeder. Different connections, tuned, untuned and detuned capacitor banks have different impacts on harmonic performance. The worst case scenario is the grounded Y capacitors connected at the end of the feeder (weakest bus). This type of connection affects even the zero sequence harmonics, and the high network impedance X s (weak bus) with large size capacitor connected (low X C ) might lead to resonance situations at low frequencies, as shown in equation (4.7). A total peak of 600 kvar capacitance is connected, a relatively large capacitor(s) compared to the feeder total reactive load. Two cases are considered, the first case is 6 capacitors 100 kvar connected at the characteristic buses and the second case is when having one capacitor bank 600 kvar connected at bus 35 (the end of the feeder). The 143

144 Current flow (A) Current flow (A) Z (ohm) S/S B sub B load B cap I_sub I_cap I_load (a) 3-bus network Harmonic number (b) Impedance Frequency scan at B load I_sub I_cap I_load I_sub I_cap I_load load spectrum (%) (c) I 5 results load spectrum (%) (d) I 11 results Figure 4-32: Simple 3-bus circuit resonance analysis average day displacement power factor for the first case is improved to 0.97 and for the second to 0.99 compared to 0.94 before capacitor switching. The capacitors MVAr output is considered continues variable to simplify the problem, i.e. the capacitor output is based on the reactive power loading curve at the corresponding connection buses. Frequency scans are performed for the characteristic buses for the two cases as shown in Figure

145 Z (Ω) Z (Ω) B 35 B 28 B 01 B 19 B Harmonic number (a) Single 600 kvar capacitor at Bus 35 B 35 B 28 B 01 B 19 B Harmonic number (b) Six 100 kvar capacitors at the characteristic buses Figure 4-33: Characteristic buses frequency scan with capacitors connected It can be clearly seen that the resonance phenomenon is not a local phenomenon, i.e. the resonance frequency is approximately the same for all buses (range between h=16.1 for Bus 01 and h=16.5 for Bus 35 for the second peak in Figure 4-33 (a)) but with different peak values, especially for the buses at the mid and end of the feeder. Therefore, regardless the actual location of the harmonic source, the resonance can be excited by injection in almost every bus in the feeder. The presence of capacitors in the feeder also affects the estimation based on the readings of the head of the feeder only. The harmonic currents seen at the head of the feeder are not the total harmonic currents injected by the sources, as the capacitors provide low impedance paths as well as the substation. For the first case, where small capacitors are located at different buses on the feeder and assuming that no resonance is excited, the probabilistic estimation can still yield accepted, although overestimated, results in terms of accuracy, Figure As shown in the figure, the THD, 9 th, 11 th and 13 th harmonic 145

146 voltages 95 th percentile are estimated with around 10-15% errors in the 95 th percentile values. The overestimation is as a result of not considering the harmonic angle information in the estimation, hence the magnitude of the current flowing to the grid (inductance current) is actually higher than the injection currents (total current) as explained in the simple LC circuit in This is in addition to the reduction in the fundamental current magnitude supplied from the grid due to capacitors compensation. Therefore, the initial average estimated spectra (I h, S/S /I fund, S/S ) is always overestimated. Figure 4-34: Estimated/true 95 th percentiles for THD, V9, V11 and V13 six capacitors (100 kvar) connected For the second case, when only 1 large capacitor is connected at a weak bus, the result of estimation based on head of the feeder readings is shown in Figure As it can be seen the errors reach more than 40% in THD estimation. This is due to two reasons. First high portion of the harmonic currents does not reach the substation. Second the resonance at the 9 th harmonic is excited. To improve the estimation performance, a second monitor is connected at the capacitor bus, recording readings of the harmonic currents flushed through the capacitor. The first educated guess of currents injected by the load (total current) can be done based on the difference between the substation readings 146

147 Figure 4-35: Estimated/true 95 th percentiles for THD, V9, V11 and V13 not monitored single capacitor (600 kvar) at Bus 35 (inductance current) and the capacitor readings (capacitance current). The initial spectra calculated and used as average for the probabilistic estimation is calculated as shown in equation (4.8) where: Spectra avr = I sub,h I cap,h 2 2 I sub,1 + I cap,1 I sub,h = substation monitor h th harmonic current reading, I cap,h = capacitor monitor h th harmonic current reading, I sub,1 = substation monitor fundamental current reading, I cap,1 = capacitor monitor fundamental current reading. 100 % for all h (4.8) Note that the dominator is an approximate estimate of the total load fundamental current, under the assumption that the fundamental current at the substation and at the capacitor point are perpendicular (i.e. due to highly MVAr compensated feeder the current angle at the substation is approximately 0 ). By including the extra monitor information, the 147

148 estimation for all harmonic voltages except at the resonant frequency is significantly improved. As shown in Figure 4-36 the errors in estimating the 11 th and 13 th harmonic s 95 th percentile voltages are improved to less than 5%. However, due to the resonance at the 9 th harmonic, the estimation at this frequency is underestimated regardless the extra information. Recording errors of around 30%, and this high error is subsequently transferred to the calculated THD, resulting in errors in the range of 20-30%. Figure 4-36: Estimated/true 95 th percentiles for THD, V9, V11 and V13 monitored single capacitor (600 kvar) at Bus 35 The inaccuracy in estimation caused by excited resonance can be solved by including current angle information in the initial guess of resonance harmonic spectrum. Equation (4.8) above can be rewritten in its full vector form as the following Spectra avr,hr = I sub,hr < θ sub,hr + I cap,hr < θ cap,hr I sub,1 < θ sub,1 + I cap,1 < θ cap,1 100 % at h r (4.9) where h r is the resonance frequency harmonic. The problem is that the angle information is not readily available for most of the distribution feeders, therefore, equation (4.9) is approximated as equation (4.10) below 148

149 Spectra avr,hr = I sub,hr I cap,hr /α 2 2 I sub,hr + I cap,hr 100 % at h r (4.10) where α is an attenuation factor to retune the initial spectrum at the resonance frequency. This is performed simply by running simulated generic (equal known injections at all nodes as in ) two or three iterations at the peak MVAr injection from the capacitor and then selecting the value of α based on the recorded underestimated error (see Figure 4-36). For example if the first run gives an underestimate error of 40%, the assumed harmonic current absorbed by the capacitor I cap,hr will be reduced by 40%, before performing the MC estimation. A flag is set to indicate the measurements throughout the study day that showed recordings for predefined resonance harmonic in addition to MVAr injections from the capacitor exceeding a certain threshold. This insures that the harmonic performance at these measurement times is actually affected by excited resonance before applying equation (4.10). The probabilistic estimation is significantly improved by considering retuned initial harmonic spectrum at the resonance frequency as shown in Figure 4-37, where the maximum error in the 95 th percentiles THD, for the buses under study, is less than 3%. Figure 4-37: Estimated/true 95 th percentiles for THD, V9, V11 and V13 monitored single capacitor (600 kvar) at Bus 35 (resonance considered) 149

150 4.3 Summary This chapter presented the harmonic evaluation methodologies. The evaluations are based on different case studies, considering different levels, types, sizes and distribution of DG units and EV locations. A number of uncertainties affects the harmonic evaluation due to high temporal and spatial variations of such harmonic sources in modern distribution networks. Monte Carlo simulation was therefore adopted to cater for different types of uncertainty, like harmonic injections, angle of injections, locations of sources, switching frequency, etc. The harmonic performance was analysed at network level, using heat maps and statistical measures. Also, bus level evaluation is presented using harmonic bar charts, PDFs, CDFs and numerical tables. Two study periods were considered; one day hour-byhour evaluation and annual evaluation based on selected hours of the year. The results of the different case studies showed that distributed generation (single phase generation in particular) could contribute significantly to the increase of harmonics in the network (beyond standard specified limits) in spite of reasonably small harmonic generation by individual DG units and a conservative DG penetration level of about 30% of total feeder load. The results show that even though the overall increase in harmonics in the network for some of the case studies is insignificant some buses experience more than 100% increase in THD. The increase of harmonic levels (measured by THD) is not confined to location of DG. Quite contrary, the most affected areas are those where DG harmonic contribution combines with existing contribution of non-linear loads or where harmonic injections by DG excite a harmonic resonance. The overall effect of DG seems to be the increase of harmonic levels across the network and in particular in the areas which are already exposed to higher harmonics due to the presence of non-linear loads. Furthermore, the presence of 150

151 DG increases the range of THD variation. MC simulations result in higher standard deviation of THD PDFs, i.e., highlighting the increase in the uncertainty of expected THD. The study demonstrated that the impact of DG on harmonic levels in the network needs to be studied by considering different types of both, DG and non-linear loads and different operating conditions during the year. Considering the estimated increase of THD in this study due to the operation of DG it is essential to use as realistic parameters (harmonic injection levels) as possible for both DG and non-linear loads. Due to high variability of the types of DG and non-linear loads and their location (EV can change location during the day, and DG effectively change location by, for example, not operating during part of the day/night even though they remain connected at the same bus) and operating conditions during the day and year the probabilistic studies are inevitable. Although computationally expensive, they can enable more realistic estimation of harmonic levels in the network, identify the worst performing areas and facilitate development of appropriate mitigating solutions. This chapter also presented simple and effective methodologies for estimation of harmonic propagation and analysis for radial feeders based on monitored substations and basic knowledge about the network parameters. The approximate (coarse) estimation can be performed by plotting look-up curves that help indicating levels of distortion based on electric distance to the monitored bus. The probabilistic estimation can provide the THD and harmonic voltage levels at all buses along the feeder with the errors in the worst case scenario, below 15% for 95 th percentile values of THD and below 10% for individual harmonic voltages. The accurate knowledge of harmonic levels at distribution level buses is still not required by electricity market regulators in many countries except for facilitating new load 151

152 and generation connections. With the current trend towards integrating more renewables at distribution level, proliferation of power electronics interfaced load and generation and storage technologies, and the increased attention for power quality performance; the ability to quickly assess harmonic performance of the network based on existing limited monitoring can prove essential for the planning of cleaner and more flexible distribution networks. In this context, the proposed methodology can be used very effectively as it provides a simple and cost effective way to indicate the level of DG penetration, load growth and new load connections that might lead to violation of harmonic planning levels. It can also identify, based on simple estimates and look up curves, locations along the feeder where further more accurate harmonic analysis should be performed. The impact of capacitors presence and the effects of resonance conditions in the estimation are also considered, and solutions to improve the accuracy under these situations are discussed. In this chapter, two further original contributions of the thesis are presented: Contribution 3 the long term probabilistic estimation of harmonics in distribution networks with DGs; Contribution 4 the new method for estimation of harmonics in radial feeders based on limited monitoring. 152

153 5 Global PQ Evaluation Indices A high level of delivered Power Quality (PQ) is becoming one of the key performance indicators for both contemporary and future power networks. The increased proliferation of converter connected generation and load in power networks, increased sensitivity to network disturbances of some of these new types of devices and requirements for more flexible operation of power networks have led to the revision of some PQ standards and to the introduction of modified, or in some cases, new requirements for PQ compliance. DSOs have also started to pay significant attention to the levels of the PQ provided in their networks for many reasons, such as, the current deregulated electricity markets, increased DG penetration and new regulatory frameworks. Although almost all PQ phenomena, with the exception of voltage transients, are well defined and although appropriate thresholds for individual phenomena are set in international standards, there is no standardized nor commonly accepted way to describe and evaluate the overall PQ performance of network buses. 153

154 This chapter presents two methodologies for overall PQ evaluation; a methodology to comprehensively evaluate the PQ for a network considering both the suggested planning levels set by the utility and the different PQ requirements expected by the customers, is presented. The suggested PQ reserve (PQR) index is calculated using numerical consolidation of several separate indices. The PQR is applied to evaluate the voltage harmonics, unbalance and flicker combined performance. In addition, an analytic hierarchy process (AHP) inspired methodology to calculate an overall PQ index is proposed. The overall PQ performance of the bus with respect to voltage sag, harmonics and voltage unbalance is presented using newly defined Compound Bus PQ Index (CBPQI). Both methodologies are illustrated in the GDN distribution network. 5.1 The need for unified PQ indices In the past, different voltage and current disturbances and variations were dealt with separately and under different names, but in the late 1980s were brought together under the common name Power Quality (PQ). Since then, PQ has become one of the most talked about and analysed performance indicators of power networks. This can be broadly attributed to the re-regulation of the electric power industry and to the introduction of competition in the electricity generation and supply business (occurring at approximately the same time), to increased customer awareness about different PQ issues (that has led to industrial process interruptions and consequential financial losses) and to the introduction of highly sensitive equipment and devices in customer plants and premises. Since then, both utilities and customers have started to pay much more attention to different phenomena and disturbances brought together under the common name of PQ (see Chapter 1). Even though the common name has existed since the late 1980s, different phenomena have continued to be addressed individually by the utilities and customers. Many customers, those having very sensitive industrial processes in particular, have tried to 154

155 tackle the disturbances they are exposed to locally without much interest in how their equipment or solutions impact the system [11]. More recently, it has been agreed between all the parties involved, from the producers and suppliers of electricity to the end users that PQ in general can be addressed using system approaches rather than individual solutions. Nevertheless PQ is still considered a consumer-driven issue and the main concerns, when studying this issue, are the compatibility between the customers equipment and PQ disturbances originating in the network. These disturbances often lead to equipment and process/customer-activity misoperation or complete interruption resulting in very high financial losses. Utilities are trying to minimize their costs while respecting different customers needs for certain level of PQ. The Council of European Energy Regulators considers the assessment of the impact on customers as the main reason for monitoring and regulating PQ [10]. All the main PQ phenomena, voltage regulation, voltage flicker, voltage sags and swells, harmonics, over and undervoltages and transients (to a lesser extent though) are well defined in the international standards. Indices for evaluating, methodologies for calculation and measuring the phenomena and other characterizing parameters are usually defined in these standards. For example, harmonic performance planning and compatibility levels are described in IEC and IEEE 519, the voltage unbalance calculations and emission limits are described in EN and IEC , also the voltage flicker planning and compatibility levels can be found in IEC and IEC and the voltage sag indices and immunity levels can be found in IEC and IEEE The compatibility levels defined in international standards also serve as a guideline for the network operators to deliver PQ below these limits, and for the equipment producer to ensure immunity level above these limits, see Chapter 2 for further details. 155

156 In spite of all the extensive past work and efforts in the PQ area, there is still no standard way to describe the performance of a bus or a network in terms of overall PQ performance. In general, different parts of the network exhibit different types and levels of PQ disturbances and have different customers (with equipment having different immunity levels) connected to it, so the process of comprehensive PQ evaluation is not trivial. Even though different equipment has different sensitivity thresholds to different PQ disturbances and even these individual limits can be modelled, there are often many types of equipment used in customer premises whose individual thresholds are different. The overall characterisation of PQ at the point of connection of different customers, at least as global indication of delivered PQ, would be highly beneficial both from a customer s prospective and for the purpose of network PQ performance assessment and benchmarking. Once the global PQ performance is assessed, it would be easier to develop cost effective mitigating solutions at network and customer level. 5.2 PQ reserve index (PQR) An index based on the normalization and numerical consolidation of different PQ phenomena reserves is proposed in this section. The methodology proposed here takes into account the perspectives of the utilities regarding the suggested accepted levels (planning levels) and customers concerns regarding the different required levels of performances (immunity/compatibility levels). By comparing the planning levels with spatially probabilistically varying immunity levels, using heat maps, the areas of expected inadequate PQ can be pinpointed. The methodology is demonstrated using the case study of the GDN network considering voltage harmonics, unbalance and flicker as the phenomena of interest. 156

157 5.2.1 The Framework The general framework of the proposed methodology is shown in Figure 5-1. The overall PQ index is calculated in three steps: - PQ measurements/simulation: in this step the different PQ phenomena are analysed separately, by using the indices proposed in the standards. For each phenomenon, the indices are averaged over short periods (e.g. 10-min averages), then for a longer study period (e.g. a week) the 95 th percentiles are taken as the measures of the phenomena, as suggested in the standards. Harmonics Meas. THD PQ Measurements/Simulations Unbalance Meas. VUF Flicker Meas. Meas. taken over the same period Averaged every 10 m 95 th percentiles taken to represent the period P st PQ Evaluations and Bus Sensitivities Compare PQ levels with Thresholds Determine Phenomena weightings based on planning priorities Thresholds Weighting Factors Standards/ Customers requirements Planning Priorities - Rank Buses - Compare Areas - Id Worst Area... PQ Overall Evaluation Consider priorities of phenomena Compare to bus sensitivity PQR calculation Numerical Consolidation Figure 5-1: Framework for the overall PQ evaluation using PQR - PQ Evaluations and Bus Sensitivities: in this step the separate phenomena performances are compared to predefined thresholds, either from the standards or those based on customers requirements. Then, the different phenomena are weighted for the overall evaluation based on the utility s planning levels. The planning levels can be suggested based on the importance of the bus or the bus dominant equipment type, e.g. for three phase induction motor dominant buses (customer type) the most weighted phenomenon 157

158 will be unbalance. They can also be selected based on expected customers financial losses (customer sensitivity), or the levels can be set based on expected disturbances (increase of DG or EV will impact certain PQ phenomena). - Overall PQ evaluation: in this step the performances of different phenomena are combined considering different weights, to come up with a single index for the overall PQ performance. Different mathematical techniques are available for combining the indices (see Chapter 1). For the PQR, the index is calculated using numerical consolidation Planning levels The planning levels reflect the utility perspectives regarding the different PQ phenomena. They are usually less or equal to the compatibility levels. Planning levels are difficult to determine as they can vary from network to network based on importance or regulators codes. Therefore, they are only indicated in the standards. To illustrate the proposed methodology, the selected planning levels are uniform for all network buses and arbitrary selected to be 2.5% Total harmonic Distortion (THD), 1.6% Voltage unbalance Factor (VUF) and 0.7 p.u. Short term Flicker index (P st ) for the harmonics, unbalance and flicker, respectively. These planning levels, when compared to standard compatibility levels, i.e., 5% THD, 2% VUF and 1 p.u. P st [60] show the planned performance distance from thresholds, i.e. PQ reserves. By applying the general equation for reserve calculation [45] equation (5.1) r = (1 m ) 100 (5.1) g where r is the PQ reserve, m is the actual performance and g is the assigned threshold., the expected minimum reserves for the considered phenomena are 50%, 20% and 30% for harmonics, unbalance and flicker respectively. Based on this selection, it can be concluded 158

159 that the harmonics seems to be the most important phenomenon for the utility, as shown by the selected conservative planning level, or high planned reserve, for the phenomenon. The ratios of these planned reserves to the sum of reserves of the considered phenomena are adopted as weighting factors for the phenomena when evaluating the overall performance. This gives the following weighting factors w har =0.5, w unb =0.2 and w flk =0.3 (e.g w har =50/ ) Immunity levels (Thresholds) Due to the variations in the load sensitivities, the PQ disturbance immunity levels may vary between different types of customers. Furthermore, although the suggested compatibility levels in the standards are usually adopted by equipment designers and manufacturers, there will be still uncertainties and variations in the equipment behavior in practical applications. For example, for the same equipment type, some can sustain normal operation beyond the threshold while others may trip or mal-operate under levels below the thresholds [6]. Therefore, the thresholds are best described on a probabilistic basis. In the proposed methodology, the immunity levels (thresholds) are varied between different buses randomly, suggesting that loads have different requirements (spatial variation of thresholds). However, the assigned values are constant throughout the study period, without considering the uncertainty of the threshold itself or the temporal variation, for the sake of simplicity. For the harmonics and unbalance phenomena, thresholds are sampled from normally distributed ranges. The distributions have mean values at the standard compatibility levels, i.e., 5% THD (IEEE 519 [14]) and 2% VUF (EN [60]) with standard deviations of 0.33% THD and 0.133% VUF which are selected to have 99.7 % of the distribution covered by +20% of the mean, as shown in Figure 5-2. For the buses under evaluation the VUF immunity levels range between 0.9% and 3.1% while the THD 159

160 immunity levels range between 3.1% and 6.6%. For the flicker, a uniform immunity (equal to the standard compatibility level) is adopted, i.e. P st = 1.0 p.u. (IEC [4]). Figure 5-2: Harmonics and unbalance sampled thresholds levels (THD red/dashed, VUF blue/solid) Calculating PQR Based on annual PQ simulations of separate PQ performances (see chapters 3 and 4), and by applying equation (5.1) using the 95 th percentiles of the considered indices, the reserve of each phenomenon is calculated, considering that the thresholds g i in equation (5.1) vary at different buses. By comparing the calculated reserves for the considered phenomena, the PQR index is calculated differently for each of the following two cases. In the case of no exceedance (+ve reserves), the minimum reserve is selected as the index for the PQ performance. In the case of ve reserve(s), the index is calculated by taking the weighted average of the exceeding phenomena reserves, equation (5.2). The weighting factors are selected based on the planning reserves as illustrated in min r j,i all r j,i > 0 PQR i = { j w j,i r j,i w r j,i < 0 (5.2) where r j,i is the PQ reserve at bus i for phenomenon j and w j,i is the weighting factor of phenomenon j at bus i. For example, for a bus with THD = 3%, VUF = 1% and Pst = 0.8 pu, all reserves are positive and can be calculated by (5.1) to give r har = 40%, r unb = 50% and r flk = 20%. In this case the PQR of the bus is the minimum reserve, i.e. 20%, as a result 160

161 of the flicker performance. Another example is a bus with THD = 5.5%, VUF = 2.5% and Pst = 0.5 pu. In this case harmonics and unbalance exceed the limits so the PQR is calculated based on the weighted average of the exceeding phenomena. The harmonics and unbalance reserves in this case can be calculated by (5.1) to yield r har = -10% and r unb = - 25%. Recalling that the weightings factors are calculated based on the planning levels and equal to w har =0.5 and w unb =0.2, the PQR can now be calculated using (5.2) PQR = ( ) + ( ) 0.7 = 0.14 pu As shown in the example, in case of exceeding phenomena, the PQR is always negative and is calculated based on the violating phenomena only; while in case of all phenomena above the limits the PQR is based on the worse performing phenomenon only PQR results and applications Different heat maps of the GDN are used to present the results and applications of the methodology. Figure 5-3 shows a heat map that identifies the sensitive buses, i.e., the buses where performance at the planning levels is not adequate for certain phenomena. The map is plotted by applying the proposed methodology (PQR index) under the assumption of uniform performance (equal to planning levels) for all the considered phenomena at all buses. As can be seen from the figure some pockets of negative reserves are dispersed all over the network, due to the randomness of the selected variable thresholds. Figure 5-3: Sensitive areas 161

162 Figure 5-4 (a), (b), (c) and (d) show the overall, harmonic, flicker and unbalance performances respectively. It can be clearly seen that the overall performance follows the trend of the harmonic performance, as a result of the high weight of the phenomenon. Furthermore, the circled group of buses shows a good example of the performance of the index, and its capability of considering the points of view of both utilities and the customers. Due to the high weight of the harmonics but relatively good performance, and the low weight of unbalance but very poor performance coinciding with relatively high customer requirements (buses B_35 and B_36, see Table 5-1 (a)), the average overall index is recorded in the yellow zone. However, some buses with moderate unbalance performance when averaged with good harmonic performing areas are totally masked and pushed up to the green areas, i.e., acceptable overall PQ performance. Heat maps are efficient visualization tool for both the PQ performance and sensitivity results. They facilitate the quick and easy identification of weak performing and sensitive areas in the network. Heat maps are first used in [66] to represent PQ performance and they are currently available in some of the modern commercial power system simulation packages for different application. From the numerical results shown in Table 5-1 (a) and (b), the worst performing buses with the recorded reserves (in p.u.) are ranked when evaluated separately and overall. In the overall ranking, Bus 219 appears third worst, although it has not been recorded as one of the worst performing buses in the separate evaluation. This is due to its average performance in the two important phenomena, harmonics and flicker. 162

163 (a) Overall PQ performance (PQR reserve) (b) Harmonic performance (THD reserve) (c) Flicker performance (P st reserve) (d) Unbalance performance (VUF reserve) Figure 5-4: Heat maps for overall and separate PQ GDN performance (Note: The heat bars have different ranges and colour for 0% reserve) 163

164 Table 5-1: Worst performing buses based on separate and PQR rankings of reserves (in p.u.) (a) Separate evaluation ranking Harmonics Flicker Unbalance Bus r har Bus r flk Bus r unb B_ B_ B_ B_ B_ B_ B_ B_ B_ B_ B_ B_ B_ B_ B_ (b) Overall evaluation ranking Harmonics Flicker Unbalance PQR THD Thrs.* r har P st Thrs.* r flk VUF Thrs.* r unb B_ B_ B_ B_ B_ *Threshold The main application of the index is the comparison between buses and areas based on overall PQ performance considering different phenomena with different levels of importance from the utilities and customers perspectives, sometimes conflicting perspectives. The utility might be interested in planning high reserves for certain phenomena to meet expected levels of disturbances, e.g., having high reserves on harmonics and unbalance will allow higher penetration of single phase, converter based DG units without standards violations. On the other hand, the sensitive equipment at end user facilities determines the customers perspectives. Due to averaging and weighting, the index does not describe the exact reserve level of the worst phenomenon at individual buses and cannot be used for analysing a single phenomenon or identify the exact level of exceedance. Another application of the PQR is the optimization of PQ solutions. The selection of a single index to optimize PQ performance in the network will significantly reduce the complexity of the optimization problem. Also, by pinpointing the areas where higher PQ requirement coincide with poor PQ performance, the complexity of the optimization problem for PQ mitigation can be reduced. Figure 5-5 shows a comparison between the sensitive areas map (Figure 5-3) and the overall performance map (Figure

165 (a)), indicating the areas which will most probably need to be prioritised in the PQ reinforcement planning. Although it is convenient and simple to pinpoint problematic areas in this way, it is more of an indicative way for further analysis. For example, B_210 is in one of the overlapping sensitive/poor overall PQ areas in Figure 5-5; however it appears in a sensitive area due to tight unbalance thresholds and in a poorly performing area due to poor flicker performance. It is inevitable to have some masking or overestimating of the overall PQ when combining indices in PQR, however, the main advantage of reducing the problem and the amount of information required from 233 buses to less than 20 buses can significantly improve the efficiency of the PQ evaluation problem. B_210 Figure 5-5: PQ sensitive and poor performing overlapped areas 5.3 Compound bus PQ index (CBPQI) This section presents the second methodology to evaluate the overall PQ performance of the network. The Compound Bus PQ index (CBPQI) is calculated using Analytic Hierarchy Process (AHP) model. The PQ phenomena selected for illustration of this index and assessment methodology are harmonics, unbalance and voltage sag, showing that this index is capable of considering both event-type and continuous-type phenomena. Critical states which are derived from thresholds specified in standards or user requirements are incorporated in the AHP while developing CBPQI. The presented 165

166 approach allows the comparative scores of actual state and critical states to be combined and a numerical score which represents the overall PQ performance to be produced. Different weights are considered for different phenomena, as well as, for different characteristics of a single phenomenon. Therefore, the proposed index has added flexibility in that it considers different levels of detail in the evaluated phenomena. The application of the CBPQI is also illustrated on the GDN network Analytic Hierarchy Process (AHP) The AHP is one of the common mathematical models for multi criteria decision making (mcdm) problems. It solves the problem of selecting a goal from a number of alternatives based on a number of selecting criteria. Different selection criteria will have different weights on the final decision. Each selecting criterion can also have a number of sub-criteria, which again can have different weights in the main selecting criterion. Based on the different weights, each criterion has a different priority on the final decision. The alternatives have different scores for each selecting criteria, and, based on the criteria relative priorities, the final score will be given to the alternatives and the final decision will be made. Figure 5-6 shows the block diagram of the AHP model to solve a general problem of hiring a new manager. As shown in the figure, the selection criteria are qualifications, personal skills and experience. Each criterion will have a pre-defined weight on the final selection. The alternatives Candidate 1, 2,, n will have different scores in each selecting criterion, and based on the weighting of the different criteria, the total score for each candidate will be calculated. The candidate with the highest score can be selected as the final goal. More details about the mathematical models can be found in [139]. 166

167 Level 1 Goal Select Manager Level 2 Criteria Qualifications Personal Skills Experience Level 3 Alternatives Candidate 1... Candidate n Figure 5-6: AHP model for manager selection example The AHP structure is modified to accommodate the CBPQI calculation, as shown in Figure 5-7. The CBPQI calculation is based on a comparison of PQ performance of any bus Bus i and the specified reference bus Bus Reference (all considered PQ phenomena within specified limits) at the alternative level. Level 1 Goal CBPQI Level 2 Criteria Voltage Sag Voltage Unbalance Harmonics Level 2.1 Subcriteria THD V h Level 3 Alternatives Bus i Bus Reference Figure 5-7: AHP model for calculating CBPQI The comparison at the alternative level is performed using a pair-wise comparison matrix for each phenomenon, and taking the principle eigenvector absolute normalized values as the relative scores for the buses/phenomena. Table 5-2 shows the construction of the comparison matrix for the unbalance, where Score i and Score Threshold are the measures of how far (above or below) Bus i performance is from the threshold. In the case of performance at the threshold Score i and Score Threshold are both equal to 0.5, while in the 167

168 case of limit violation, Score i will be greater than Score Threshold. In this way, the critical states derived from the thresholds specified by standards, or user requirements, are included in the evaluation and contribute to the final comparative scores. It is believed that the inclusion of standard specified thresholds in PQ evaluation is essential for keeping the methodology as relevant to industrial practice as possible. Table 5-2: Comparison of alternatives pairwise matrix Unbalance BUS i BUS Reference Eigenvector BUS i 1 VUF Threshold /VUF i Score i BUS Reference VUF i /VUF Threshold 1 Score Threshold The score calculated in the first step for each phenomenon for the bus is multiplied by the priority of the phenomenon from the criteria level. At the criteria level, the priority of each phenomenon is calculated based on weighting factors assigned to each phenomenon. Again, the different phenomenon priorities are calculated from the principle eigenvector of the pair-wise comparison matrix, Table 5-3 shows the priorities comparison where w sag, w har and w unb are the weighting factors of the voltage sag, harmonics and unbalance respectively. Table 5-3: Comparison of priorities pairwise matrix Sag Harmonics Unbalance Priorities (Eigenvector) Sag 1 w sag /w har w sag /w unb priority sag Harmonics w har /w sag 1 w har /w unb priority har Unbalance w unb /w sag w unb /w har 1 priority unb In the same manner, at the sub-criteria level, different weights are assigned to different sub-criteria. For example, the different harmonic voltage characteristics (e.g. negative/positive sequence, high/low frequencies, near/far from resonance) will have different impacts on the sensitive loads, even if they have equal THD [140]. Therefore, if two buses have similar THD, but one of them, for example, contains harmonic orders close to the system resonance frequencies, it should be ranked as more critical in terms of 168

169 harmonic performance as it is more likely to have more severe impact. Similarly, different weights can be considered for other harmonic indices that are more relative to certain loads, e.g., zero crossing for electronic clocks and contactors or the Crest Factor (V peak /V rms ) for considering the impact on insulation. Table 5-4 shows an example of calculation of the sub-priorities for harmonics. The overall priority (weighting) of the harmonics is composed of the THD and the considered number of harmonic voltages V n. Table 5-4: Comparison of sub-priorities pairwise matrix THD V 5 V 7 Eigenvector THD 1 w THD /w V5 w THD /w V7 subpriority THD V 5 w V5 /w THD 1 w V5 /w V7 subpriority V5 V 7 w V7 /w THD w V7 /w V5 1 subpriority V7 The total score for Bus i is calculated as the sum of the multiplications of the scores of the bus in each criterion (sub-criterion) by the priorities (sub-priorities) of the phenomenon. Then, the CBPQI i is calculated by comparing the total score of Bus i to the total score of Bus Reference, as shown in equation (5.3) where N is the total number of considered phenomena, CBPQI i = N n=1 N n=1 score i,n priority n score Threshold,n priority n (5.3) In equation (5.3), the overall contribution of a phenomenon having sub levels will contribute to the CBPQI i by considering each sub-criterion index weighted by the subpriorities, and then by the global priority of the phenomenon. Equation (5.4) shows the example of Bus i harmonic performance contribution to the CBPQI i, with the THD, V 5 and V 7 representing the sub-criteria considered in the evaluation of Bus Har i. Bus Har i = (score THD i subpriority THD + score V5 V7 i subpriority V5 + score i subpriority V7 ) priority har (5.4) 169

170 For conventional AHP methodology, the scores calculated by N n=1 score i,n priority n are taken as the final results and used to compare the PQ performance among different buses/alternatives, which is acceptable if only one set of thresholds is used as the reference for all alternatives. However, if the critical thresholds are set to different values based on spatial requirements, conventional AHP methodology will be insufficient. Equation (5.3) provides a way to solve this issue. It incorporates the comparative results of two states, i.e. critical state and standard state, into one numerical index which represents how good the actual PQ performance is compared to the corresponding critical state that is obtained from the thresholds specified by standards or user requirements locally. The framework, inputs and outputs for the overall methodology are summarized in Figure 5-8. As shown in the figure, designated PQ thresholds at certain buses can be used as inputs to the model in lieu of the standards compatibility levels. This is to consider the types of customers for which the common compatibility levels specified in standards are not adequate. These customers typically have highly sensitive equipment/processes which would still underperform, in spite of meeting PQ threshold standards. PQ Measurements/Simulations Harmonics Simulation THD V h Unbalance Simulation VUF Sag Simulation BPI S PQ Evaluations and Bus Sensitivities Compare PQ levels with Thresholds Select Phenomena weightings Thresholds Standards/ PQ Agreement Meas. taken over the same period Averaged every 10 m 95 th percentiles taken to represent the period Weighting Factors Bus Load Equipment - Rank Buses - Compare Areas - Id Worst Area... PQ Overall Evaluation Consider priorities of phenomena Compare to Reference Bus CBPQI Calculated AHP Model Figure 5-8: Framework for the overall PQ evaluation using CBPQI 170

171 Numerical example of CBPQI calculation Consider three phenomena a, b and c. Phenomenon a is 1.5 times more important than phenomenon b, and 3 times more important than phenomenon c. In the case of the consistent weighting model (i.e. w a,c = w a,b. w b,c ), phenomenon b is 2 times more important than phenomenon c. The priorities can be calculated by normalizing the absolute values of the principle eigenvector as shown in Table 5-5 below, the priority of phenomenon a is 1.5 times the priority of phenomenon b and 3 times the priority of phenomenon c. Table 5-5: Example priorities calculation a b c Eigenvector priorities a b 1/ c 1/3 1/ Now let us consider the two buses under evaluation to demonstrate the following cases 1) Only the most important phenomenon exceeds the limit, and 2) the moderate and least important phenomena exceed the limits and the most important phenomenon is well below the limit. Table 5-6 shows the example variables (note that units are ignored as all phenomena are normalized throughout the solution procedure). Table 5-6: Example buses performances Bus 1 Bus 2 Performance Threshold Performance Threshold Phenomenon a Phenomenon b Phenomenon c The first step to the solution is to calculate the score of the buses regarding each phenomenon. The score is the ratio between the performance and the threshold; it is a measure of how far the bus performance is from the threshold (note that only in the case of exceedance, is the score of the bus higher than the threshold score, as in Bus 1 171

172 phenomenon a, and Bus 2 phenomena b and c). The scores are calculated by the same algorithm used for the weighting matrix calculation, as shown in Table 5-7. Table 5-7: Example score calculation (a) Phenomenon a Bus 1 Threshold 1 Score Bus 2 Threshold 2 Score Bus 1 1 3/ Bus / Threshold 1 2.5/ Threshold 2 3/ (b) Phenomenon b Bus 1 Threshold 1 Score Bus 2 Threshold 2 Score Bus 1 1 2/3 0.4 Bus 2 1 2/ Threshold 1 3/ Threshold 2 1.5/ (c) Phenomenon c Bus 1 Threshold 1 Score Bus 2 Threshold 2 Score Bus / Bus /1 0.6 Threshold 1 2/ Threshold 2 1/ The second step is to weight the performance in each phenomenon by the corresponding priority (calculated in Table 5-5): Table 5-8: Example score weighting Bus 1 Threshold 1 Bus 2 Threshold 2 priority score s*p score s*p priority score s*p score s*p phenomenon a phenomenon a phenomenon b phenomenon b phenomenon c phenomenon c Sum Sum Note: s*p = score x priority The final step is to apply equation (5.3) (reproduced here), using the sum values calculated in Table 5-8. which gives, CBPQI i = N n=1 N n=1 score i,n priority n score Threshold,n priority n (5.3) CBPQI 1 = 0.48/0.52 = p.u. CBPQI 2 = 0.455/0.545 = p.u. Thus, by comparing the two values of the index, we can reach the conclusion that Bus 1 has a worse performance than Bus 2. Note that the index did not capture the violation of thresholds in either of the cases (i.e. recording higher than 1), neither the violation for the most important phenomenon, nor the violation for any of the two least important phenomena. 172

173 The CBPQI is primarily intended for ranking purposes and is based on the weighted averaging of different indices. In some cases therefore (considering that it assesses global PQ performance) it can mask violation of particular threshold by a certain phenomenon that was assigned low weight in the procedure or masked by the good performance of remaining phenomena. If limits violation is the priority, higher weights should be assigned to exceeding phenomena in the calculations. This numerical presentation is straightforward in terms of comparing PQ performance of different buses when accounting for different thresholds. It is also useful in terms of mitigation planning as it helps in identifying weak PQ performing areas of the network. Specifically, if the PQ performance is to be compared to predefined thresholds that vary spatially Case study and CBPQI results Separate PQ evaluation PQ Monte Carlo simulations as presented in Chapter 3 and Chapter 4 are applied and the results are utilised here to illustrate the application of CBPQI. An analysis of harmonics was carried out by varying the injections of sources through a study period, and the 95 th percentile of the voltage THD was recorded considering the results for all simulation hours over the study period. In this case study, to keep the methodology presentation as simple as possible, the sub-criteria level for the harmonics was not applied. Figure 5-9 shows a heat map of the THD performance of the buses under evaluation, with the calculated 95 th percentiles THD values, for all simulation hours, ranging from 0.21% to 7.13%; the threshold for the harmonic performance is set to THD=5 %. The load unbalance is created by varying the reactive power only in accordance with sampled power factors. The unbalance performance of the buses under study is shown by the heat map in Figure 5-10, where the 95 th percentile of VUF, for all simulation hours of the study period, ranged between 0.12% and 2.17%; the threshold for the unbalance performance is set to 173

174 VUF=2 %. The sag evaluation is based on the frequency and severity of all possible sag incidents on the GDN network using the BPI S as an index. The sag bus performance index BPI S ranged from 0.1 p.u. to 2.7 p.u. In the study, the threshold for sag performance is set to BPI S Thr=3 p.u., which is subject to modification and here it is chosen based on the worst bus performance when critical operating conditions are applied. Figure 5-11 shows the sag performance of the network. Figure 5-9: GDN harmonic performance Figure 5-10: GDN unbalance performance Figure 5-11: GDN voltage sag performance 174

175 Overall PQ evaluation The overall PQ performance of each bus is based on the calculation of CBPQI of the bus following the methodology illustrated above in The weights for three different phenomena considered are sampled from uniformly distributed ranges. These ranges should be overlapped to cater for the different sensitivities to different phenomena at different times for the bus. In other words, the importance rank of phenomena may vary at some samples in the simulation, Figure 5-12 shows an example of three overlapping uniformly distributed weighting ranges. Density 1/(w 2, max -w 2, min ) 1/(w 3, max -w 3, min ) 1/(w 1, max -w 1, min ) w3, min w2, min w1, min w3, max w2, max w1, max Weightings Figure 5-12: Uniformly distributed weighting ranges for PQ phenomena A Monte Carlo simulation technique is adopted to consider the uncertainty in the importance of the phenomena. The typical MC approach relies on a high number of repeated simulations with sampled uncertain input(s). The number of simulations usually depends on the required confidence level that the error of the output mean is below a given threshold [141]. To illustrate the methodology with reasonable computational burden, five hundred Monte Carlo simulations are performed to calculate the CBPQI by sampling different weights from the weight ranges and applying the AHP model. The input indices to the AHP model (Figure 5-8) are the 95 th percentiles values of THD, VUF and BPI S from annual performances calculated by separate probabilistic evaluations for each phenomenon. The PDF of the CBPQI is obtained by the Monte Carlo simulations and the 175

176 most probable value is taken as the final CBPQI. The calculated CBPQI for the buses under evaluation ranged from 0.07 p.u. to 0.76 p.u., see Figure Figure 5-13: Overall PQ performance based on CBPQI The weight ranges for different phenomena are randomly selected to illustrate the methodology and to have the sag as the most important (weights 15-20), the harmonics the second most important (weights 10-15) and the unbalance the least important (weights 6 10). In practical applications, the immunity levels of different loads connected to different buses to different phenomena should be considered to determine the relative thresholds of the phenomena for the bus. Furthermore, the expected losses resulting from PQ disturbances at different buses can be used to weight the different phenomena, e.g. the cost of sag trips compared to the cost of extra losses from harmonics or the cost of load derating from the unbalance. This approach, however, would require very detailed information about the customers connected to different buses and the immunity of their equipment to different phenomena. Therefore, the selection/assignment of the weights to different phenomena can be either qualitative or quantitative. The qualitative selection is guided by expert opinion or the operator s experience of relative importance of different phenomena in general or at specific locations. The quantitative selection is based on measured PQ losses (technical or economical) due to different phenomena at different locations or on the ratio of sensitive loads at different locations. The Monte Carlo 176

177 simulations for the selection of weights from different ranges are adopted in this case to cater for the uncertainty and temporal variation of the types of sensitive loads connected at the bus throughout the study period. If, on the other hand, equal weights and standard thresholds are used for different PQ phenomena considered (voltage sags, harmonics and unbalance in this example), the overall technical PQ performance of the network can be assessed, disregarding the financial consequences to the customers. This level of technical PQ performance can be used to compare different parts of the network or overall network performance against other networks in a similar way as suggested in [41, 42], i.e., by aggregating the CBPQI sample values for a number of buses (area) or for the whole feeder and taking the relevant statistical measure (e.g. average, 95 th percentile, maximum, etc.) of the sample to represent the whole area, feeder or network performance. The inclusion of customer information for these types of comparisons is not necessary and would only be required if mitigating solutions are to be developed. This is because the choice of solution will be strongly influenced by costs induced by PQ disturbances to different customers, which are, on the other hand, strongly related to equipment immunity and types of customer activities. Table 5-9 (a) shows the performance of the five worst performing buses based on each phenomenon separately and the corresponding CBPQI calculated for each bus (4 buses are among the worst performing in both harmonics and unbalance, therefore only 11 buses are shown). The normalized values in the table are normalized based on the thresholds adopted for each index and the CBPQI is normalized based on the performance of the worst performing bus in the network. Regarding the good performence, Table 5-9 (b) shows the best performing buses based on CBPQI. The best performing bus is Bus 87 scoring CBPQI=0.074 p.u. (0.098 normalized). Interestingly, this bus is not the best performing bus in any separate evaluation yet it showed the best overall performance. The 177

178 best performing bus in terms of sag only is Bus 59 scoring CBPQI=0.089 p.u. (0.118 normalized). The best bus in terms of harmonics only is Bus 226 scoring CBPQI=0.101 p.u. (0.133 normalized). The best bus in terms of unbalance is Bus 13 scoring CBPQI=0.185 p.u. (0.244 normalized). Table 5-9: Buses ranked based on CBPQI (a) Worst performing buses (ranked in descending order from the worst bus) Sags Harmonics Unbalance CBPQI Bus Bus Bus Bus Bus Bus Bus Bus Bus Bus Bus (b) Best performing buses (ranked in descending order from the best bus) Sags Harmonics Unbalance CBPQI Bus Bus Bus Bus Bus Bus Bus Bus Bus Bus The values from Table 5-9 (a) are depicted in Figure The effect of the weighting factors can be clearly seen by comparing the individual phenomenon heat maps (Figure 5-9, Figure 5-10 and Figure 5-11) with the overall performance shown in the heat map in Figure The worst performing areas in terms of sag are the worst areas after unifying the indices. Moreover, some of the worst performing areas in terms of unbalance are completely masked in the overall heat map showing average overall performance. This 178

179 Performance (p.u.) impact can also be noted in Figure 5-14, where the CBPQI bars (striped) follow the trend of the sag bars (black). The first five buses (193, 210, 196, 194 and 195) which show poor performances for both sag and harmonics and good performances for unbalance scored very high CBPQI, while the next five buses (136, 138, 137, 134 and 135), which show good performances in sag but poor performances in both unbalance and harmonics, still scored relatively high in the CBPQI. For Bus 36, the fifth worst bus in terms of unbalance but with good performances in both sag and harmonics, the CBPQI is low indicating good overall performance sag har unb CBPQI Bus 193 Bus 210 Bus 196 Bus 194 Bus 195 Bus 136 Bus 138 Bus 137 Bus 134 Bus 135 Bus 36 Figure 5-14: Worst performing buses (normalized) Another important aspect can be seen from the results in Table 5-9, the PQ performance varies slightly (second or third decimal place) between some buses; this is intuitive because the geographical and electrical proximities between some buses cause them to experience the same types and levels of PQ disturbances, unless a certain bus has special operating conditions (e.g. DG connected and/or capacitors bank connected). This fact can be used for a zonal or area based PQ evaluation rather than bus by bus evaluation, especially if the evaluation is performed to identify the weak areas of the network for the purpose of PQ mitigation. This is because the mitigation solutions will also affect zones or 179

180 areas rather than affecting the connection buses only. Considering this zonal behavior of different PQ phenomena and consequently overall PQ in the network as well as all the uncertainties involved in assessment, the practical approach would be to identify ranges of CBPQI and group the buses into classes, e.g very poor performance, acceptable performance and good performance, rather than insisting on individual ranking of buses. Heat maps used for illustration of the results in the figures above are a good example of identifying critical areas of the network rather than critical buses per se. 5.4 Application of PQ indices in real measurements Measurement synthesizing The PQ measurements of 8 sites are adopted for the application of the PQ indices. The sites are of approximately equal load (based on equal service transformer size) and of different types (4 domestic, 2 commercial and 2 mixed type). The measurements are taken in the period from 2/9/2013 to 23/11/2014 covering 64 weeks with 10 minute resolutions (i.e measurements per phenomenon). The recorded measurement types are the 3 phases RMS voltage, 3 phases THD, 3 phases Pst, VUF, and 3 phase 24 harmonic voltages; which in total are 82 records per site. The PQ measurements plots of all sites for the THD, VUF and Pst are shown in the Appendix D. Some readings are missing from the measurement database, or considered to be bad data. The missing data are replaced by the same week of the other year if available (the 12 week measurements between 2/9 and 23/11 are recorded for two years), or by measurements from the following week at the same time. Although the number of missing measurement vary between sites, all 8 sites are considered equally in the PQ evaluation and comparisons, as the missing data did not exceed 5% for any site for one phenomenon (Site 7 has 5.1% missing measurements of all records), bearing in mind that, according to 180

181 the standards, a 5% time threshold exceedance is usually allowed. Table 5-10 shows the summary of the measurements PQ performance analysis The considered phenomena for evaluation are harmonics (THD only), unbalance (VUF) and flicker (Pst), for phase A of the sites, the Compound Bus PQ Index (CBPQI) is calculated at each measurement point using these indices, having all indices (separate and combined) with equal sample size. The indices are handled statistically, the following measures are calculated for each site; the minimum, maximum, average, median, 95 th percentile and the mode (the most frequent value). The statistical measures adopted for comparison between sites are the 95 th percentiles and the most frequent values. To represent the data in PDFs, the fitting distribution adopted in all data analysis is the normal distribution for the sake of simplicity of comparison (although not the best fit for some data samples) Table 5-10: Measurements Summary Starting date and time "2/9/2013" 00:00 Measurements period minutes (64 weeks) Resolution 10 minutes Measurements types 82 (3xV RMS +3xTHD+VUF+3xP st +3x24xV h ) Voltage level LV (230 V L-N) Total number of readings Sites Site Type Missing readings Ratio of missing (%) Site 1 Domestic Site 2 Commercial Site 3 Domestic Site 4 Domestic Site 5 Commercial Site 6 Domestic Site 7 Mixed Site 8 Mixed The flicker phenomenon, in particular, showed extreme cases of performance for some sites. The Pst reached very high values for some sites, more than 100 p.u. in some 181

182 cases (compared to the standard threshold of 1 p.u. [4]). This could be due to faults, switching of capacitors or the failed start-up of big motors or simply bad data from the meters. However, further investigation was not carried out as it is beyond the scope of this study. Instead, in the interest of better statistical comparisons and of reducing the effects of outliers in the measures, the extreme values are chopped to Pst=2 p.u. i.e. 100% exceedance or double the standard threshold. This assumption basically means the impact of Pst=8 p.u. and Pst=100 pu, for example, is not worse than 100% exceedance of the threshold. As this assumption is applied to all sites, the main objective of the study, i.e. sites comparison, is fairly unaffected. Two methodologies are adopted to perform the analysis of sites. The first is the comparison based on the average performance of all phenomena by using the CBPQI. The second methodology is by comparing the available margins of normalized separate performances to 1 p.u. (i.e. PQ reserve); the most critical phenomenon with minimum margin represents the site performance index. The normalized performances are based on the adopted standards thresholds, i.e., THD=5% [14], VUF=2% [60] and Pst=1 p.u. [4] Reporting PQ based on EN50160 Based on EN50160 [60] the considered PQ phenomena in LV site evaluation are the RMS voltage, voltage unbalance (VUF), total harmonic distortion (THD), long term flicker (P lt ) and the individual harmonic voltages up to the 25 th harmonic (even harmonics are not considered here due to low readings during the whole measurement period). In the following analysis, all performances (except for unbalance) are considered per phase and the worst performing phase is taken as the site performance. In the case of the RMS voltage performance the phase with the highest absolute deviation RMS dev from ideal case (230 V) is taken as the representing phase. The PQ performance variation for Site 1 is shown in Figure 5-15 with the aid of box plots. The CBPQI was calculated for each 182

183 measurement week based on the 95 th percentile of the different phenomena performances with equal phenomena weights. In addition, to facilitate the comparisons, the phenomena were normalized based on their respective limits from EN50160 (the RMS dev is normalized based on 23 V limit, i.e. the +10% maximum allowed deviation from the 230 V nominal voltage suggested in EN50160) and shown in Figure 5-16 colour matrix in pu. The CBPQI is shown in the top row and it is already normalized in the calculation process (see equation 5.3). The colour matrix shown in Figure 5-16 consists of 60 columns Figure 5-15: Site 1 PQ phenomena variation Figure 5-16: Site 1 PQ performance (normalized based on thresholds) 183

184 corresponding to each measurement week, and of 17 rows corresponding to the 16 measured phenomena plus the calculated global index. The colour matrix is useful in identifying seasons patterns of PQ performance behaviour or identifying certain measurement week with noticeable change in performance. For example after week 46, noticeable changes were recorded in the harmonic performance. The performance of the 15 th voltage harmonic improved while an increase in the 3 rd voltage harmonic was recorded from week 46 onward. As a result, the THD has slightly increased from that week on. This change in the behaviour could be a result of the connection of new type of nonlinear load Flexibility of the indices Two ways of introducing flexibility to the evaluation are suggested in this section. Firstly, different levels of phenomena importance can be introduced to the calculation of the CBPQI in the form of weightings based on site type (see Table 5-3). Secondly, the importance of sites can be introduced by multiplying the calculated indices based on an importance factor of the site, which changes the final rank. The importance of sites is studied in two conflicting ways; the site is more important if it has more customers, or the site is more important if the p.u. size of customer connected (number of customers/service transformer size) is higher. By introducing the flexibility in the calculation of the index, the PQ evaluation can be more focused on particular PQ phenomena or particular sensitive customers. Consequently, the final ranking of buses can provide information about the areas in need of PQ improvement depending on different priorities set by network operator Weighting of phenomena Three levels of importance are applied to weight the phenomena. The assigned weights, Table 5-11, are selected with the aid of a normal distribution function, segmented based on proximity to the average, see Figure This can be explained in the following way, with a load type that is sensitive to certain phenomena, and these sensitivities (failure 184

185 probability) can be represented by normal distribution functions with coinciding averages; the phenomenon with high probability of failure that falls into the first segment (σ to σ) is considered as the most important, with the weighting factor equal to the probability within the first range, i.e Similarly, for the moderate important phenomenon, with a probability of failure of the second range, the weighting factor is equal to the probability, i.e Finally, the least important phenomenon with a probability of failure falls in the third segment, the weighting factor is Table 5-11: Normal distribution segments for weightings calculations Probability range Probability inside range Weighting factors High importance μ ± σ Moderate importance 2*(2σ σ) Low importance 2*(3σ 2σ) Figure 5-17: Segmented normal distribution function for weighting factors calculations (adopted from [142]) The importance of the three considered phenomena is correlated to the type of the site. The phenomena importance of the domestic sites is taken in this order, flicker is the most important (w flk =0.68), harmonics with moderate importance (w har =0.27) and unbalance with the least importance (w unb =0.04). For the commercial sites, the importance of the considered phenomena is taken in the following order: harmonics, unbalance and flicker, applying the same weights. For the case of mixed sites, the weights are averaged based on the ratio of each type from the total load. As shown in Table 5-12 and Figure 185

186 Weightings (pu) 5-18, when averaging the weights based on distribution of types, as in Site 7, the harmonics remains the most important phenomenon, while in Site 8 where the distribution of the domestic/commercial ratio is the flicker is the most important phenomenon. Table 5-12: Weighting factors of phenomena for each site Site 1 Site 2 Site 3 Site 4 Site 5 Site 6 Site 7 Site 8 Type Dom Com Dom Dom Com Dom Dom Com Dom Com Ratio (%) w flk w har w unb Flk Har Unb Site 1 (Dom) Site 2 (Com) Site 3 (Dom) Site 4 (Dom) Site 5 (Com) Site 6 (Dom) Site 7 (Mix) Site 8 (Mix) Figure 5-18: Weighting factors of phenomena for each site Weighting of sites The importance of sites is included based on two criteria: the first is the per-unit (pu) customer size. This is calculated based on the assumption that all customers are of equal size, and the number of customers divided by the service transformer size (630 kva) gives the p.u. customer size at a certain site. Then the final values of indices are multiplied (i.e. weighted) by the ratio of the p.u. size to the total served capacity (8 sites* 630 kva = 5040 kva) to consider the site importance in the final rank, Table 5-13 (a). The second criterion is to consider the importance of sites based on the number of customers connected. The final indices are multiplied (weighted) by the ratio of the customers 186

187 connected to a site to the total number of served customers, Table 5-13 (b). The two criteria give opposite direction of the importance, and each has its advantages and disadvantages. The first criterion (p.u. size of customers) is more suitable when comparing different types of sites, for example when having a service transformer dedicated to a single customer (as the case of Site 6). This usually means a more important site, in terms of PQ requirements, as it is usually small industries or big commercial compounds which have this way of connection; this criterion can be considered as a qualitative way of considering importance. The second criterion is suitable when comparing similar types of site, where all the customers have similar PQ requirements and the site with a higher number of customers will have higher importance, simply because a higher number of customers might be affected by the PQ performance; the second criterion can be considered a quantitative way of measuring importance. Table 5-13: Importance of sites considered (a) Based on p.u. size of customers (Criterion I) Site 1 Site 2 Site 3 Site 4 Site 5 Site 6 Site 7 Site 8 p.u. size of customers Importance of site (b) Based on number of customers (Criterion II) Site 1 Site 2 Site 3 Site 4 Site 5 Site 6 Site 7 Site 8 Number of customers Importance of site Results of the analysis Single site analysis Site 6 is selected to illustrate the PQ analysis performed for each site. The CBPQI is used in the performance analysis to study: - The impact of a year s growth of load and increased DG penetration. - The impact of different seasons on the PQ performance. 187

188 The impact of a year s growth is studied by comparing the performance of the same weeks (the 12 weeks between 2/9 and 23/11) between the two years. The considered phenomena and the CBPQI samples are plotted for the years 2013 and 2014 in Figure The study enables the DSO to measure the expected change in performance of the site and the effects (degradation or improvement) that the reinforcements or increased penetration may have on the overall performance. As shown in Figure 5-19, a clear improvement was recorded in both the unbalance and harmonics phenomena, albeit the harmonics recording a higher variation in performance. A slight degradation was also recorded in the flicker performance. This change in performance was accurately captured by the CBPQI, in terms of overall performance and variation of the PQ. Figure 5-19: 2013 vs 2014 PQ performances To study the impact of the seasons in the PQ performance, the study period is divided into 4 seasons: the periods between December to February, March to May, June to August and September to November. The measurements at each period for the considered phenomena and the corresponding calculated CBPQI were analysed separately and compared to each other, see Figure The figure shows almost identical flicker 188

189 performances throughout the year, while the unbalance performance shows worse performance during the winter period (December to February) due to higher load in general and possibly less evenly distributed. The average harmonic performances of all periods are almost the same, though the period from March to May shows a higher variation in the performance, and the maximum recoded THD value of all periods. The CBPQI indicates a similar seasonal performance in the year of study, with the winter period recording slightly higher CBPQI values (worse performance). Figure 5-20: Seasonal PQ performances Comparison of sites performance The main objective of the CBPQI is to perform comprehensive PQ comparisons between sites, to decide which ones require more attention from the DSO in terms of PQ performance. This section provides a comparison between the performances of indices in deciding the PQ rank of sites. The effect of considering flexibility in evaluating the sites is also presented. Two statistical measures are adopted for the comparisons. The first is the mode which represents the level of performance that took place most of the time during the study 189

190 period. The most frequent values give a realistic estimate of which level of PQ losses occurred. This is due to the fact that levels of PQ performance were recorded over long periods of time, especially for the continuous phenomena like unbalance and harmonics. The second measure adopted for the comparisons is the 95 th percentile. The percentiles give the maximum levels of the performance that were recorded over 95% of the time of the study period. The 95 th percentiles are usually adopted by the standards as thresholds of phenomena. The percentiles can capture the performance of the phenomena with high variation levels which are disregarded if the mode is adopted. However, the 95 th percentiles are highly affected by the outliers and extreme cases in the evaluation. Figure 5-21 and Figure 5-22 show the performances of the considered phenomena for Site 5 and Site 6. Part (a) of the figures show the histograms of the phenomena and CBPQI, while part (b) of the figures show the results fitted into normally distribution functions (even if it is not the best fit to facilitate visual comparisons). In the case of Site 5, note the high frequency of the Pst performance at 2 pu; this is because all the exceeding values beyond Pst=2 p.u. (100% exceedance) were assigned this value, to simplify the statistical fittings and comparison. This can also be clearly seen in Figure 5-21 (b) by the wide flicker phenomenon distribution function (high σ) compared to the flicker distribution function in Figure 5-22 (b). The available reserve in each of the indices is indicated by the margins to the unity of the normalized indices, as shown in Figure 5-21 (b) and Figure 5-22 (b). As mentioned above, the mode measure is not affected by the extreme cases as much as the 95 th percentile, and actually similar performances of these two sites can be noted using the mode as a measure. However, this is not the case when using the 95 th percentile as the measure (the mode for Site 5 is Pst= 0.35 p.u. and for Site 6 is Pst= 0.26 pu, where the 95 th percentile for Site 5 is Pst=1.2 p.u. and for Site 6 is Pst=0.34 pu). 190

191 (a) Histograms of considered phenomena and CBPQI (b) Normally distributed fits of the phenomena and CBPQI Figure 5-21: Site 5 performances (a) Histograms of considered phenomena and CBPQI 191

192 (b) Normally distributed fits of the phenomena and CBPQI Figure 5-22: Site 6 performances Table 5-14 presents the numerical results of the performance analysis, using the most frequent value as a measure. The normalized performances are normalized based on individual phenomena thresholds; therefore the margin to thresholds is the distance to unity for all phenomena (see Figure 5-21 (b) and Figure 5-22 (b)). The last two columns of Table 5-14 (a) show the unifying indices, CBPQI considers all phenomena in the calculation, and the minimum margin index represents the site based on its most critical phenomenon. Table 5-14 (b) and (c) show the sites performances ranks based on each type of overall indices, ranking from the best performing sites (Note that the best performance, based on CBPQI, corresponds to the site with the smallest value of the index, while the best performance based on the minimum margin corresponds to the site with the highest value of the index). Similarly, Table 5-15 (a), (b) and (c) represent the same analysis, but based on the 95 th percentiles as measures. By comparing Table 5-14 (b) and (c) with Table 5-15 (b) and (c), the effects of phenomena with a high rate of variation can be clearly noted. The rank based on the critical phenomenon has significantly changed, and the critical phenomena representing some sites have also changed. For example, Site 1 s minimum margin, when taking the mode as a measure, is and is recorded for the harmonics phenomenon, but when applying the 95 th percentile as a measure, Site 1 s most 192

193 critical phenomenon changed to the flicker with a margin of In general, the statistical measure and adopted index do not affect the rank of the extreme cases, i.e., if there is a significant difference in performance of some sites. For example, as shown in Table 5-14 and Table 5-15 Site 6 was always the best performing site regardless the selected index or statistical measure. Similar conclusion can be drawn for Site 5 which was the worst performing site in three ranks and second worst in the fourth rank. Table 5-14: PQ performance comparison based on the mode as a measure (a) Sites performances for all considered phenomena Normalized performance Margins to thresholds THD VUF P ST M_THD M_VUF M_Pst CBPQI Minimum Margin Site Site Site Site Site Site Site Site (b) Rank from the best site based on minimum margin (c) Rank from the best site based on CBPQI Minimum margin CBPQI Minimum margin CBPQI Site Site Site Site Site Site Site Site Site Site Site Site Site Site Site Site

194 Table 5-15: PQ performance comparison based on the 95 th percentile as a measure (a) Sites performances for all considered phenomena Normalized performance Margins to thresholds THD VUF P ST M_THD M_VUF M_Pst CBPQI Minimum Margin Site Site Site Site Site Site Site Site (b) Rank from the best site based on minimum margin (c) Rank from the best site based on CBPQI Minimum margin CBPQI Minimum margin CBPQI Site Site Site Site Site Site Site Site Site Site Site Site Site Site Site Site In the case where different levels of importance, of phenomena and sites, are to be considered (see Section 5.4.3), the CBPQI 95 th percentiles are adopted for the comparison of the sites. Six different cases are analysed: - Case 1: Phenomena not weighted, Sites not weighted, Table 5-16 (a), - Case 2: Phenomena weighted, Sites not weighted Table 5-16 (b), - Case 3: Phenomena not weighted, Sites weighted (Criterion I), Table 5-16 (c), - Case 4: Phenomena weighted, Sites weighted (Criterion I), Table 5-16 (d), - Case 5: Phenomena not weighted, Sites weighted (Criterion II), Table 5-16 (e), - Case 6: Phenomena weighted, Sites weighted (Criterion II), Table 5-16 (f). 194

195 Table 5-16: Ranks of site based on different cases of weighting phenomena and sites (best site on top) (a) Case 1 (b) Case 2 Phenomena NOT weighted sites NOT weighted Phenomena weighted sites NOT weighted Site Site Site Site Site Site Site Site Site Site Site Site Site Site Site Site (c) Case 3 (d) Case 4 Phenomena NOT weighted sites weighted (p.u. size of customer) Phenomena weighted sites weighted (p.u. size of customer) Site Site Site Site Site Site Site Site Site Site Site Site Site Site Site Site (e) Case 5 (f) Case 6 Phenomena NOT weighted sites weighted (number of customers) Phenomena weighted sites weighted (number of customers) Site Site Site Site Site Site Site Site Site Site Site Site Site Site Site Site It can be seen from the different cases presented in Table 5-16, considering the importance of sites, in both criteria, masks the effects of weighting the phenomena. In particular, it can be seen that identical ranks are recorded for cases 5 and 6, and that there 195

196 is a slight difference in the ranks of cases 3 and 4. The ranks of cases 1 and 2 are significantly different since different levels of phenomena importance are considered at different sites. To compare different levels of flexibility, two cases of weighting are compared with the original case. The original case is the case where all PQ phenomena (based on EN [60]) are considered with equal weights, i.e., to compare the performance of sites based on standards thresholds only. The cases where flexibility is considered with different levels are the case when three phenomena measures (VUF, THD and Plt) are weighted equally for all sites throughout the measurement weeks, and the case when the considered phenomena are weighted based on site type (see Table 5-12). Figure 5-23 and Figure 5-24 (a) and (b) show the weekly ranks and the average rank of sites in each case. The dark red represents the worst site and the dark blue represents the best site, while black boxes in the figure represent missing measurement weeks. By comparing Figure 5-23 and Figure 5-24 it can be seen that there is a high discrepancy between average ranks. Furthermore, the weekly ranks for the weighted cases are not as consistent as the original case. This is mainly because fewer phenomena are taken into account and their variations are therefore more pronounced in the averaging process. It is clear that adding different types of flexibility in PQ evaluation shifts the attention from some sites to another (except for extreme performances, e.g., Site 5 remains the worst performing in all cases). Nevertheless, adding different types of flexibility in PQ evaluation can be used to retune the index so that greater focus can be given to certain PQ disturbances or types of customers or other operational requirements (it is always possible, though, to adopt the original case index to benchmark performance with the standards.) Table 5-17 presents the comparison between the rankings of the three case studies. 196

197 Weekly Ranks Average Rank Figure 5-23: Average and weekly sites rank (original case) Weekly Ranks Average Rank (a) Constant weights of phenomena Weekly Ranks Average Rank (b) Variable weights based on site type Figure 5-24: Average and weekly sites ranks (flexible cases) 197

198 Table 5-17: Average site ranking (from worst site) for different case studies Original Case Constant weights Weighting based on site type Worst Site 5 Site 5 Site 5 Site 4 Site 3 Site 1 Site 1 Site 1 Site 4 Site 7 Site 2 Site 3 Site 3 Site 4 Site 8 Site 8 Site 8 Site 6 Site 2 Site 7 Site 2 Best Site 6 Site 6 Site Summary The abundance of data collected for different PQ phenomena, from different areas, represented by different indices over long study periods has still not answered the question of how to evaluate the PQ in the network in general, i.e., taking all relevant PQ phenomena into account simultaneously so that weak performing areas can be identified. This chapter attempts to point out the importance of having a standard way, between all stake holders DSOs/customers/regulators, to describe a bus s performance in terms of PQ as good, adequate or poor. It is especially important due to the increased application of new technologies in distribution networks, and due to the different levels of impact that these technologies have on PQ. PQ performance is commonly understood in terms of specific phenomena but as this chapter has demonstrated, it can be analyzed more comprehensively. The exposure of different buses to different network disturbances (many of which cause PQ disturbances) throughout the year raises the question of the performance of the whole network or part of it in terms of PQ adequacy and standards compliance. Identifying areas of the network (collection of adjacent buses) which perform better or worse than the other areas can directly influence the network s asset maintenance and the development of the network s maintenance strategy. Describing network performance using different indices for different phenomena can lead to suboptimal solutions both technical and economical. 198

199 This chapter presents two methodologies to evaluate the overall PQ at a site or a bus, using a single index, for the purpose of benchmarking and network reinforcement planning. The first index, PQR, is calculated based on the concept of PQ reserves, and it considers different levels of importance of different phenomena in the evaluation, based on utilities and customers perspectives. It tries to find the balance between the requirements of both parties in the overall calculated index. The index considers the expected variable PQ requirements between different customers in different areas of the network. These PQ requirements may also vary with time. The temporal variation in thresholds is not considered but can be easily incorporated in the proposed methodology and so further enhance its applicability to PQ evaluation in future power networks with spatial and temporal variation of generation and loads. The second index for the overall PQ evaluation is the CBPQI. It evaluates the PQ performance comprehensively based on both, event-type (voltage sags) and continuoustype phenomena (unbalance and harmonics) considering different weights for each phenomenon in the overall evaluation. The methodology can include sub-level evaluations for the phenomena considered, e.g. the harmonics can be evaluated based on THD and selected harmonic voltages or any other relevant index like the Crest Factor or zero crossing, depending on the sensitive equipment and the more relevant evaluation indices. The index can be used to identify the weak areas of the network in terms of overall PQ, to provide a useful comparison tool between the buses and to give indicative information about how far a PQ performance of a certain bus is from the PQ limits. The unified PQ indices are applied to real PQ measurements, comparing the performances of indices and considering different flexibility levels on weighting phenomena and sites for the overall PQ evaluation. The new indices developed for the global evaluation of PQ performance of network buses, PQR and CBPQI, are the fifth original contributions of the thesis. 199

200 6 The concept of Provision of differentiated PQ This chapter presents the concept of provision of differentiated quality of electricity supply based on customers requirements in distribution networks. As discussed in Chapter 2, the mitigation of PQ can be either network based or device based. The network based solutions can be broadly described as preventive solutions where the PQ performance is enhanced by reducing or preventing the causes of the PQ disturbances. On the other hand the device based PQ solutions can be described as corrective actions, i.e., by eliminating or reducing the impacts of PQ disturbances after they occur. Moreover, the device based solutions can have zonal effects, i.e., they can improve the PQ performance in the connection bus and some neighbouring buses. Therefore, the selection of mitigation devices location can be performed based on optimisation techniques, taking into account the whole network performance, to achieve the global optimal PQ performance. In this chapter, the provision of differentiated PQ is presented by applying device based PQ 200

201 mitigation solutions, optimally selected based on different PQ requirements at different zones in the network. The methodology is illustrated on the GDN network taking the CBPQI as the measure of the overall PQ performance of the network buses. 6.1 General concept The general concept of the PQ in different zones including PQ temporal and spatial threshold variations is illustrated in Figure 6-1. The DSO in this example studies three levels of quality in each of the three zones: high power quality (HPQ), moderate power quality (MPQ) and low power quality (LPQ). The levels may not necessarily be equal between zones (LPQ A =/ LPQ B =/ LPQ C ). For the study period (day, month, year etc.) and due to different uncertainties (DG penetration, loading levels, faults probabilities, ) in different zones, the PQ can vary between different levels, for example Zone A shows higher variation in PQ, where zones B and C show less variation (the red solid circle shows the current PQ level, the dotted circles shows the levels that were recorded through the study period). Additionally, different load types can be considered (three in this example). For the same study period and due to uncertainties in load sensitivity, variable thresholds Thr Variation HPQ A Threshold 2 MPQ A Zone A PQ Variation Thr Variation Threshold 2 HPQ B MPQ B Threshold 1 Threshold 1 LPQ A LPQ B Zone B PQ Variation Load Type 1 (LT1) Load Type 2 (LT2) Zone C Thr Variation Load Type 3 (LT3) HPQ C MPQ C Threshold 2 LPQ C Threshold 1 PQ Variation Figure 6-1: The concept of differentiated PQ provision to different types of loads 201

202 will be recorded. From the figure, several observations can be noted to determine the possibility and feasibility of applying differentiated PQ and premium contracts: - In Zone A, as a result of the overlap between the PQ variation and the threshold variation, the DSO must plan mitigation solutions (depending on the customer s losses and their willingness to pay) to ensure that the PQ level stays at HPQ value with variation not exceeding the Threshold 2 of LT1. - In Zone B, regardless of the fact that the PQ level is the highest, it is still probable that LT2 will encounter PQ problems. Individual plant level solutions will be recommended for this type of customer. - In Zone C, as there is no overlap between the PQ and the threshold variations, no PQ problems will arise for LT3 in this zone. - If the levels of PQ are equal between zones, LT3 will not face PQ problems in either zones B or C, while in Zone A the DSO needs to take actions to provide LT3 with the required PQ. LT2 will face PQ problems in all the zones unless plant level actions are taken to increase the immunity of the plant (lowering Threshold 2 below HPQ x ) before agreeing on PQ provision by the DSO. The above example is simplified for demonstration purposes; in fact, the problem of providing different levels of PQ can be much more complicated. For instance, it is possible to have customers with high requirements dispersed all over the network, and also it is possible to have zones with certain levels of PQ enclosing islands of totally different levels of PQ (higher or lower levels). These problems, the conflicts in the PQ requirements between neighbouring customers and the demarcation of different performance zones, can highly impact the feasibility of the premium contract application and hinder the application of the differentiation of PQ. 202

203 The main concern of the application of differentiated PQ contracts is the availability of suitable market models. In the Council of European Energy Regulators (CEER) report [10] the following is stated: In a liberalised electricity market, the customer concludes either a single contract with the supplier (SP) or separate contracts with the supplier and the distribution system operator (DSO), according to the existing national regulations. Therefore, basically the customers and DSOs are allowed to adopt contractual agreements of certain levels PQ, commonly referred to as Premium Power Quality contracts. A report [57] gives different examples of these types of contracts in France, Italy, USA and South Africa (see Chapter 1). In some European countries where the national regulators prohibit a direct contact between the DSO and customers, the role of aggregators, ancillary service providers, or virtual power plants (VPP) owners can be studied to perform the provision of premium PQ. Figure 6-2 shows the possible arrangements of PQ contracts in today s deregulated electricity markets. Retailor PQ Contract PQ Contract Sags/ Interruptions Sags/ Interruptions PQ Contract TSO Cap Switching Transients Harmonics Regulation, Unbalance, Flicker DSO PQ Contract Cap Switching Transients Harmonics Regulation, Unbalance, Flicker PQ Productivity Performances Power Factor, Harmonic Control Power Factor, Harmonic Control, Unbalance, Fluctuations Customer Agg/VPP Figure 6-2: PQ agreements in deregulated electricity market environment (modified from [58]) 203

204 6.2 Application of CBPQI for provision of differentiated PQ To illustrate the application of the CBPQI (Chapter 5) in differentiated PQ provision, the GDN is divided into three zones with different PQ requirements, as shown in Figure 6-3. The locations of the unbalanced loads, fixed non-linear loads and different types of distributed generators are marked by different labels in the figure. The network is divided into three zones circled by solid red lines. The zone division and zonal PQ requirements are set here for illustrative purposes only. They are based on the distribution of different classes of customers and the assumed sensitivities of different classes of customers to PQ disturbances. The industrial loads are mainly located in zone 2, thus zone 2 is assigned the most rigorous PQ requirement in the study, with the CBPQI threshold (CBPQI TH,2 ) set to CBPQI TH in zones 1 and 3 are set to and respectively, to represent the differentiated levels of PQ requirements. Based on the evaluated PQ performance using CBPQI as a measure and objective function, an optimization technique is developed to locate optimally different types of FACTS and filter devices at selected buses of the GDN network. G 400kV kV kV N 277 O kV A B C D E H I J K L kV kV Zone Zone Zone-3 F G Figure 6-3: GDN different PQ zones 134 The optimization problem is defined to minimize the gap between the received PQ performance and the zonal thresholds for violating buses. To facilitate the concept of 204

205 provision of differentiated PQ levels, the gap indices are proposed to present the distance between the PQ performance of the buses from their zonal thresholds. The thresholds with respect to voltage sags, harmonics and unbalance phenomena in PQ zone i are denoted as BPI TH,i, THD TH,i and VUF TH,i respectively. If the PQ phenomena are considered individually, three gap indices can be derived. Sag Gap Index (SGI), which presents the gap between the received voltage sag performance and the imposed zonal sag requirements, can be defined as: N B j SGI = i=1 ( BPI i,j BPI TH,i BPIi,j ) (6.1) j=1 >BPI TH,i where B j denotes the total number of buses within PQ zone i; and BPI i,j denotes BPI of the j th bus in zone i. The same principle is applied to the phenomena of harmonics and unbalance respectively, Harmonic Gap Index (HGI) and Unbalance Gap Index (UGI) can be derived as below: N B j HGI = i=1 ( THD i,j THD TH,i THDi,j ) (6.2) j=1 >THD TH,i N B j UGI = i=1 ( VUF i,j VUF TH,i VUFi,j ) (6.3) j=1 >VUF TH,i Similarly, given the aggregated PQ performance and zonal PQ thresholds CBPQI TH,i, the gap between the received PQ and the zonal PQ thresholds can be defined as: N B i PQGI CBPQI = i=1 ( j=1 CBPQI i,j CBPQI TH,i CBPQIi,j ) (6.4) >CBPQI TH,i In the study, passive filters (PF) and FACTS devices including Static VAR Compensator (SVC), Static Compensator (STATCOM) and Dynamic Voltage Restorer (DVR) are investigated for PQ mitigation. SVC is a shunt device that regulates the voltage by controlling the reactive power generated into or absorbed from the power system. STATCOM regulates the voltage by adjusting the amount of reactive and active power transmitted between the power system and the Voltage Source Converter (VSC). DVR 205

206 connected in series with the grid is capable of protecting sensitive loads against the voltage variations or disturbances via a VSC that injects a dynamically controlled voltage in series with the supply voltage through three single-phase transformers for correcting the load voltage. With these pre-selected locations, greedy algorithm [143] is used to search the optimal placement of FACTS devices and their optimal rating settings. Greedy algorithm is chosen due to its simplicity of implementation. It divides the problem into different consecutive stages and solves the problem heuristically by making the local optimal choice greedily at each stage. The proposed methodology includes three parts, global selection, zonal selection and optimization, as illustrated in Figure 6-4. In order to place available devices optimally, potentially effective locations for their placement are selected based on the analysis of PQ performance and sensitivity analysis. The potential locations are chosen globally (i.e., based on the whole network) and zonally (i.e., based on zonal information) respectively. These locations form a pool of available locations for optimization algorithm to select the optimal device placement. 1) Global selection. Buses are sorted according to performance indices BPI, VUF, N B THD, V j j=1 and V j j=1 in descending order, respectively (step 1 in Figure 6-4). Q N B P The ranking index of bus B i with respect to BPI is denoted as R BPI (B i ), and the same applies to other variables. Then R BPI (B i )=1 suggests that bus B i is experiencing the worst sag performance, and R V/ Q (B i ) = 1 that the bus voltage in the network is the most sensitive to the injection of reactive power at bus B i. The buses having R BPI =1, R VUF =1, the smallest R BPI + R V/ Q, the smallest R VUF + R V/ Q and the smallest R BPI +R VUF are selected as the potential locations for installing SVC (step 2 in Figure 6-4). The same 206

207 Begin Set potential device set U G =Ø, in which each element is a pair of the device type and its associated location 1) Rank all buses according to BPI, VUF, THD, V j / Q And V j / P in descending order; obtain R BPI, R VUF R THD, R V/ Q and R V/ P for each bus B i Global selection 2) U G =U G {SVC and location of bus having R BPI =1} {SVC and bus with R VUF =1} {SVC and bus with the smallest R VUF +R V/ Q } {SVC and bus with R BPI +R VUF } 3) U G =U G {STATCOM and location of bus having R BPI =1} {STATCOM and bus with R VUF =1} {STATCOM and bus with R BPI +(R V/ Q +R V/ P )/2} {STATCOM and bus with the smallest R VUF +(R V/ Q +R V/ P )/2} 4) Repeat step 3 with STATCOM replaced with DVR 5) U G =U G {PF and location of bus having R THD =1} 6) U G =U G {PF and location of intersections of branches} 7) Place SVCs of U G in the network, then perform steps 1 and 2; Place STATCOMs of U G in the network, then perform steps 1 and 3; Place DVRs of U G in the network, then perform steps 1 and 4: Place PF then perform steps 1 and 5 Zonal selection 8) For each zone Z i, i=1,,n, set potential zonal device set U zi =Ø, perform steps 1-5 and 7 with bus ranking performed within zone Z i only U T =U G U z1... U zn X=U; Γ=Φ; Install covered devices Γ; Update X by reselect rating randomly within its associated interval Optimisation Select sϵx that minimizes objective function F; X=X-{all elements in X which have the same location and type of device as s}; Γ=Γ {s}; Reach stop criteria? No Yes End Figure 6-4: Flowchart of the optimization methodology procedure is applied to select the potential locations for installing STATCOM (step 3 in Figure 6-4) and DVR (step 4 in Figure 6-4), while in this case the selection is based on R BPI =1, R VUF =1, the smallest R BPI + R V/ P +R V/ Q, the smallest R 2 VUF + R V/ P +R V/ Q and 2 207

208 the smallest R BPI +R VUF. It can be seen that instead of using R V/ Q, the R V/ P +R V/ Q 2 is used in this case as both STATCOM and DVR can transmit both active and reactive power between the devices and the grid. To initialize the placement of PF, the same selection procedure mentioned above is performed to select the potential locations, while the buses are ranked based on R THD (steps 1 and 5 in Figure 6-4). For each type of devices, following the selected devices of the same type are preliminarily placed at the selected potential locations, the selection procedure introduced above is then performed again to select the second set of potential locations (step 7 in Figure 6-4). Besides, the intersections of two branches which have more than three buses in the downstream branches are also initially made available for placement of PF (step 6 in Figure 6-4), as the PF located at the intersections can prevent the harmonic current flowing from one branch to another. 2) Zonal selection. To ensure the capability of providing certain PQ levels required in different zones, the potential locations should also be selected zonally (step 8 in Figure 6-4). For zonal selection, the procedure is the same as the global selection, except that the ranking procedure is performed within the zones rather than within the whole network. Geography feasibility could be also taken into account during the process of selecting potential locations. A pool of potential solutions, denoted as set U, should be determined based on the initial placement/locations and rating constraints. With this initial set U, the greedy algorithm is applied to select the optimal mitigation solution. The flowchart of the application of greedy algorithm for this allocation problem is also shown in Figure 6-4, where s is the chosen solution corresponding to the minimum objective value evaluated at each stage; Γ denotes the devices selected so far; and X is the updated pool of potential solutions at each stage. At each stage, X is updated by removing its elements which have 208

209 the same location and type of devise as the selected s. The optimization procedure can be terminated if the size of Γ reaches the preset maximum number of devices allowed to be installed, or if the improvement of PQ performance between two sequential stages is smaller than a preset threshold. Set Γ is selected as the final optimal mitigation solution. 6.3 Optimal provision of differentiated PQ The impact of the adopted objective functions on the selection of the mitigation scheme and ultimately their impact on the final mitigated PQ performance are investigated through different scenarios. Four scenarios are introduced here: Case 1: Optimization based on SGI. Case 2: Optimization based on HGI. Case 3: Optimization based on UGI. Case 4: Optimization based on PQGI CBPQI. In cases 1-3, each PQ phenomenon is tackled individually. For these three cases, the optimization procedure terminates if the evaluated gap index reaches zeros. In case 1, SGI reaches zero with the installation of 4 devices; in case 2, HGI reaches zero with 7 devices; and in case 3, 4 devices are required to reduce the UGI to zero. If all of these devices mentioned above are enabled simultaneously during the simulation, the load flow calculation cannot converge. It suggests that in PQ mitigation planning, it is more appropriate to consider the related critical phenomena simultaneously. Otherwise, the solution which directly combines the optimal schemes obtained from individual PQ phenomenon respectively could be infeasible. In case 4, the performance of the three PQ phenomena is aggregated and the PQGI CBPQI is used as the objective function. The optimization procedure terminates when the improvement of PQGI CBPQI index is <0.2%. CBPQI, BPI, THD and VUF evaluated at all buses in various cases are provided in Figure 6-5 (a)-(d) respectively. The results obtained in cases 1-3 are also provided in Figure

210 (b)-(d) to illustrate the difference between the performance of BPI, THD and VUF obtained based on the aggregated PQ performance and that obtained based on individual PQ phenomenon. (a) Overall PQ (b) Voltage sag (c) Harmonics (d) Unbalance Figure 6-5: CBPQI, BPI, THD and VUF performances before and after mitigation To compare between the different solutions, the optimal solutions obtained by optimization-based selection rules for different cases are listed in Table 6-1. The optimal solution obtained for case 1 consists of three DVR and one SVC, which shows the preference of DVR for sag mitigation. It can be seen from case 2 that harmonic mitigation solution favors STATCOM. Apart from the functionality of harmonic mitigation, PF, working along with other active devices, can also be used to compensate reactive power 210

211 and ultimately to regulate voltages. For case 4, the optimal mitigation solution consists of different types of devices, including STATCOM, SVC, DVR and PF, due to that all three PQ phenomena are considered simultaneously in this case. Table 6-1: Optimal solutions for different cases Cases type (size MVA) location Case 1 DVR(4.40) at B72; DVR(7.71) at B210; DVR(6.01) at B291; SVC (6.42) at B165 Case 2 STATCOM (1.34) at B29; STATCOM (4.38) at B42; STATCOM (7.27) at B124; STATCOM (7.18) at B210; SVC (6.66) at B196; PF (6.50) at B116; PF (4.52) at B232 Case 3 SVC (6.98) at B29; STATCOM (4.20) at B28; PF (7.77) at B102; PF (5.59) at B136 Case 4 STATCOM (7.85) at B48; STATCOM (6.88) at B138; SVC (5.47) at B36; SVC (6.55) at B72; PF (3.62) at B181; DVR (6.01) at B291 Moreover, hour-by-hour simulations are presented to validate the comprehensiveness of the selected solutions. The 24 hour simulations for a selected day are performed and the CBPQI is calculated for every bus under evaluation at every hour of the day. The results are shown in Figure 6-6, from which it can be seen that some buses violate the requirements of the zone at certain hours, even after mitigation. However, the common measures for long periods of studies are the percentiles, in particular the 95 th percentile; therefore these levels of performance can be accepted based on the commonly adopted standards procedure. A comparison of the heat maps in Figure 6-7 (a) and (b) shows an overall improvement of around 50% in the PQ performance of the network at the peak load hour of the study year when the optimum mitigation solutions are applied. The heatmaps of CBPQI obtained with 7 devices (3 SVCs, 3 DVRs and 1 passive filter) are plotted in Figure 6-7 (b) with optimal location of devices shown in the figure. The critical area marked in red was exposed to severe PQ disruption, as shown in Figure 6-7 (a), and it is greatly improved with the mitigation scheme obtained using the proposed mitigation methodology. It can be clearly seen that the optimal locations of devices, selected by the methodology, are at the areas with high concentration of disturbing loads or/and DGs. 211

212 H2 O2 H2 O2 Figure 6-6: Hourly PQ performance over a day before and after application of mitigation solution A B C D E H I J K L Photovoltaic Fuel Cell Wind Turbine Unbalance load Non-linear load CBPQI = 0 pu (a) Before applying mitigation solution CBPQI = 1 pu A B C D E H I J K L CBPQI = 0 pu Photovoltaic Fuel Cell Wind Turbine STATCOM Filter Unbalance load Non-linear load (b) After applying optimum mitigation solutions Figure 6-7: Peak hour PQ overall performance before and after application of mitigation solutions (CBPQI normalized based on worst bus) DVR SVC CBPQI = 1 pu 212

213 6.4 Summary The concept of provision of differentiated PQ was discussed in this chapter. The concept is introduced first and then an optimisation technique was applied to deliver optimal differentiated PQ in the network. The application of the CBPQI for developing optimal PQ mitigation solutions was illustrated on the GDN network. The considered PQ mitigation solutions include passive filters and different types of FACTS devices. The results showed that the proposed methodology yields promising mitigation scheme which ensures the received PQ performance meets the zonal thresholds. The main challenges of implementing the methodology are the requirement for substantial data and sufficient observability in the network regarding PQ requirements and performances. The development of the concept of differentiated PQ and methodology for the delivering of differentiated PQ through optimal placement of mitigation devices is the sixth original contribution of the thesis. 213

214 7 Conclusions and Future Work The main aim of this research was to develop a methodology for the optimal provision of PQ to customers with different PQ requirements in the MV of distribution networks. Three phenomena are considered as the main measures of PQ performance of buses and feeders, namely, voltage sag, harmonics and voltage unbalance. The concept of differentiated PQ is discussed at three levels: - Developing models and case studies to simulate considered PQ phenomena based on expected levels in typical contemporary distribution networks. - Evaluation of the PQ based on the developed models and the expected impact of increased DG and EV penetration in future networks. This evaluation is aimed at identifying the poorly performing areas of the network and benchmarking based on both international standards (regulators requirements) and the proposed thresholds from a PQ requirements classification step (customers requirements). - Suggestion of PQ enhancement solutions based on the simulated levels of PQ 214

215 performance and the different PQ requirements assigned to different areas in the test distribution network. 7.1 Major conclusions The thesis discussed the three considered phenomena in sufficient detail for MV distribution networks. The voltage sag, harmonics and voltage unbalance were discussed in Chapter 2, presenting the main causes of the phenomena, the negative impacts and the applicable mitigation solutions. The phenomena evaluation indices and the limits in the relevant international standards are also presented in the chapter. Chapter 3 then presented the test networks, simulation models and case studies for the probabilistic evaluation of the considered PQ phenomena. Characteristic studies are presented to select the appropriate models for the PQ simulation. All phenomena were simulated probabilistically considering a number of typical uncertainties that usually arise in distribution network operations. Chapter 3 also presented newly developed probabilistic sag index and a site sag index. The new sag event index, Sag Severity Index (SSI), is able to translate physical sag parameters (sag magnitude and duration) into voltage sag severity levels. The variation/uncertainty of equipment sensitivity to voltage sags is taken into account by adopting a set of normally distributed voltage tolerance curves, within the predefined range of variation/uncertainty, instead of a single curve. For the sag site index, Bus Performance Index (BPI S ) is derived based on simulating various fault types and locations in the network. The duration and frequency of sags are derived based on protection relay reliability and fault rates of different components in the network. Then the BPI S is calculated to quantify the sag performance by combining the occurrence frequency and the SSI. The voltage sag performance is presented using heat maps to facilitate the visualization and comparison between different areas of the network. Development of the new probabilistic sag indices and graphical presentation of sag, and the other considered PQ phenomena, are the first 215

216 and the second original contributions of the thesis. Chapter 4 discusses the harmonic phenomenon in further detail. Monte Carlo simulation was adopted to evaluate the phenomenon probabilistically. The harmonic performance was assessed using the THD as the main measure. Different periods of study and different penetration levels of DG were considered in the simulation. The main results show that the increased levels of DG, in particular the single phase units, can significantly impact the harmonic performance of the network. The long term probabilistic estimation of harmonics is the third original contribution of the thesis. To overcome the problem of limited harmonic monitoring, a simple and efficient technique for evaluating the radial feeder performance based on the substation readings only is proposed. It is demonstrated that the accuracy of the estimation can be further improved by using additional monitors at the large connected capacitors and by considering the resonance in the mathematical model. The newly developed method for estimation of harmonics in radial feeders based on limited monitoring is the fourth original contribution of the thesis. Chapter 5 presents two new global PQ indices developed to compress the large amount of PQ data into a single index with different aggregation levels. The raw data of each PQ phenomenon is statistically processed first and the resulting weekly 95 th percentiles of the PQ individual indices are combined into the global indices, namely the compound bus PQ index CBPQI and the PQ reserve index PQR. Different techniques of introducing flexibility to the evaluation based on the CBPQI are investigated. PQ phenomena are weighted based on their individual importance or expected losses, or based on the type of the load connected at different buses. Moreover, the sites themselves can be weighted in the overall evaluation based on their global PQ importance. The importance of sites is determined based on the number of customers connected or the p.u. size of the 216

217 connected customers. The application of global PQ evaluation is also illustrated using real measurements at eight LV sites. The development of the CBPQI and PQR is the fifth original contribution of the thesis. Chapter 6 presents the concept of provision of differentiated PQ based on customers requirements. Gap indices are proposed to present the differences between the received PQ compared to adopted thresholds. The thresholds are set based on either the individual PQ phenomenon or based on the aggregated PQ performance. Given the set of FACTS devices, a greedy algorithm is adopted to search the optimal mitigation scheme in order to minimize the gap between the actual received PQ performance and the imposed PQ thresholds and such provide optimal PQ mitigation solution. The proposed methodology particularly suits the distribution networks which have the characteristic of zonal centralization of customers of the same type. The simulation results demonstrate that the proposed methodology yields promising mitigation scheme which ensures the received PQ performance meets the imposed thresholds. Although when validating the optimum solutions based on individual evaluation of PQ phenomena some thresholds exceedances remain for certain buses and certain phenomena, the simplicity of the methodology still makes it attractive for the initial PQ mitigation analysis. For the areas in the network where zonal mitigation cannot provide the required levels of PQ, local PQ solutions must be investigated. The optimum locating of PQ solutions considering zonal requirements in the network is the sixth original contribution of the thesis. 7.2 Recommendations Distribution network operators are fully aware that in the future they will be operating in an increasingly competitive market environment and with increasing levels of uncertainty. They are already increasing the level of observability of PQ in their networks through more extensive monitoring campaigns and paying more attention to customers 217

218 requirements and satisfaction. In addition, new players in the energy market are emerging, the entities like aggregators, ancillary service providers or VPP owners can participate, contribute to and benefit from the PQ enhancement schemes. In order to apply the concept of differentiated PQ proposed in this thesis efficiently and to ensure the consensus among all parties involved, the following recommendations are made: - Keep historical reports and data about different PQ phenomena, to provide sufficient statistical information about past network performance and facilitate evaluation and prediction/forecasting of PQ levels in the future. - Collect information about the customers having sensitive industrial/commercial processes susceptible to interruptions or mal-operation due to PQ phenomena, their equipment and processes sensitive to different phenomena, sensitivity thresholds if possible and the resulting financial losses due to inadequate PQ. - Develop further and implement techniques for compressing the huge amount of PQ monitoring data. Data mining and classification methodologies should be particularly considered for processing the large amount of both historic and online monitoring data. - Develop further effective visualization techniques for faster and easier analysis of data. - Consider the required levels of PQ for different classes of customers in the future network planning, e.g., the suitable types of solutions, the costs of reinforcements, locations of high quality zones, etc. - Continuously analyse the consequences of increased penetration of DG and temporally and spatially varying loads (e.g., EV) in the network with the aim to establish potential effects on PQ performance of the network as a whole, or its parts. - Participate in the development of regulatory requirements and constraints (e.g., possibility of DG curtailment, levels of customer compensations, etc.) and expected market constraints (e.g., conflict between PQ requirements by different customers and 218

219 different valuation of PQ by different market players). This will facilitate the development of appropriate structure and contents of the PQ agreements and premium PQ contracts, and ensure the transparency and fairness in these agreements. 7.3 Future work The potential future steps in implementing the general concept of differentiated PQ are discussed first in this section. To start with, performing cost-benefit analysis to evaluate the profitability of the concept is an essential step. It can be challenging though to have a generic cost-benefit analysis model for such a concept with a high number of variables and uncertainties, yet it is very important to include the market model in the evaluation of the concept. The emerging players in the energy market, like ancillary service providers, might benefit from and contribute to such a concept and can be an interface between customers and utilities in implementing differentiated PQ concept. As previously mentioned, the current regulatory framework may hinder the application of the concept. Therefore, suggestions and involvement of all parties in the regulations set up and engineering guidance and codes preparations should be welcomed and considered. Some barriers to the scalability and replicability of the concept should also be identified. Customer awareness and willingness to participate could be a main barrier that needs to be overcome. Another main barrier for the wide application of the concept is the high cost of the PQ enhancement solutions based on FACTS devices. Further research in more cost effective solutions could provide cheaper alternatives and faster acceptance of global PQ enhancement schemes. Important direction for the future research is also the work on the standardization of PQ evaluation techniques, sensitivity of equipment and processes in particular to PQ disturbances and assessment of expected financial losses due to PQ disturbances. Discrepancies in these assessments will only lead to disagreements between the parties involved and hinder the adoption of the concept. 219

220 The increased deployment of smart meters in customers premises and the increased connection of LV monitors are leading to increasing amounts of data which require further handling and restructuring. These data can prove very useful in improving the estimation and analysis of PQ performance if properly structured, studied and shared between different parties in the power system field. Therefore a special consideration should be paid to application of data analytics techniques to general area of PQ. Finally, the five main technical areas considered in this thesis are presented in the following sub-sections, with presentations for the current assumptions or limitations and suggestions for future research directions in each of the areas Assessment of PQ requirements Current For modelling different requirements at different zones in the test networks, PQ thresholds for each zone are selected randomly from predefined ranges. Thresholds are time independent, i.e., only spatial variation in thresholds is considered. Future Considering temporal variation in PQ thresholds may further minimize the cost of the optimum PQ solutions. For the areas containing different types of loads, probabilistic representation of PQ requirements can be adopted to address the differences in the sensitivity levels and the importance of PQ phenomena for each load type Network modelling for PQ studies Current In all the case studies presented in the thesis, loads are modelled as constant power for all load types. Load buses are either a combination of domestic and commercial loads or industrial loads. All loads are connected to 11/33/132 kv. The differences between load types are only in the loading curves and PQ requirements rather than the model itself. 220

221 Regarding the DG models, all DG units are modelled as real power injections only with no controllers modelled. The same DG types have same annual output curves for all units all over the network, i.e., DG units location has no impact in the power output. For modelling of harmonic sources, the model adopted is current source with predefined injection ratio of fundamental currents at different harmonic frequency. The buses containing harmonic sources are considered to be 100% non-linear loads. The angles of injected harmonics are assumed to be uniformly distributed between degrees. Future For more accurate network and harmonic source modelling for PQ studies, customer behaviour can be included in the modelling, e.g., EV charging times of the day and DG disconnection periods throughout the year. Moreover, sensitive customers with different PQ requirements can be connected at the LV level (0.4 kv connected), therefore the network models between the 11 kv bus and the PCC must be included (e.g. service transformer, PF correction capacitors, PQ mitigation devices). Load models should also be selected based on dominant load equipment (motors, resistive, power electronics interfaced, etc) to consider different impacts of disturbances on different types of loads. Finally, the amplification or cancellation of harmonics must be considered based on better modelling of harmonic injection angles Uncertainties Current The main uncertainties considered in this research are: i) Harmonic current injection, ii) Unbalance levels of loads, iii) Loads sensitivity to sag events (probabilistic CBEMA curves) and iv) Importance of different phenomena at different buses (probabilistic weighting of phenomena) Future 221

222 Based on the network under study, the followings might be needed to be modelled as uncertainties: i) Bus loads, ii) Network parameters, iii) Network topology and iv) DG outputs based on location PQ estimation/evaluation Current In this thesis, the individual PQ phenomenon evaluation is based on single index per phenomenon, e.g., THD for harmonic evaluation and VUF for unbalance evaluation. For the overall evaluation, the CBPQI has values in open range, i.e., PQ thresholds exceedance can be masked in the overall evaluation. Also in the global PQ evaluation the phenomena weights are selected arbitrarily to illustrate the proposed concept. In the proposed estimation of harmonics methodology, full knowledge about the network components states is assumed (e.g. capacitors, switches, etc), which cannot always be the case in distribution networks. Future For the further work in this area, the sub-criteria level in the CBPQI calculation can be introduced to provide more detailed evaluation, for example the Crest Factor and Zero Crossing factor can be considered for the further detailed harmonic evaluation. On the other hand, the CBPQI can be modified to clearly show if there is exceedance in any phenomenon, even if the phenomenon is lightly weighted. Consistent, objective methodologies of weighting PQ phenomena in the overall evaluation are also needed to improve the methodology. Disturbance Costs evaluation is also important to assess the long term impact of disturbances, especially for the continuous type phenomena (e.g., is it worse to have 4.9% THD the whole week or to have 8% for a few hours and less than 1% for the rest of the week). 222

223 7.3.5 Optimizing PQ mitigation Current The optimization is performed assuming the full knowledge of the PQ performance, i.e., sources of harmonics, sources of unbalance, expected number of faults, etc. The optimization problem adopted is a single objective optimization. Each of the technical performance and the cost of the solution can be either an objective function or a constraint, i.e., multi-objective optimization is not considered. The adopted optimization is a constrained optimization problem; constraints like number of devices, size ranges of devices and candidate location buses are considered to reduce the computational burden and time. Future For the future work in this area, the optimization must be based on estimated PQ performance to reflect the fact that there is limited knowledge about actual PQ performance of the network. Investigating how does the accuracy of the estimation and measurement errors impact the selected solutions is important. Moreover, the optimization based on technical performance or PQ costs can give different set of solutions. Cost-benefit analysis of the suggested different solutions can establish which set of potential solutions is the most beneficial for all the parties involved. 223

224 References [1] Math H.J. Bollen, Understanding Power Quality Problems. New Jersey: John Wiley & Sons, 2000 [2] Electromagnetic compatibility (EMC) Part 4-30: Testing and measurement techniques Power quality measurement methods. IEC Standard , 2003 [3] IEEE Recommended Practice for Powering and Grounding Electronic Equipment. IEEE Standard , 2006 [4] Environment - Compatibility levels for low-frequency conducted disturbances and signalling in public low-voltage power supply systems. IEC :2002, 2002 [5] Glossary of Electrotechnical, power, telecommunication, electronics, lighting and colour terms Part 1: Terms common to power, telecommunications and electronics. IEC 50(161), 1990 [6] Jhan Yhee Chan, "Framework for Assessment of Economic Feasibility of Voltage Sag Mitigation Solutions," PhD, The University of Manchester, Manchester, 2010 [7] Jose Thomas, "Development of Methodology for Online Provision of Differentiated Quality of Electricity Supply," MSc, The University of Manchester, Manchester, 2012 [8] Yan Zhang, "Techno-economic Assessment of Voltage Sag Performance and Mitigation," PhD, University of Manchester, Manchester, 2008 [9] D. Lineweber and S. McNulty, "The Cost of Power Disturbances to Industrial & Digital Economy Companies," EPRI s Consortium for Electric Infrastructure for a Digital Society (CEIDS) [10] Council of European Energy Regulators, "5th CEER Benchmarking Report on The Quality of Electricity Supply," CEER [11] Roger C. Dugan, Mark F. McGranaghan, Surya Santoso, and H. Wayne Beaty, Electrical Power Systems Quality, 2nd ed. New York: McGraw-Hill, 2002 [12] Vanya Ignatova. Power Quality: Measuring to Manage [Online]. Available: [13] Task Force on HarrnonicS Modeling and Simulation, "Modeling and simulation of the propagation of harmonics in electric power networks. I. Concepts, models, and simulation techniques," IEEE Trans. Power Delivery, vol. 11, no. 1, pp , [14] IEEE Recommended Practices and Requirements for Harmonic Control in Electrical Power Systems. IEEE Standard 519:2014, 2014 [15] Yahia Baghzouz and Owen T. Tan, "Probabilistic Modeling of Power System Harmonics," IEEE Trans. Industry Applications, vol. IA-23, no. 1, pp , [16] Au Mau Teng and J. V. Milanovic, "Establishing Harmonic Distortion Level of Distribution Network Based on Stochastic Aggregate Harmonic Load Models," IEEE Trans. Power Delivery, vol. 22, no. 2, pp ,

225 [17] J. M. Crucq and A. Robert, "Statistical approach for harmonics measurements and calculations," in Proc. Electricity Distribution, CIRED th International Conference on, 1989, pp vol.2. [18] Assessment of Emission Limits for Distorting Loads in MV and HV Power Systems. IEC Standard :1996, 1996 [19] Lawrence L. Lapin, Modern Engineering Statistics. Belmont: Wadsworth Publishing Company, 1997 [20] PennState Eberly College of Science. STAT 414 Intro Probability Theory [Online]. Available: [21] D. A. Robinson, V. J. Gosbell, B. S. P. Perera, and D. J. Mannix, "Establishment of typical harmonic voltage levels in radial distribution systems," in Proc. Harmonics and Quality of Power, Proceedings. Ninth International Conference on, 2000, vol. 3, pp vol.3. [22] European Regulators' Group for Electricity and Gas, "ERGEG Public Consultation Towards Voltage Quality Regulation in Europe: Evaluation of the Comments Received," Bruxelles, E07-EQS-15-04, [23] Y. Baghzouz, R. F. Burch, A. Capasso, A. Cavallini, A. E. Emanuel, M. Halpin, A. Imece, A. Ludbrook, G. Montanari, K. J. Olejniczak, P. Ribeiro, S. Rios-Marcuello, L. Tang, R. Thaliam, and P. Verde, "Time-varying harmonics. I. Characterizing measured data," IEEE Trans. Power Delivery, vol. 13, no. 3, pp , [24] ENA Engineering Recommendation, "ER G5/4-1: Planning levels for harmonic voltage distortion and the connection of non-linear equipment to transmission systems and distribution networks in the United Kingdom," [25] J. E. Farach, W. M. Grady, and A. Arapostathis, "An optimal procedure for placing sensors and estimating the locations of harmonic sources in power systems," IEEE Trans. Power Delivery, vol. 8, no. 3, pp , [26] G. T. Heydt, "Identification of harmonic sources by a state estimation technique," IEEE Trans. Power Delivery, vol. 4, no. 1, pp , [27] A. P. S. Meliopoulos, Zhang Fan, and S. Zelingher, "Power system harmonic state estimation," IEEE Trans. Power Delivery, vol. 9, no. 3, pp , [28] C. Rakpenthai, S. Uatrongjit, N. R. Watson, and S. Premrudeepreechacharn, "On harmonic state estimation of power system with uncertain network parameters," IEEE Trans. Power Systems, vol. 28, no. 4, pp , [29] Ali Abur and Antonio Gomez Exposito, Power System State Estimation. New York: Marcel Dekker, 2004 [30] A. Arefi, M. R. Haghifam, and S. H. Fathi, "Distribution harmonic state estimation based on a modified PSO considering parameters uncertainty," in Proc. IEEE PowerTech conf.,trondheim, 2011, pp [31] G. D'Antona, C. Muscas, and S. Sulis, "State estimation for the localization of harmonic sources in electric distribution systems," in Proc. IEEE Instrumentation and Measurement Technology Conf., IMTC,, 2008, pp

226 [32] N. Okada and K. Yukihira, "Harmonic state estimation in distribution network based on fifth harmonic current characteristic," in Proc. IEEE 16th Int. Conf. on Harmonics and Quality of Power (ICHQP), 2014, pp [33] ELECTROTEK CONCEPTS, "Reliability benchmarking methodology," EPRI [34] T. Hiyama, M. S. A. A. Hammam, and T. H. Ortmeyer, "Distribution system modeling with distributed harmonic sources," IEEE Trans. Power Delivery, vol. 4, no. 2, pp , [35] M. F. McGranaghan, R. C. Dugan, Jack A. King, and W. T. Jewell, "Distribution Feeder Harmonic Study Methodology," IEEE Trans. Power Apparatus and Systems, vol. PAS-103, no. 12, pp , [36] A. E. Emanuel, J. Janczak, D. J. Pileggi, E. M. Gulachenski, C. E. Root, M. Breen, and T. J. Gentile, "Voltage distortion in distribution feeders with nonlinear loads," IEEE Trans. Power Delivery, vol. 9, no. 1, pp , [37] A. Bhowmik, A. Maitra, S. M. Halpin, and J. E. Schatz, "Determination of allowable penetration levels of distributed generation resources based on harmonic limit considerations," IEEE Trans. Power Delivery, vol. 18, no. 2, pp , [38] N. R. Watson, C. K. Ying, and C. P. Arnold, "A global power quality index for aperiodic waveforms," in Proc. Harmonics and Quality of Power, Proceedings. Ninth International Conference on, 2000, vol. 3, pp vol.3. [39] J. Kilter, J. Meyer, B. Howe, F. Zavoda, L. Tenti, J. V. Milanovic, M. Bollen, P. F. Ribeiro, P. Doyle, and J. M. Romero Gordon, "Current practice and future challenges for power quality monitoring - CIGRE WG C4.112 perspective," in Proc. Harmonics and Quality of Power (ICHQP), 2012 IEEE 15th International Conference on, 2012, pp [40] S. Nourollah and M. Moallem, "A data mining method for obtaining global power quality index," in Proc. Electric Power and Energy Conversion Systems (EPECS), nd International Conference on, 2011, pp [41] V.J. Gosbbell. B.S.P. Perera and H.M.S.C. Herath, "New Framework for Utility Power Quality (PQ) Data Analysis," in Proc. AUPEC'01 Perth, Australia, September, 2001, pp [42] V. J. Gosbell, B. S. P. Perera, and H. M. S. C. Herath, "Unified power quality index (UPQI) for continuous disturbances," in Proc. Harmonics and Quality of Power, th International Conference on, 2002, vol. 1, pp vol.1. [43] H. M. S. C. Herath, V. J. Gosbell, and S. Perera, "Power quality (PQ) survey reporting: discrete disturbance limits," IEEE Trans. Power Delivery, vol. 20, no. 2, pp , [44] G. Carpinelli, P. Caramia, P. Varilone, P. Verde, R. Chiumeo, I. Mastrandrea, F. Tarsia, and O. Ornago, "A global index for discrete voltage disturbances," in Proc. Electrical Power Quality and Utilisation, EPQU th International Conference on, 2007, pp [45] Jan Meyer, Peter Schegner, Gert Winkler, Michael Muhlwitz, Drewag Stadtwerke, and Lutz Schulze, "Efficient method for power quality surveying in distribution networks," in Proc. Electricity Distribution, CIRED th International Conference and Exhibition on, 2005, pp [46] Jan Meyer, Hansjorg Holenstein, and Stefan Egger, "NEQUAL web based voltage quality monitoring in Switzerland," in Proc. 23rd Int. Con. and Exhib. on Electricity Distribution CIRED, 2015, pp

227 [47] Markus Kraft, "Power quality recording and evaluation in an industerial area (chemical park)," in Proc. 23rd Int. Con. and Exhib. on Electricity Distribution CIRED 2015, pp [48] S. A. Farghal, M. S. Kandil, and A. Elmitwally, "Quantifying electric power quality via fuzzy modelling and analytic hierarchy processing," IEE Proceedings- Generation, Transmission and Distribution, vol. 149, no. 1, pp , [49] W. Morsi and M. El-Hawary, "Fuzzy-wavelet-based electric power quality assessment of distribution systems under stationary and nonstationary disturbances," in Proc. Power and Energy Society General Meeting, 2010 IEEE, 2010, pp [50] A. Salarvand, B. Mirzaeian, and M. Moallem, "Obtaining a quantitative index for power quality evaluation in competitive electricity market," IET Generation, Transmission & Distribution, vol. 4, no. 7, pp , [51] G. A. Vokas, S. D. Kaminaris, P. A. Kontaxis, M. Rangoussi, G. C. Ioannidis, S. A. Papathanassiou, P. V. Malatestas, and F. V. Topalis, "Electric network power quality assessment using fuzzy expert system methodology," in Proc. Power Generation, Transmission, Distribution and Energy Conversion (MEDPOWER 2012), 8th Mediterranean Conference on, 2012, pp [52] Lee Buhm and Kim Kyoung Min, "Unified power quality index based on value-based methodology," in Proc. Power & Energy Society General Meeting, PES '09. IEEE, 2009, pp [53] Lee Buhm and Kim Kyoung Min, "Development of ideal analytic hierarchy process - application of power quality," in Proc. Fuzzy Systems, FUZZ-IEEE IEEE International Conference on, 2009, pp [54] Lee Buhm, Kim Kyoung Min, and Goh Yeongjin, "Unified power quality index using ideal AHP," in Proc. Harmonics and Quality of Power, ICHQP th International Conference on, 2008, pp [55] A. Bracale, P. Caramia, G. Carpinelli, A. Russo, and P. Verde, "Site and System Indices for Power-Quality Characterization of Distribution Networks With Distributed Generation," IEEE Trans. Power Delivery, vol. 26, no. 3, pp , [56] JWG Cigre/Cired C4.07, "Power Quality Indices and Objectives," Cigre Final WG Report, [57] E.L.M. Smeets W.T.J. Hulshorst, J.A. Wolse, "Premium Power Quality contracts and labeling," KEMA [58] M. McGranaghan, B. W. Kennedy, and M. Samotyj, "Power quality contracts in a competitive electric utility industry," in Proc. Harmonics and Quality of Power Proceedings, Proceedings. 8th International Conference On, 1998, vol. 1, pp vol.1. [59] J. Arrillaga, M. H. J. Bollen, and N. R. Watson, "Power quality following deregulation," Proceedings of the IEEE, vol. 88, no. 2, pp , [60] Voltage Characteristic of Electricity supplied by public distribution systems. CENELEC EN 50160:1999, 1999 [61] R. M. Gantiv and J. V. Milanovic, "Qualitative and quantitative analysis of voltage sags in networks with significant penetration of embedded generation," Euro. Trans. Electr. Power, no. 15, pp ,

228 [62] I. N. Santos, V. Cuk, P. M. Almeida, M. H. J. Bollen, and P. F. Ribeiro, "Considerations on hosting capacity for harmonic distortions on transmission and distribution systems," EPSR, no. 119, pp , [63] Wang Fei, J. L. Duarte, and M. A. M. Hendrix, "Analysis of harmonic interactions between DG inverters and polluted grids," in Proc. Energy Conference and Exhibition (EnergyCon), 2010 IEEE International, 2010, pp [64] D. Patel, R. K. Varma, R. Seethapathy, and M. Dang, "Impact of wind turbine generators on network resonance and harmonic distortion," in Proc. Electrical and Computer Engineering (CCECE), rd Canadian Conference on, 2010, pp [65] M. T. Arif, A. M. T. Oo, A. S. Ali, and G. Shafiullah, "Impacts of storage and solar photovoltaic on the distribution network," in Proc. Universities Power Engineering Conference (AUPEC), nd Australasian, 2012, pp [66] Cigre/Cired JWG C4.112, "Guidelines for Power Quality Monitoring - Measurement Locations, Processing and Presentation of Data," Cigre TB 596, [67] IEC, "IEC :2002 Voltage dips and short interruptions on public electric power supply systems with statistical measurement results," [68] Testing and measurement techniques Voltage dips, short interruptions and voltage variations immunity tests. IEC Standard : [69] IEEE Guide for Voltage Sag Indices. IEEE Standard , 2014 [70] IEEE Recommended Practice for Evaluating Electric Power System Compatibility With Electronic Process Equipment. IEEE Standard , 1998 [71] Assessment of emission limits for the connection of unbalanced installations to MV, HV and EHV power systems. IEC Standard , 2008 [72] Nick C Woolley, "Identification of Weak Areas and Worst Served Customers for Power Quality Issues Using Limited Monitoring and Non-Deterministic Data Processing Techniques," PhD, University of Manchester, Manchester, 2012 [73] IEEE Guide for Application of Power Electronics for Power Quality Improvement on Distribution Systems Rated 1 kv Through 38 kv. IEEE Standard , 2012 [74] EU FP 7 project 'Smart Distribution System Operation for Maximizing the Integration of Renewable Generation' SuSTAINABLE project no Deliverable 2.1, "Survey and reporting of state of the art technologies and services used by DSOs in Europe," [75] J. Arrillaga, N. R. Watson, and S. Chen, Power System Quality Assessment, 2nd ed. Chichester: John Wiley & Sons Ltd, 2000 [76] Jose Manuel Avendano Mora, "Monitor Placement for Estimation of Voltage Sags in Power Systems," PhD, University of Manchester, Manchester, 2012 [77] Zhixuan Liu, "Probabilistic Assessment of Unbalance in Distribution Networks Based on Limited Monitoring," PhD, University of Manchester, Manchester, 2014 [78] Z. Liu and J. V. Milanovic, "Probabilistic Estimation of Voltage Unbalance in MV Distribution Networks With Unbalanced Load," IEEE Trans. Power Delivery, vol. 30, no. 2, pp , [79] Jovica V. Milanovic, "Power Quality Teaching Notes," University of Manchester

229 [80] George J. Wakileh, Power Systems Harmonics: Fundamentals, Analysis and Filter Design. New York: Springer, 2001 [81] IEEE Recommended Practices and Requirements for Harmonic Control in Electrical Power Systems. IEEE 519:1992, 1992 [82] G. Lemieux, "Power system harmonic resonance: a documented case," in Proc. Pulp and Paper Industry Technical Conference, 1988., Conference Record of 1988 Annual, 1988, pp [83] L. Paulsson, B. Ekehov, S. Halen, T. Larsson, L. Palmqvist, A. A. Edris, D. Kidd, A. J. F. Keri, and B. Mehraban, "High-frequency impacts in a converter-based back-to-back tie; the Eagle Pass installation," IEEE Trans. Power Delivery, vol. 18, no. 4, pp , [84] A. A. Girgis, J. W. Nims, J. Jacomino, J. G. Dalton, and A. Bishop, "Effect of voltage harmonics on the operation of solid-state relays in industrial applications," IEEE Trans. Industry Applications, vol. 28, no. 5, pp , [85] IEEE Recommended Practice for Establishing Liquid-Filled and Dry-Type Power and Distribution Transformer Capability When Supplying Nonsinusoidal Load Currents. IEEE Standard C , 2008 [86] V. E. Wagner, J. C. Balda, D. C. Griffith, A. McEachern, T. M. Barnes, D. P. Hartmann, D. J. Phileggi, A. E. Emannuel, W. F. Horton, W. E. Reid, R. J. Ferraro, and W. T. Jewell, "Effects of harmonics on equipment," IEEE Trans. Power Delivery, vol. 8, no. 2, pp , [87] Mau Teng Au, "Stochastic Modelling and Mitigation of Harmonics in Distribution Networks," PhD, University of Manchester, Manchester, 2005 [88] M. T. Au and J. V. Milanovic, "Planning Approaches for the Strategic Placement of Passive Harmonic Filters in Radial Distribution Networks," IEEE Trans. Power Delivery, vol. 22, no. 1, pp , [89] Spec. Semicon. Process. Equipment Voltage Sag Immunity, SEMI-F [Online]. Available: www. Semi.org. [90] M. H. J. Bollen and D. D. Sabin, "International Coordination for Voltage Sag Indices," in Proc. Transmission and Distribution Conference and Exhibition, 2005/2006 IEEE PES, 2006, pp [91] A. Robert and E. De Jaeger, "Final Rep. round table on power quality at the interface T&D CIRED 2003," May. 14, [92] R. S. Thallam, "Power quality indices based on voltage sag energy values," in Proc. Proc. Power Quality Conf. Expo., Chicago, IL, [93] R. S. Thallam and G. T. Heydt, "Power acceptability and voltage sag indices in the three phase sense," in Proc. Proc. IEEE Power Eng. Soc. Summer Meeting, Seattle, WA, 2000, vol. 2, pp [94] J. Wang, S. Chen, and T. T. Lie, "System voltage sag performance estimation," IEEE Trans. Power Delivery, vol. 20, no. 2, pp , [95] Chan Nan Lu and Cheng Chieh Shen, "Estimation of Sensitive Equipment Disruptions Due to Voltage Sags," IEEE Trans. Power Delivery, vol. 22, no. 2, pp ,

230 [96] C. N. Lu and C. C. Shen, "Voltage sag immunity factor considering severity and duration," in Proc. Proc. IEEE Power Eng. Soc. Gneral Meeting, 2004, vol. 1, pp [97] Chan Nan Lu and Cheng Chieh Shen, "A Voltage Sag Index Considering Compatibility Between Equipment and Supply," IEEE Trans. Power Delivery, vol. 22, no. 2, pp , [98] J. Y. Chan, J. V. Milanovic, and A. Delahunty, "Generic Failure-Risk Assessment of Industrial Processes due to Voltage Sags," IEEE Trans Power Delivery, vol. 24, no. 4, pp , [99] Mark McGranaghan EC&M. "Dealing with voltage sags in your facility" [Online]. Available: [100] CIGRE TF C4.102, "Voltage dip evaluation and prediction tools," [101] J. Y. Chan, J. V. Milanovic, and A. Delahunty, "Risk-Based Assessment of Financial Losses Due to Voltage Sag," IEEE Trans. Power Delivery, vol. 26, no. 2, pp , [102] David Chapman, "The cost of poor power quality," Power quality application guide - Copper Development Association [103] Yan Ruifeng and T. K. Saha, "Investigation of Voltage Imbalance Due to Distribution Network Unbalanced Line Configurations and Load Levels," IEEE Trans. Power Systems, vol. 28, no. 2, pp , [104] Yan Ruifeng and T. K. Saha, "Investigation of Voltage Stability for Residential Customers Due to High Photovoltaic Penetrations," IEEE Trans. Power Systems, vol. 27, no. 2, pp , [105] J. Faiz, H. Ebrahimpour, and P. Pillay, "Influence of unbalanced voltage on the steady-state performance of a three-phase squirrel-cage induction motor," IEEE Trans. Energy Conversion, vol. 19, no. 4, pp , [106] Lee Ching-Yin, "Effects of unbalanced voltage on the operation performance of a threephase induction motor," IEEE Trans. Energy Conversion, vol. 14, no. 2, pp , [107] U. Jayatunga, S. Perera, and P. Ciufo, "Voltage unbalance emission assessment in radial power systems," in Proc. Power and Energy Society General Meeting (PES), 2013 IEEE, 2013, pp [108] D. R. Smith, H. R. Braunstein, and J. D. Borst, "Voltage unbalance in 3- and 4-wire delta secondary systems," IEEE Trans. Power Delivery, vol. 3, no. 2, pp , [109] Description of the environment - Electromagnetic environment for low-frequency conducted disturbances and signalling in public power supply systems. IEC Standard :1990, 1990 [110] A. Robert and J. Marquet, "Assessing Voltage Quality with relation to Harmonics, Flicker and Unbalance," WG 36.05, Paper , CIGRE 92. [111] W. H. Kersting, "Causes and effects of unbalanced voltages serving an induction motor," in Proc. Rural Electric Power Conference, 2000, 2000, pp. B3/1-B3/8. [112] EPRI Power Electronics Applications Center, "Input performance of ASDs during supply voltage unbalance Power quality testing network," PQTN Brief no. 28,

231 [113] IEEE Recommended Practice for Electric Power Systems in Commercial Buildings. IEEE Standard , 1991 [114] Electric Power Systems and Equipment Voltage Ratings (60 Hertz). ANSI Standard C , 1995 [115] P. Pillay and M. Manyage, "Loss of Life in Induction Machines Operating With Unbalanced Supplies," IEEE Trans. Energy Conversion, vol. 21, no. 4, pp , [116] Wang Yaw-Juen, "Analysis of effects of three-phase voltage unbalance on induction motors with emphasis on the angle of the complex voltage unbalance factor," IEEE Trans. Energy Conversion, vol. 16, no. 3, pp , [117] "Energy Analysis of a Three-Phase Induction Motor under Unbalanced Voltage Using Simulation and Symmetrical Component," in Proc. the 9th International Energy Conference,, Tehran Iran. [118] K. Lee, T. M. Jahns, T. A. Lipo, and V. Blasko, "New Control Method Including State Observer of Voltage Unbalance for Grid Voltage-Source Converters," IEEE Trans. Industrial Electronics, vol. 57, no. 6, pp , [119] K. Lee, G. Venkataramanan, and T. M. Jahns, "Source current harmonic analysis of adjustable speed drives under input voltage unbalance and sag conditions," IEEE Trans. Power Delivery, vol. 21, no. 2, pp , [120] T. Van Craenenbroeck J. Driesen, "Voltage Disturbances," Copper Development Association May [121] E. Neto R. Salustiano, M. Martine, " The Unbalanced Load Cost on Transformer Losses at A Distribution System," presented at the th International Conference on Electricity Distribution (CIRED),, Stockholm. [122] A. von Jouanne and B. B. Banerjee, "Voltage unbalance: Power quality issues, related standards and mitigation techniques," EPRI Final Rep. May [123] K. Lee, G. Venkataramanan, and T. M. Jahns, "Modeling Effects of Voltage Unbalances in Industrial Distribution Systems With Adjustable-Speed Drives," IEEE Trans. Industry Applications, vol. 44, no. 5, pp , [124] Lee Woo-Cheol, Lee Taeck-Kie, and Hyun Dong-seok, "A three-phase parallel active power filter operating with PCC voltage compensation with consideration for an unbalanced load," IEEE Trans. Power Electronics, vol. 17, no. 5, pp , [125] Li Kuang, Liu Jinjun, Wang Zhaoan, and Wei Biao, "Strategies and Operating Point Optimization of STATCOM Control for Voltage Unbalance Mitigation in Three-Phase Three-Wire Systems," IEEE Trans. Power Delivery, vol. 22, no. 1, pp , [126] R. Gupta, A. Ghosh, and A. Joshi, "Performance Comparison of VSC-Based Shunt and Series Compensators Used for Load Voltage Control in Distribution Systems," IEEE Trans. Power Delivery, vol. 26, no. 1, pp , [127] Smart Distribution System Operation for Maximizing the Integration of Renewable Generation [Online]. Available: [128] S. Bahadoorsingh, J. V. Milanovic, Zhang Yan, C. P. Gupta, and J. Dragovic, "Minimization of Voltage Sag Costs by Optimal Reconfiguration of Distribution Network Using Genetic Algorithms," IEEE Trans. Power Delivery, vol. 22, no. 4, pp ,

232 [129] J. V. Milanovic and Jingwei Lu, "Application Of artificial immune system for detecting overloaded lines and voltage collapse prone buses in distribution network," in Proc. IEEE Bucharest PowerTech, 2009, 2009, pp [130] The University of Edinburgh, "Matching Renewable Electricity Generation With Demand " Scottish Executive, Edinburgh [131] National Renewable Energy Laboratory. PVWatts Viewer [Online]. Available: [132] Roger C. Dugan, "Reference Guide - The Open Distribution System Simulator (OpenDSS)," EPRI [133] ABB, "Power factor correction and harmonic filtering in electrical plants," ABB 1SDC007107G0201 Technical Application Papers No [134] Schneider Electric, "Guide for the Design and Production of LV Power Factor Correction Cubicles," [135] Aung Myo Thu and J. V. Milanovic, "Stochastic prediction of voltage sags by considering the probability of the failure of the protection system," IEEE Trans. Power Delivery, vol. 21, no. 1, pp , [136] SMA Solar Technology. EN Type certification test result sheet [Online]. Available: [137] A. E. Emanuel, J. A. Orr, D. Cyganski, and Edward M. Gulachenski, "A survey of harmonic voltages and currents at distribution substations," IEEE Trans. Power Delivery, vol. 6, no. 4, pp , [138] T. H. Ortmeyer and T. Hiyama, "Distribution system harmonic filter planning," IEEE Trans. Power Delivery, vol. 11, no. 4, pp , [139] Luis G. Vargas Thomas L. Saaty, Models, Methods, Concepts & Applications of the Analytic Hierarchy Process. New York: Springer Science and Business Media, 2012 [140] G. T. Heydt and W. T. Jewell, "Pitfalls of electric power quality indices," IEEE Trans. Power Delivery, vol. 13, no. 2, pp , [141] R. Preece and J. V. Milanovic, "Efficient Estimation of the Probability of Small-Disturbance Instability of Large Uncertain Power Systems," IEEE Trans. Power Systems, vol. 31, no. 2, pp , [142] David L. Chandler. Explained: Sigma [Online]. Available: [143] T. H. Cormen, C. E. Leiserson, R. L. Rivest, and C. Stein, Introduction to Algorithms, 3rd ed.: MIT press,

233 Appendix A. Test Systems The 295-bus Generic Distribution Network Table A-1, Table A-2 and Table A-3 show the lines, transformers and peak loads details of the GDN network. The network is connected to an infinite external grid at bus 300. Table A-1: GDN lines details Name from to Rated kv R (Ohm) X (Ohm) R0 (Ohm) X0 (Ohm) B (us) B0 (us) HV OL HV OL HV OL HV OL HV cable HV cable HV cable HV cable HV cable HV cable HV cable HV cable HV cable HV cable HV cable HV cable HV cable HV cable HV cable HV cable HV cable HV cable HV cable HV cable HV cable HV cable OL OL OL

234 OL OL OL OL OL OL OL OL OL OL OL OL OL OL OL OL OL OL OL OL OL OL OL OL OL OL OL OL OL OL OL OL OL OL OL OL OL OL OL OL OL OL OL OL OL

235 OL OL OL OL OL OL OL OL OL OL OL OL OL OL OL OL OL OL OL OL OL OL OL OL OL OL OL OL OL OL OL OL OL OL OL OL OL OL OL OL OL OL OL OL OL

236 OL OL OL OL OL OL OL OL OL OL OL OL OL OL OL OL OL OL OL OL OL OL OL OL OL OL OL OL OL OL OL OL OL OL OL OL OL OL OL OL OL OL OL OL OL

237 OL OL OL cable cable cable cable cable cable cable cable cable cable cable cable cable cable cable cable cable cable cable cable cable cable cable cable cable cable cable cable cable cable cable cable cable cable cable cable cable cable cable cable cable cable

238 cable cable cable cable cable cable cable cable cable cable cable cable cable cable cable cable cable cable cable cable cable cable cable cable cable cable cable cable cable cable cable cable cable cable cable cable cable cable cable cable cable cable cable cable cable

239 cable cable cable cable cable cable cable cable cable cable cable cable cable cable cable cable cable cable cable cable cable cable cable cable All lines rated current is 1 ka. Name HV bus LV bus HV (kv) Table A-2: GDN transformers details LV (kv) x1 (pu) r1 (pu) Conn. HV Tx Yy HV Tx Yy HV Tx Yy HV Tx Yy HV Tx Yy HV Tx Yy HV Tx Yy HV Tx Yy HV Tx Yy HV Tx Yy HV Tx Yy HV Tx Yy HV Tx Yy HV Tx Yy HV Tx Yy min tap max tap tap pos. Add V/tap (%) 239

240 HV Tx Yy Tx Yd Tx Yd Tx Yd Tx Yd Tx Yd Tx Yd Tx Yd Tx Yd Tx Yd Tx Yd Tx Yd Tx Yd Tx Yd Tx Yd Tx Yd Tx Yd Tx Dyn Tx Dyn Tx Dyn Tx Dyn Tx Dyn Tx Yyn Tx Yyn All transformers rated MVA is 99 MVA Table A-3: GDN load details Name Bus P (MW) Q (MVAr) C C C C C C C C C C C C C C C C

241 C C C C C C C C C C C C C C C C C C C C C C C C C C C C C C C C C C C C C C C C C C C C C

242 C C C C C C C C C C C C C C C C C C C C C C C C C C C C C C C C C C C C C C C C C C C C C

243 C C C C C C C C C C C C C C C C C C C C C C C C C C C C C C C C C C C C C C C C C D D D D

244 D D D D D D D D D D D D D D D D D D D D D D D D D D D D D D D D D D D D D D D D D D D D D

245 D D D D D D D D D D D D D D D D D D D D D D D D D D D D D D D D D D D D D D D D D D D D D

246 D D D D D D D D D D D D D D D D D D D D D D D D D D D D D D D D D D D D D D D D D D D D D

247 D D D D D D D D I I I The real test feeder Table A-4, Table A-5 and Table A-6 show the lines, transformers and peak loads details of the real test network. The network is connected to an infinite external grid at bus sourcesbus. Name Rtd. Volt Table A-4: Real test feeder lines details (a) Line types Rtd Current R' X' R0' X0' B' B0' kv ka Ohm/km Ohm/km Ohm/km Ohm/km us/km us/km 15_LEHIV_120_S _LXHIOV_070_S _LXHIOV_120_S _LXHIOV_240_S (b) Lines connection details Name Type Terminal i Terminal j Length (km) ' ' 15_LXHIOV_120_S B 10 B ' ' 15_LXHIOV_120_S B 17 B ' ' 15_LXHIOV_120_S B 19 B ' ' 15_LXHIOV_070_S B 23 B ' ' 15_LXHIOV_120_S B 20 B ' ' 15_LEHIV_120_S B 15 B ' ' 15_LXHIOV_120_S B 20 B ' ' 15_LXHIOV_120_S B 16 B Line 01 15_LXHIOV_240_S B 01 B Line 02 15_LXHIOV_240_S B 02 CX Line 03 15_LEHIV_120_S CX CX

248 Line 04 15_LXHIOV_120_S CX B Line 05 15_LXHIOV_120_S B 03 CX Line 06 15_LEHIV_120_S CX B Line 07 15_LXHIOV_120_S B 04 B Line 08 15_LXHIOV_120_S B 05 CX Line 09 15_LXHIOV_120_S CX B Line 10 15_LXHIOV_120_S B 06 CX Line 11 15_LXHIOV_120_S CX B Line 12 15_LXHIOV_120_S B 07 B Line 13 15_LXHIOV_120_S B 08 B Line 14 15_LXHIOV_120_S B 09 B Line 15 15_LXHIOV_120_S B 11 B Line 16 15_LXHIOV_120_S B 12 B Line 17 15_LXHIOV_120_S B 13 B Line 18 15_LXHIOV_120_S B 14 B Line 19 15_LXHIOV_120_S B 15 B Line 20 15_LXHIOV_120_S B 20 B Line 21 15_LXHIOV_120_S B 23 CX Line 22 15_LXHIOV_120_S CX B Line 23 15_LXHIOV_120_S B 25 CX Line 24 15_LXHIOV_120_S CX B Line 25 15_LXHIOV_120_S B 26 B Line 26 15_LXHIOV_120_S B 27 B Line 27 15_LXHIOV_120_S B 28 B Line 28 15_LXHIOV_120_S B 29 B Line 29 15_LXHIOV_120_S B 30 B Line 30 15_LXHIOV_120_S B 31 B Line 31 15_LXHIOV_120_S B 32 B Line 32 15_LXHIOV_120_S B 33 CX Line 33 15_LXHIOV_120_S CX B Line 34 15_LXHIOV_120_S B 34 CX Line 35 15_LXHIOV_120_S CX B Table A-5: Real test feeder transformer details Trans S/S HV bus LV bus Rtd Pow. (MVA) HV-rtd (kv) LV-rtd (kv) x1(pu) r1(pu) conn. sourcebus B YNd11 Neu Tap Min Tap Max Tap Add V./tap (%)

249 Table A-6: Real test feeder load details Name Bus P (MW) Q (MVAr) Load 1 B Load 2 B Load 3 B Load 4 B Load 5 B Load 6 B Load 7 B Load 8 B Load 9 B Load 10 B Load 11 B Load 12 B Load 13 B Load 14 B Load 15 B Load 16 B Load 17 B Load 18 B Load 19 B Load 20 B Load 21 B Load 22 B Load 23 B Load 24 B Load 25 B Load 26 B Load 27 B Load 28 B Load 29 B Load 30 B Load 31 B Load 32 B Load 33 B Load 34 B Load 35 B Load 36 B Load 37 B Load 38 B

250 Output (p.u.) Output (pu) Output (pu) Output (pu) Output (p.u.) Output (pu) Appendix B. Wind and PV output profiles January February Hour of the month Hour of the month March April Hour of the month Hour of the month May June Hour of the month Hour of the month 250

251 Output (pu) Output (pu) Output (pu) Output (pu) Output (pu) Output (pu) July August Hour of the month Hour of the month September October Hour of the month Hour of the month November December Hour of the month Hour of the month Figure B-1: PV output monthly profiles 251

252 Output (p.u.) Output (pu) Output (pu) Output (pu) Output (p.u.) Output (pu) January February Hour of the month Hour of the month March April Hour of the month Hour of the month May June Hour of the month Hour of the month 252

253 Output (pu) Output (pu) Output (pu) Output (pu) Output (pu) Output (pu) July August Hour of the month Hour of the month September October Hour of the month Hour of the month November December Hour of the month Hour of the month Figure B-2: Wind generation monthly profiles 253

254 Appendix C. Transformers zero sequence circuits Figure C-1: Zero sequence circuits for the different transformers connections (adopted from [77]) 254

255 Appendix D. PQ Measurements The PQ measurements, used in Chapter 5, are presented in this appendix. The total harmonic distortion (THD), short term flicker (Pst) and the voltage unbalance factor (VUF) of the eight LV sites are shown in Figure D-1 overleaf. The lines in each plots represents the weeks measurement, 64 in total. The measurements are 1008 per week (10-minute readings). 255

256

257 Figure D-1: PQ measurements for harmonics, unbalance and flicker for 8 LV sites 257

UNIT-4 POWER QUALITY MONITORING

UNIT-4 POWER QUALITY MONITORING UNIT-4 POWER QUALITY MONITORING Terms and Definitions Spectrum analyzer Swept heterodyne technique FFT (or) digital technique tracking generator harmonic analyzer An instrument used for the analysis and

More information

The development of the SA grid code on Power Quality emission. Dr. Gerhard Botha 2017/08/08

The development of the SA grid code on Power Quality emission. Dr. Gerhard Botha 2017/08/08 The development of the SA grid code on Power Quality emission Dr. Gerhard Botha 2017/08/08 Overview What is the Grid Code? What is Power Quality? Power Quality Management Principles Differences Challenges

More information

On the Evaluation of Power Quality Indices in Distribution Systems with Dispersed Generation

On the Evaluation of Power Quality Indices in Distribution Systems with Dispersed Generation European Association for the Development of Renewable Energies, Environment and Power Quality International Conference on Renewable Energies and Power Quality (ICREPQ 09) Valencia (Spain), 1th to 17th

More information

Investigation of data reporting techniques and analysis of continuous power quality data in the Vector distribution network

Investigation of data reporting techniques and analysis of continuous power quality data in the Vector distribution network University of Wollongong Research Online University of Wollongong Thesis Collection 1954-2016 University of Wollongong Thesis Collections 2006 Investigation of data reporting techniques and analysis of

More information

CHAPTER 4 PV-UPQC BASED HARMONICS REDUCTION IN POWER DISTRIBUTION SYSTEMS

CHAPTER 4 PV-UPQC BASED HARMONICS REDUCTION IN POWER DISTRIBUTION SYSTEMS 66 CHAPTER 4 PV-UPQC BASED HARMONICS REDUCTION IN POWER DISTRIBUTION SYSTEMS INTRODUCTION The use of electronic controllers in the electric power supply system has become very common. These electronic

More information

Roadmap For Power Quality Standards Development

Roadmap For Power Quality Standards Development Roadmap For Power Quality Standards Development IEEE Power Quality Standards Coordinating Committee Authors: David B. Vannoy, P.E., Chair Mark F. McGranghan, Vice Chair S. Mark Halpin, Vice Chair D. Daniel

More information

Voltage Sags Evaluating Methods, Power Quality and Voltage Sags Assessment regarding Voltage Dip Immunity of Equipment

Voltage Sags Evaluating Methods, Power Quality and Voltage Sags Assessment regarding Voltage Dip Immunity of Equipment s Evaluating Methods, Power Quality and s Assessment regarding Voltage Dip Immunity of Equipment ANTON BELÁŇ, MARTIN LIŠKA, BORIS CINTULA, ŽANETA ELESCHOVÁ Institute of Power and Applied Electrical Engineering

More information

p. 1 p. 6 p. 22 p. 46 p. 58

p. 1 p. 6 p. 22 p. 46 p. 58 Comparing power factor and displacement power factor corrections based on IEEE Std. 18-2002 Harmonic problems produced from the use of adjustable speed drives in industrial plants : case study Theory for

More information

Harmonic impact of photovoltaic inverter systems on low and medium voltage distribution systems

Harmonic impact of photovoltaic inverter systems on low and medium voltage distribution systems University of Wollongong Research Online University of Wollongong Thesis Collection 1954-2016 University of Wollongong Thesis Collections 2006 Harmonic impact of photovoltaic inverter systems on low and

More information

CHAPTER 1 INTRODUCTION

CHAPTER 1 INTRODUCTION 1 CHAPTER 1 INTRODUCTION 1.1 BACKGROUND The increased use of non-linear loads and the occurrence of fault on the power system have resulted in deterioration in the quality of power supplied to the customers.

More information

Harmonic distortion analysis on the MV and LV distribution networks: problems, influencing factors and possible solutions

Harmonic distortion analysis on the MV and LV distribution networks: problems, influencing factors and possible solutions Harmonic distortion analysis on the MV and LV distribution networks: problems, influencing factors and possible solutions Fernando Bastião and Humberto Jorge Department of Electrical Engineering and Computers

More information

Harmonic Distortion Levels Measured at The Enmax Substations

Harmonic Distortion Levels Measured at The Enmax Substations Harmonic Distortion Levels Measured at The Enmax Substations This report documents the findings on the harmonic voltage and current levels at ENMAX Power Corporation (EPC) substations. ENMAX is concerned

More information

Reliability and Power Quality Indices for Premium Power Contracts

Reliability and Power Quality Indices for Premium Power Contracts Mark McGranaghan Daniel Brooks Electrotek Concepts, Inc. Phone 423-470-9222, Fax 423-470-9223, email markm@electrotek.com 408 North Cedar Bluff Road, Suite 500 Knoxville, Tennessee 37923 Abstract Deregulation

More information

IEEE sion/1547revision_index.html

IEEE sion/1547revision_index.html IEEE 1547 IEEE 1547: Standard for Interconnection and Interoperability of Distributed Energy Resources with Associated Electric Power Systems Interfaces http://grouper.ieee.org/groups/scc21/1547_revi sion/1547revision_index.html

More information

NEQUAL -WEB-BASED VOLTAGE QUALITY MONITORING IN SWITZERLAND

NEQUAL -WEB-BASED VOLTAGE QUALITY MONITORING IN SWITZERLAND NEQUAL -WEB-BASED VOLTAGE QUALITY MONITORING IN SWITZERLAND Jan MEYER Hansjörg HOLENSTEIN Stefan EGGER Etienne GASCH, Max DOMAGK Association of St. Gallisch-Appenzellische Technische Universitaet Dresden

More information

Southern Company Power Quality Policy

Southern Company Power Quality Policy Southern Company Power Quality Policy Alabama Power Georgia Power Gulf Power Mississippi Power i Table of Contents: Southern Company Power Quality Policy SCOPE AND PURPOSE... 1 DEFINITIONS... 2 I. HARMONICS...

More information

CHAPTER 4 POWER QUALITY AND VAR COMPENSATION IN DISTRIBUTION SYSTEMS

CHAPTER 4 POWER QUALITY AND VAR COMPENSATION IN DISTRIBUTION SYSTEMS 84 CHAPTER 4 POWER QUALITY AND VAR COMPENSATION IN DISTRIBUTION SYSTEMS 4.1 INTRODUCTION Now a days, the growth of digital economy implies a widespread use of electronic equipment not only in the industrial

More information

Power Conditioning Equipment for Improvement of Power Quality in Distribution Systems M. Weinhold R. Zurowski T. Mangold L. Voss

Power Conditioning Equipment for Improvement of Power Quality in Distribution Systems M. Weinhold R. Zurowski T. Mangold L. Voss Power Conditioning Equipment for Improvement of Power Quality in Distribution Systems M. Weinhold R. Zurowski T. Mangold L. Voss Siemens AG, EV NP3 P.O. Box 3220 91050 Erlangen, Germany e-mail: Michael.Weinhold@erls04.siemens.de

More information

1. Introduction to Power Quality

1. Introduction to Power Quality 1.1. Define the term Quality A Standard IEEE1100 defines power quality (PQ) as the concept of powering and grounding sensitive electronic equipment in a manner suitable for the equipment. A simpler and

More information

ECE 528 Understanding Power Quality

ECE 528 Understanding Power Quality ECE 528 Understanding Power Quality http://www.ece.uidaho.edu/ee/power/ece528/ Paul Ortmann portmann@uidaho.edu 208-733-7972 (voice) Lecture 19 1 Today Flicker Power quality and reliability benchmarking

More information

Power quality report. A Manufacturing Plant

Power quality report. A Manufacturing Plant Power quality report Prepared for A Manufacturing Plant 6 May 2016 by Dr Angelo De Francesco Power Quality Consultant Page 1 Contents 1 EXECUTIVE SUMMARY... 4 2 INTRODUCTION... 5 2.1 SITE MONITORED...

More information

Power Quality Requirements for Connection to the Transmission System

Power Quality Requirements for Connection to the Transmission System Power Quality Requirements for Connection to the Transmission System Revision: 1.0 Date: September 2015 Introduction and Purpose of this Document The purpose of this document is to provide clarity to Customers

More information

III CHAPTER. Analysis Methodology. regarding data sampling, benchmarking indices and overall program layouts.

III CHAPTER. Analysis Methodology. regarding data sampling, benchmarking indices and overall program layouts. III CHAPTER Analysis Methodology CHAPTER 3 ANALYSIS METHODOLOGY: Provides information on preliminary design aspects regarding data sampling, benchmarking indices and overall program layouts. 3.1 INTRODUCTION

More information

Power Quality Report. A Manufacturing Plant

Power Quality Report. A Manufacturing Plant Power Quality Report Prepared for A Manufacturing Plant 6 May 2016 by Dr Angelo De Francesco Power Quality Consultant CHK Power Quality Pty Ltd Page 1 Contents 1 EXECUTIVE SUMMARY... 4 2 INTRODUCTION...

More information

PQ Monitoring Standards

PQ Monitoring Standards Characterization of Power Quality Events Charles Perry, EPRI Chair, Task Force for PQ Characterization E. R. Randy Collins, Clemson University Chair, Working Group for Monitoring Electric Power Quality

More information

IEEE 1547: Standard for Interconnection and Interoperability of Distributed Energy Resources with Associated Electric Power Systems Interfaces

IEEE 1547: Standard for Interconnection and Interoperability of Distributed Energy Resources with Associated Electric Power Systems Interfaces IEEE PES Boston Chapter Technical Meeting IEEE 1547: Standard for Interconnection and Interoperability of Distributed Energy Resources with Associated Electric Power Systems Interfaces P1547 Chair David

More information

Assessing network compliance for power quality performance

Assessing network compliance for power quality performance University of Wollongong Research Online Faculty of Engineering and Information Sciences - Papers: Part A Faculty of Engineering and Information Sciences 214 Assessing network compliance for power quality

More information

PROVISION OF DIFFERENTIATED POWER QUALITY USING NETWORK BASED MITIGATION SOLUTIONS

PROVISION OF DIFFERENTIATED POWER QUALITY USING NETWORK BASED MITIGATION SOLUTIONS PROVISION OF DIFFERENTIATED POWER QUALITY USING NETWORK BASED MITIGATION SOLUTIONS Huilian LIAO Jahangir SAIF J. V. MILANOVIĆ University of Manchester U.K. UK Power Networks U.K. University of Manchester

More information

Monitoring Locations in Smart Grids 14PESGM2391

Monitoring Locations in Smart Grids 14PESGM2391 1 Panel Session PQ Monitoring in the Era of Smart Grids Monitoring Locations in Smart Grids 14PESGM2391 Francisc Zavoda IREQ (HQ) QUÉBEC, CANADA Power System and Monitoring Locations 2 Power System Classic

More information

Impact of Harmonic Resonance and V-THD in Sohar Industrial Port C Substation

Impact of Harmonic Resonance and V-THD in Sohar Industrial Port C Substation Impact of Harmonic Resonance and V-THD in Sohar Industrial Port C Substation R. S. Al Abri, M. H. Albadi, M. H. Al Abri, U. K. Al Rasbi, M. H. Al Hasni, S. M. Al Shidi Abstract This paper presents an analysis

More information

ISO Rules Part 500 Facilities Division 502 Technical Requirements Section Wind Aggregated Generating Facilities Technical Requirements

ISO Rules Part 500 Facilities Division 502 Technical Requirements Section Wind Aggregated Generating Facilities Technical Requirements Applicability 1(1) Section 502.1 applies to the ISO, and subject to the provisions of subsections 1(2), (3) and (4) to any: (a) a new wind aggregated generating facility to be connected to the transmission

More information

Impact of Distributed Generation on Voltage Regulation by ULTC Transformer using Various Existing Methods

Impact of Distributed Generation on Voltage Regulation by ULTC Transformer using Various Existing Methods Proceedings of the th WSEAS International Conference on Power Systems, Beijing, China, September -, 200 Impact of Distributed Generation on Voltage Regulation by ULTC Transformer using Various Existing

More information

MEASUREMENT CAMPAIGN AND ASSESSMENT OF THE QUALITY OF SUPPLY IN RES AND DG FACILITIES IN SPAIN

MEASUREMENT CAMPAIGN AND ASSESSMENT OF THE QUALITY OF SUPPLY IN RES AND DG FACILITIES IN SPAIN MEASUREMENT CAMPAIGN AND ASSESSMENT OF THE QUALITY OF SUPPLY IN RES AND DG FACILITIES IN SPAIN Eugenio PEREA*, Eduardo ZABALA*, J. Emilio RODRÍGUEZ*, Asier GIL DE MURO*, Hugo GAGO * * * Fundación LABEIN,

More information

th International Conference on Harmonics and Quality of Power (ICHQP 2016)

th International Conference on Harmonics and Quality of Power (ICHQP 2016) 2016 17th International Conference on Harmonics and Quality of Power (ICHQP 2016) Belo Horizonte, Brazil 16-19 October 2016 s 1-512 IEEE Catalog : ISBN: CFP16CHP-POD 978-1-5090-3793-3 1/2 Copyright 2016

More information

Power Quality and Digital Protection Relays

Power Quality and Digital Protection Relays Power Quality and Digital Protection Relays I. Zamora 1, A.J. Mazón 2, V. Valverde, E. Torres, A. Dyśko (*) Department of Electrical Engineering - University of the Basque Country Alda. Urquijo s/n, 48013

More information

Power Quality Summary

Power Quality Summary Power Quality Summary This article provides an overview of how voltage harmonic distortion is managed on the distribution network and focuses on the current at future issues surround the connection of

More information

ISO Rules Part 500 Facilities Division 502 Technical Requirements Section Aggregated Generating Facilities Technical Requirements

ISO Rules Part 500 Facilities Division 502 Technical Requirements Section Aggregated Generating Facilities Technical Requirements Division 502 Technical Applicability 1(1) Section 502.1 applies to: Expedited Filing Draft August 22, 2017 the legal owner of an aggregated generating facility directly connected to the transmission system

More information

QUESTION BANK PART - A

QUESTION BANK PART - A QUESTION BANK SUBJECT: EE6005-Power Quality SEM / YEAR: VII SEMESTER / ACADEMIC YEAR 08-09 UNIT I - INTRODUCTION TO POWER QUALITY Terms and definitions: Overloading - under voltage - over voltage. Concepts

More information

Harmonic distortion Blackouts Under or over voltage Dips (or sags) and surges, Transients.

Harmonic distortion Blackouts Under or over voltage Dips (or sags) and surges, Transients. Power Quality Standards in India Power Quality is a measure of an ideal power supply system. It can be defined as any power problem manifested in voltage, current and frequency deviations that result in

More information

POWER QUALITY SPECIFICATIONS AND GUIDELINES FOR CUSTOMERS ENGINEERING STANDARDS CITY OF LETHBRIDGE ELECTRIC

POWER QUALITY SPECIFICATIONS AND GUIDELINES FOR CUSTOMERS ENGINEERING STANDARDS CITY OF LETHBRIDGE ELECTRIC CITY OF LETHBRIDGE ELECTRIC ENGINEERING STANDARDS POWER QUALITY SPECIFICATIONS AND GUIDELINES FOR CUSTOMERS The City of Lethbridge acknowledges the use of other utility industry and industry committee

More information

HARMONICS ANALYSIS USING SEQUENTIAL-TIME SIMULATION FOR ADDRESSING SMART GRID CHALLENGES

HARMONICS ANALYSIS USING SEQUENTIAL-TIME SIMULATION FOR ADDRESSING SMART GRID CHALLENGES HARMONICS ANALYSIS USING SEQUENTIAL-TIME SIMULATION FOR ADDRESSING SMART GRID CHALLENGES Davis MONTENEGRO Roger DUGAN Gustavo RAMOS Universidad de los Andes Colombia EPRI U.S.A. Universidad de los Andes

More information

POWER QUALITY IMPACTS AND MITIGATION OF DISTRIBUTED SOLAR POWER

POWER QUALITY IMPACTS AND MITIGATION OF DISTRIBUTED SOLAR POWER POWER QUALITY IMPACTS AND MITIGATION OF DISTRIBUTED SOLAR POWER Presented by Ric Austria, Principal at Pterra Consulting to the IEEE San Francisco Chapter Feb 17, 2016 California Public Utilities Commission,

More information

Hybrid Anti-Islanding Algorithm for Utility Interconnection of Distributed Generation

Hybrid Anti-Islanding Algorithm for Utility Interconnection of Distributed Generation Hybrid Anti-Islanding Algorithm for Utility Interconnection of Distributed Generation Maher G. M. Abdolrasol maher_photo@yahoo.com Dept. of Electrical Engineering University of Malaya Lembah Pantai, 50603

More information

Harmonic Aggregation Techniques for Power Quality Assesment A review of different methods

Harmonic Aggregation Techniques for Power Quality Assesment A review of different methods Harmonic Aggregation Techniques for Power Quality Assesment A review of different methods M.M.Share Pasand Department of Electrical and Electronics Engineering Standard Research Institute- SRI Alborz,

More information

POWER CORPORATION. Power Quality. Specifications and Guidelines for Customers. Phone: Fax:

POWER CORPORATION. Power Quality. Specifications and Guidelines for Customers. Phone: Fax: POWER CORPORATION Power Quality Specifications and Guidelines for Customers Phone: 403-514-3700 Fax: 403-514-3719 1 GENERAL OVERVIEW........................................ 1.1 WHAT DOES THIS SPECIFICATION

More information

PROVISION OF DIFFERENTIATED VOLTAGE SAG PERFORMANCE USING FACTS DEVICES

PROVISION OF DIFFERENTIATED VOLTAGE SAG PERFORMANCE USING FACTS DEVICES rd International Conference on Electricity Distribution Lyon, - June Paper PROVISIO OF DIFFERETIATED VOLTAGE SAG PERFORMACE USIG FACTS DEVICES Huilian LIAO Sami ABDELRAHMA Jovica V. MILAOVIĆ University

More information

ASSESSMENT OF HARMONIC DISTORTION LEVELS IN LV NETWORKS WITH INCREASING PENETRATION LEVELS OF INVERTER CONNECTED EMBEDDED GENERATION

ASSESSMENT OF HARMONIC DISTORTION LEVELS IN LV NETWORKS WITH INCREASING PENETRATION LEVELS OF INVERTER CONNECTED EMBEDDED GENERATION ASSESSMENT OF HARMONIC DISTORTION LEVELS IN LV NETWORKS WITH INCREASING PENETRATION LEVELS OF INVERTER CONNECTED EMBEDDED GENERATION Adam DYŚKO, Graeme M. BRT, James R. McDONALD niversity of Strathclyde

More information

ELECTRICITY ASSOCIATION SERVICES LIMITED 2001

ELECTRICITY ASSOCIATION SERVICES LIMITED 2001 ELECTRICITY ASSOCIATION SERVICES LIMITED 2001 All rights reserved. No part of this publication may be reproduced, stored in a retrieval system, or transmitted in any form or by any means, electronic, mechanical,

More information

Wind Power Facility Technical Requirements CHANGE HISTORY

Wind Power Facility Technical Requirements CHANGE HISTORY CHANGE HISTORY DATE VERSION DETAIL CHANGED BY November 15, 2004 Page 2 of 24 TABLE OF CONTENTS LIST OF TABLES...5 LIST OF FIGURES...5 1.0 INTRODUCTION...6 1.1 Purpose of the Wind Power Facility Technical

More information

Emicon Engineering Consultants L.L.C.

Emicon Engineering Consultants L.L.C. Emicon Engineering Consultants L.L.C. Power Quality Consulting & Solutions Presentation / Pre-Qualification Emicon, Specialised in Power Quality Consulting and Pollution Control on Electrical Network www.emiconconsultants.com

More information

Multiconverter Unified Power-Quality Conditioning System: MC-UPQC T.Charan Singh, L.Kishore, T.Sripal Reddy

Multiconverter Unified Power-Quality Conditioning System: MC-UPQC T.Charan Singh, L.Kishore, T.Sripal Reddy Multiconverter Unified Power-Quality Conditioning System: MC-UPQC T.Charan Singh, L.Kishore, T.Sripal Reddy Abstract This paper presents a new unified power-quality conditioning system (MC-UPQC), capable

More information

Harmonics Issues that Limit Solar Photovoltaic Generation on Distribution Circuits

Harmonics Issues that Limit Solar Photovoltaic Generation on Distribution Circuits WREF 01 Paper # 048 Harmonics Issues that Limit Solar Photovoltaic Generation on Distribution Circuits Ketut Dartawan Ricardo Austria, Le Hui and Mark Suehiro* Pterra Consulting Maui Electric Company*

More information

Modelling of emission of PV inverters and Electric vehicles based on measurements

Modelling of emission of PV inverters and Electric vehicles based on measurements 1 Modelling of emission of PV inverters and Electric vehicles based on measurements Panel on Harmonics from 2 khz to 150 khz: Immunity, Emission, Assessment and Compatibility 15PESGM0912 Jan Meyer Matthias

More information

Modelling of emission of PV inverters and Electric vehicles based on measurements

Modelling of emission of PV inverters and Electric vehicles based on measurements 1 Modelling of emission of PV inverters and Electric vehicles based on measurements Panel on Harmonics from 2 khz to 150 khz: Immunity, Emission, Assessment and Compatibility 15PESGM0912 Jan Meyer Matthias

More information

Harmonic Planning Levels for Australian Distribution Systems

Harmonic Planning Levels for Australian Distribution Systems Abstract Harmonic Planning Levels for Australian Distribution Systems V.J. Gosbell 1, V.W. Smith 1, D. Robinson 1 and W. Miller 2 1 Integral Energy Power Quality Centre, University of Wollongong 2 Standards

More information

Huilian Liao, Member, IEEE, Sami Abdelrahman, Graduate Student Member, IEEE, and Jovica V. Milanović, Fellow, IEEE

Huilian Liao, Member, IEEE, Sami Abdelrahman, Graduate Student Member, IEEE, and Jovica V. Milanović, Fellow, IEEE >ACCEPTED VERSION OF THE PAPER< Zonal Mitigation of Power Quality Using FACTS Devices for Provision of Differentiated Quality of Electricity Supply in Networks with Renewable Generation Huilian Liao, Member,

More information

THE IMPACT OF NETWORK SPLITTING ON FAULT LEVELS AND OTHER PERFORMANCE MEASURES

THE IMPACT OF NETWORK SPLITTING ON FAULT LEVELS AND OTHER PERFORMANCE MEASURES THE IMPACT OF NETWORK SPLITTING ON FAULT LEVELS AND OTHER PERFORMANCE MEASURES C.E.T. Foote*, G.W. Ault*, J.R. McDonald*, A.J. Beddoes *University of Strathclyde, UK EA Technology Limited, UK c.foote@eee.strath.ac.uk

More information

Section 11: Power Quality Considerations Bill Brown, P.E., Square D Engineering Services

Section 11: Power Quality Considerations Bill Brown, P.E., Square D Engineering Services Section 11: Power Quality Considerations Bill Brown, P.E., Square D Engineering Services Introduction The term power quality may take on any one of several definitions. The strict definition of power quality

More information

North-West University, 2 Eskom SOC South Africa

North-West University, 2 Eskom SOC South Africa Electricity Supply to Africa and Developing Economies. Challenges and opportunities. Technology solutions and innovations for developing economies Advances in Power Quality Requirements for RPPs MG BOTHA

More information

Modified Three-Phase Four-Wire UPQC Topology with Reduced DC-Link Voltage Rating

Modified Three-Phase Four-Wire UPQC Topology with Reduced DC-Link Voltage Rating Modified Three-Phase Four-Wire UPQC Topology with Reduced DC-Link Voltage Rating P.Ankineedu Prasad 1, N.Venkateswarlu 2. V.Ramesh 3, L.V.Narasimharao 4 Assistant Professor 12 & Professor 4& Research Scholar

More information

PowerMonitor 5000 Family Advanced Metering Functionality

PowerMonitor 5000 Family Advanced Metering Functionality PowerMonitor 5000 Family Advanced Metering Functionality Steve Lombardi, Rockwell Automation The PowerMonitor 5000 is the new generation of high-end electrical power metering products from Rockwell Automation.

More information

Impact of Power Quality Issues and their Improvement in a Cogeneration Plant

Impact of Power Quality Issues and their Improvement in a Cogeneration Plant Impact of Power Quality Issues and their Improvement in a Cogeneration Plant Keerthi Jayaraj PG Student, M.Tech [Power Sytems], Dept. of Electrical and Electronics, Saintgits College of Engineering, Kottayam,

More information

Table of Contents. Introduction... 1

Table of Contents. Introduction... 1 Table of Contents Introduction... 1 1 Connection Impact Assessment Initial Review... 2 1.1 Facility Design Overview... 2 1.1.1 Single Line Diagram ( SLD )... 2 1.1.2 Point of Disconnection - Safety...

More information

Improvement of Power Quality Using a Hybrid Interline UPQC

Improvement of Power Quality Using a Hybrid Interline UPQC Improvement of Power Quality Using a Hybrid Interline UPQC M.K.Elango 1, C.Vengatesh Department of Electrical and Electronics Engineering K.S.Rangasamy College of Technology Tiruchengode, Tamilnadu, India

More information

A robust voltage unbalance allocation methodology based on the IEC/TR guidelines

A robust voltage unbalance allocation methodology based on the IEC/TR guidelines University of Wollongong Research Online Faculty of Engineering - Papers (Archive) Faculty of Engineering and Information Sciences 2009 A robust voltage unbalance allocation methodology based on the IEC/TR

More information

IMPLEMENTING A SYSTEMATIC APPROACH TOWARDS SOLVING POWER QUALITY COMPLAINTS FROM A NETWORK OPERATOR S PERSPECTIVE

IMPLEMENTING A SYSTEMATIC APPROACH TOWARDS SOLVING POWER QUALITY COMPLAINTS FROM A NETWORK OPERATOR S PERSPECTIVE IMPLEMENTING A SYSTEMATIC APPROACH TOWARDS SOLVING POWER QUALITY COMPLAINTS FROM A NETWORK OPERATOR S PERSPECTIVE Sharmistha BHATTACHARYYA Endinet The Netherlands sharmirb@yahoo.com ABSTRACT In Europe,

More information

Monitoring power quality beyond EN and IEC

Monitoring power quality beyond EN and IEC Monitoring power quality beyond EN 50160 and IEC 61000-4-30 by A Broshi and E Kadec, Elspec, Israel The standards currently in place provide minimum requirements, since they want to create a level playing

More information

OVERVIEW OF IEEE STD GUIDE FOR VOLTAGE SAG INDICES

OVERVIEW OF IEEE STD GUIDE FOR VOLTAGE SAG INDICES OVERVIEW OF IEEE STD 1564-2014 GUIDE FOR VOLTAGE SAG INDICES ABSTRACT Daniel SABIN Electrotek Concepts USA d.sabin@ieee.org IEEE Std 1564-2014 Guide for Voltage Sag Indices is a new standard that identifies

More information

Power Quality implications of new residential appliances. EEA Conference & Exhibition 2010, June 2010, Christchurch

Power Quality implications of new residential appliances. EEA Conference & Exhibition 2010, June 2010, Christchurch Power Quality implications of new residential appliances Stewart Hardie 1 and Neville Watson 2 1 EPECentre, Christchurch, New Zealand 2 University of Canterbury, Christchurch, New Zealand Presenter: Stewart

More information

Adaptive ANN based STATCOM and DVR for optimal integration of wind energy with grid using permanent magnet synchronous generator

Adaptive ANN based STATCOM and DVR for optimal integration of wind energy with grid using permanent magnet synchronous generator Adaptive ANN based STATCOM and DVR for optimal integration of wind energy with grid using permanent magnet synchronous generator Priyanka Sahu Columbia Institute of Engineering and Technology, Raipur,

More information

INCREASING NETWORK CAPACITY BY OPTIMISING VOLTAGE REGULATION ON MEDIUM AND LOW VOLTAGE FEEDERS

INCREASING NETWORK CAPACITY BY OPTIMISING VOLTAGE REGULATION ON MEDIUM AND LOW VOLTAGE FEEDERS INCREASING NETWORK CAPACITY BY OPTIMISING VOLTAGE REGULATION ON MEDIUM AND LOW VOLTAGE FEEDERS Carter-Brown Clinton Eskom Distribution - South Africa cartercg@eskom.co.za Gaunt CT University of Cape Town

More information

INTERIM ARRANGEMENTS FOR GRID TIED DISTRIBUTED ENERGY RESOURCES. Technical Requirements for Grid-Tied DERs

INTERIM ARRANGEMENTS FOR GRID TIED DISTRIBUTED ENERGY RESOURCES. Technical Requirements for Grid-Tied DERs INTERIM ARRANGEMENTS FOR GRID TIED DISTRIBUTED ENERGY RESOURCES Technical Requirements for Grid-Tied DERs Projects Division 6/29/2017 Contents 1 Definitions and Acronyms... 1 2 Technical Interconnection

More information

Power Quality in Metering

Power Quality in Metering Power Quality in Metering Ming T. Cheng Directory of Asian Operations 10737 Lexington Drive Knoxville, TN 37932 Phone: (865) 218.5885 PQsynergy2012 www.powermetrix.com Focus of this Presentation How power

More information

The effect of data aggregation interval on voltage results

The effect of data aggregation interval on voltage results University of Wollongong Research Online Faculty of Engineering - Papers (Archive) Faculty of Engineering and Information Sciences 2007 The effect of data aggregation interval on voltage results Sean Elphick

More information

Simulation and Comparison of DVR and DSTATCOM Used For Voltage Sag Mitigation at Distribution Side

Simulation and Comparison of DVR and DSTATCOM Used For Voltage Sag Mitigation at Distribution Side Simulation and Comparison of DVR and DSTATCOM Used For Voltage Sag Mitigation at Distribution Side 1 Jaykant Vishwakarma, 2 Dr. Arvind Kumar Sharma 1 PG Student, High voltage and Power system, Jabalpur

More information

Technical Requirements for Connecting Small Scale PV (sspv) Systems to Low Voltage Distribution Networks

Technical Requirements for Connecting Small Scale PV (sspv) Systems to Low Voltage Distribution Networks 2014 Technical Requirements for Connecting Small Scale PV (sspv) Systems to Low Voltage Distribution Networks This document specifies the technical requirement for connecting sspv to the low voltage distribution

More information

SIGNAL PROCESSING OF POWER QUALITY DISTURBANCES

SIGNAL PROCESSING OF POWER QUALITY DISTURBANCES SIGNAL PROCESSING OF POWER QUALITY DISTURBANCES MATH H. J. BOLLEN IRENE YU-HUA GU IEEE PRESS SERIES I 0N POWER ENGINEERING IEEE PRESS SERIES ON POWER ENGINEERING MOHAMED E. EL-HAWARY, SERIES EDITOR IEEE

More information

Intermittent Renewable Resources (Wind and PV) Distribution Connection Code (DCC) At Medium Voltage (MV)

Intermittent Renewable Resources (Wind and PV) Distribution Connection Code (DCC) At Medium Voltage (MV) Intermittent Renewable Resources (Wind and PV) Distribution Connection Code (DCC) At Medium Voltage (MV) IRR-DCC-MV 1. Introduction 1 IRR-DCC-MV 2. Scope 1 IRR-DCC-MV 2.1. General 1 IRR-DCC-MV 2.2. Affected

More information

Recommendations for Harmonic and Power Measurements in Electrical Networks

Recommendations for Harmonic and Power Measurements in Electrical Networks Recommendations for Harmonic and Power Measurements in Electrical Networks Analysis implementation and reporting requirements regarding grid power quality and equipment loads Power Capacitor Division Recommendations

More information

MEASUREMENT METHODS OF THE CHARACTERISTICS AND TARGET VALUES OF THE VOLTAGE QUALITY SUPPLIED BY HYDRO-QUEBEC SYSTEM

MEASUREMENT METHODS OF THE CHARACTERISTICS AND TARGET VALUES OF THE VOLTAGE QUALITY SUPPLIED BY HYDRO-QUEBEC SYSTEM MEASREMENT METHODS OF THE CHARACTERISTICS AND TARGET VALES OF THE VOLTAGE QALITY SPPLIED BY HYDRO-QEBEC SYSTEM Report no.: IREQ-99-0 Revision #1 This document is a translated version of the French document

More information

INVESTIGATING THE BENEFITS OF MESHING REAL UK LV NETWORKS

INVESTIGATING THE BENEFITS OF MESHING REAL UK LV NETWORKS INVESTIGATING THE BENEFITS OF MESHING REAL UK LV NETWORKS Muhammed S. AYDIN Alejandro NAVARRO Espinosa Luis F. OCHOA The University of Manchester UK The University of Manchester UK The University of Manchester

More information

VOLTAGE MANAGEMENT BY THE APPORTIONMENT OF TOTAL VOLTAGE DROP IN THE PLANNING AND OPERATION OF COMBINED MEDIUM AND LOW VOLTAGE DISTRIBUTION SYSTEMS

VOLTAGE MANAGEMENT BY THE APPORTIONMENT OF TOTAL VOLTAGE DROP IN THE PLANNING AND OPERATION OF COMBINED MEDIUM AND LOW VOLTAGE DISTRIBUTION SYSTEMS 66 SOUTH AFRICAN INSTITUTE OF ELECTRICAL ENGINEERS Vol.97(1) March 2006 VOLTAGE MANAGEMENT BY THE APPORTIONMENT OF TOTAL VOLTAGE DROP IN THE PLANNING AND OPERATION OF COMBINED MEDIUM AND LOW VOLTAGE DISTRIBUTION

More information

Fault Ride Through Technical Assessment Report Template

Fault Ride Through Technical Assessment Report Template Fault Ride Through Technical Assessment Report Template Notes: 1. This template is intended to provide guidelines into the minimum content and scope of the technical studies required to demonstrate compliance

More information

Power Quality Survey in a Distribution System, Standard Procedures and Limitations. H. Mokhtari S. Hasani and M. Masoudi

Power Quality Survey in a Distribution System, Standard Procedures and Limitations. H. Mokhtari S. Hasani and M. Masoudi THD Voltage Ubc Power Quality Survey in a Distribution System, Standard Procedures and Limitations H. Mokhtari S. Hasani and M. Masoudi Associate Professor Department of Electrical Engineering Sharif University

More information

Sensor Accuracy and Data Management Issues

Sensor Accuracy and Data Management Issues 8/2/21 1 Sensor Accuracy and Data Management Issues Panel on Experiences and System Requirements for Power Quality Data Analytics 1PESGM912 Jan Meyer Robert Stiegler Etienne Gasch Max Domagk Technische

More information

Emission on the low voltage grid: measurements in an urban area

Emission on the low voltage grid: measurements in an urban area International Conference on Renewable Energies and Power Quality (ICREPQ 4) Cordoba (Spain), 8 th to th April, 4 exçxãtuäx XÇxÜzç tçw céãxü dâtä àç ]ÉâÜÇtÄ (RE&PQJ) ISSN 7-38 X, No., April 4 Emission on

More information

Symmetrical Components in Analysis of Switching Event and Fault Condition for Overcurrent Protection in Electrical Machines

Symmetrical Components in Analysis of Switching Event and Fault Condition for Overcurrent Protection in Electrical Machines Symmetrical Components in Analysis of Switching Event and Fault Condition for Overcurrent Protection in Electrical Machines Dhanashree Kotkar 1, N. B. Wagh 2 1 M.Tech.Research Scholar, PEPS, SDCOE, Wardha(M.S.),India

More information

Power Quality Disturbaces Clasification And Automatic Detection Using Wavelet And ANN Techniques

Power Quality Disturbaces Clasification And Automatic Detection Using Wavelet And ANN Techniques International Journal of Engineering Research and Development e-issn: 2278-067X, p-issn: 2278-800X, www.ijerd.com Volume 13, Issue 6 (June 2017), PP.61-67 Power Quality Disturbaces Clasification And Automatic

More information

B.Tech Academic Projects EEE (Simulation)

B.Tech Academic Projects EEE (Simulation) B.Tech Academic Projects EEE (Simulation) Head office: 2 nd floor, Solitaire plaza, beside Image Hospital, Ameerpet Ameerpet : 040-44433434, email id : info@kresttechnology.com Dilsukhnagar : 9000404181,

More information

MODELING THE EFFECTIVENESS OF POWER ELECTRONICS BASED VOLTAGE REGULATORS ON DISTRIBUTION VOLTAGE DISTURBANCES

MODELING THE EFFECTIVENESS OF POWER ELECTRONICS BASED VOLTAGE REGULATORS ON DISTRIBUTION VOLTAGE DISTURBANCES MODELING THE EFFECTIVENESS OF POWER ELECTRONICS BASED VOLTAGE REGULATORS ON DISTRIBUTION VOLTAGE DISTURBANCES James SIMONELLI Olivia LEITERMANN Jing HUANG Gridco Systems USA Gridco Systems USA Gridco Systems

More information

PQ for Industrial Benchmarking with various methods to improve. Tushar Mogre.

PQ for Industrial Benchmarking with various methods to improve. Tushar Mogre. General PQ: Power Quality has multiple issues involved. Thus, need to have some benchmarking standards. Very little is spoken about the LT supply installation within an industry. There is need to understand

More information

PRC Generator Relay Loadability. Guidelines and Technical Basis Draft 4: (June 10, 2013) Page 1 of 75

PRC Generator Relay Loadability. Guidelines and Technical Basis Draft 4: (June 10, 2013) Page 1 of 75 PRC-025-1 Introduction The document, Power Plant and Transmission System Protection Coordination, published by the NERC System Protection and Control Subcommittee (SPCS) provides extensive general discussion

More information

Fundamentals of Power Quality

Fundamentals of Power Quality NWEMS Fundamentals of Power Quality August 20 24, 2018 Seattle, WA Track D Anaisha Jaykumar (SEL) Class Content» Introduction to power quality (PQ)» Causes of poor PQ and impact of application» PQ characteristics»

More information

Compensation of Reactive Power Case Study

Compensation of Reactive Power Case Study 15 Compensation of Reactive Power Case Study Stefan Fassbinder and Alan Ascolari C15.1 FLUORESCENT LAMP Fluorescent lamps are the only common device where putting the most efficient way of compensation,

More information

Measurement tools at heart of Smart Grid need calibration to ensure reliability

Measurement tools at heart of Smart Grid need calibration to ensure reliability Measurement tools at heart of Smart Grid need calibration to ensure reliability Smart grid; PMU calibration position 1 The North American interconnections, or electric transmission grids, operate as a

More information

Case Studies in PQ Monitoring

Case Studies in PQ Monitoring 1 Case Studies in PQ Monitoring Panel on PQ Monitoring in the Era of the Smart Grid 14PESGM2766 Jan Meyer Etienne Gasch Matthias Klatt Technische Universitaet Dresden, Germany 2 Agenda Objectives of Power

More information

Discussion on the Deterministic Approaches for Evaluating the Voltage Deviation due to Distributed Generation

Discussion on the Deterministic Approaches for Evaluating the Voltage Deviation due to Distributed Generation Discussion on the Deterministic Approaches for Evaluating the Voltage Deviation due to Distributed Generation TSAI-HSIANG CHEN a NIEN-CHE YANG b Department of Electrical Engineering National Taiwan University

More information

Power Quality - 1. Introduction to Power Quality. Content. Course. Ljubljana, Slovenia 2013/14. Prof. dr. Igor Papič

Power Quality - 1. Introduction to Power Quality. Content. Course. Ljubljana, Slovenia 2013/14. Prof. dr. Igor Papič Course Power Quality - 1 Ljubljana, Slovenia 2013/14 Prof. dr. Igor Papič igor.papic@fe.uni-lj.si Introduction to Power Quality Content Session 1 Session 2 Session 3 Session 4 1st day 2nd day 3rd day 4th

More information

A Practical Application of Low Voltage DC Distribution Network Within Buildings

A Practical Application of Low Voltage DC Distribution Network Within Buildings A Practical Application of Low Voltage DC Distribution Network Within Buildings B Marah 1 Hoare Lea LLP. London, UK bmarah@theiet.org Y R Bhavanam 2, G A Taylor 2, M K Darwish 2 Brunel University London.

More information

Voltage Sag and Swell Mitigation Using Dynamic Voltage Restore (DVR)

Voltage Sag and Swell Mitigation Using Dynamic Voltage Restore (DVR) Voltage Sag and Swell Mitigation Using Dynamic Voltage Restore (DVR) Mr. A. S. Patil Mr. S. K. Patil Department of Electrical Engg. Department of Electrical Engg. I. C. R. E. Gargoti I. C. R. E. Gargoti

More information