Low Voltage System State Estimation Using Smart Meters

Transcription

1 Low Voltage System State Estimation Using Smart Meters Ahmad Abdel-Majeed IEH, University of Stuttgart, Germany Martin Braun IEH, University of Stuttgart, Germany Fraunhofer IWES, Germany Abstract Accurate and reliable state estimation for low voltage networks is the core stone for flexible operation and control in the current transaction from passive conventional to active smart grids. The development in the energy supply has revealed a rapid increase of controllable distributed generators, consumer s installation, stationary storage systems and electric vehicles. This development leads to a significantly different kind of system behavior which must be understood first, and then to make suggestions for operational network improvements in order to increase the security and efficiency of the distribution system operation. For this, a high chronological and topological resolution of information for the system state estimation in the low voltage level is necessary. However, measurement data are necessary for state estimation, these measurements can be obtained either from the distribution system measurement infrastructure or from the smart meters at the connection points of the customer. The focus of this paper is to obtain the technical feasibility of using smart meter and their measurements for low voltage network observability and controllability through state estimation techniques. Also to analyse the impact of high accuracy measurement data provided from smart meters on the state estimation output accuracy for both voltage and it's phase angle. Index Terms Simulation, Smart Meter, State Estimation. I. INTRODUCTION The current development in the energy supply revealed a rapid increase of controllable distributed generation systems, consumer s installation, stationary storage systems and electric vehicles. This development leads to a significantly different kind of system behaviour that must be understood first, and then make suggestions for improvements of network operations in order to increase the security and efficiency of system operation. In Germany, the number of small scale distributed generation (DG) units especially from renewable energy has increased sharply in the recent years and will grow up significantly in the near future especially the share of wind and solar energy in the total amount of energy generation through a targeted support from the federal government. Due to the low production capacities from DG units from renewable energy, they are normally connected to the medium or low voltage network, this results in problems of power quality and compliance with the voltage range [1]. For the distribution network (medium and low voltage) there are (up to date) a few studies such as cost effective network expansion and operation taking into consideration the future integration of more DG units in Germany. To optimise the integration of renewable energies in the management of distribution networks through distributed generation the active and reactive power control can be taken over a number of other functions in addition to the pure active power infeed. More severe demands on the distributed generation can be found in the AR4105 guidelines for the Generators connected to the low-voltage distribution network []. Some aspects, for example, the active power control, provision of reactive power and the possibility of intervention by the network operator. This requires further study in order to quantify the optimal economic solution. Initial investigations show that by reactive power control of PV systems, the capacity of the low voltage networks can be increased in some cases for more than 40% [3]. A critical element in determining the possible optimisation is the accuracy of state estimation in low and medium voltage networks which is generally designed without measurements. However, currently the distribution network operator has thus no information about the network status, so the possible developments could come through the use of the customer s network ancillary services, and smart meters. II. CHALLENGES In transmission networks, state estimation has been since 1970 the state of the art [4], and now it is considered as a routine task. However these methodologies cannot be directly applied to distribution systems because of these reasons: 1. There are very few real-time measurements available in medium and low voltage levels (several thousand nodes usually have a few measurements at the head of the feeder).. The modeling of complex multi-phase asymmetric distribution represents a challenge for developing efficient and robust estimation algorithms which are suitable for different types of measurements.

2 These two reasons arise from the following fundamental differences between transmission and distribution grids: Transmission grids are meshed and must be analysed as a whole, while distribution networks are usually radially constructed and can be analysed separately as sub-networks. In transmission systems the impedance ration R/X << 1 and can be ignored, but in distribution systems the R/X> 1 or R=X in some cases, and can be no longer neglected. Table.1. provides an overview of typical line data in different voltage levels [4]. Voltage Level Material Overhead line conductors: 750 kv 110 kv Cables: kv 1 kv Table.1. Typical line data Cross section [mm²] Al/St 805/10 R' [Ohm/ km] X' [Ohm/ km] R/X Al/St 435/ Type Al NAXSY 3x50sm 6/10kV Cu NXSY 3x150rm 1/0kV Cu NXSY 1x500R M 18/30kV Al NAXY 3x50sm 0.6/1kV Al NAXY 3x150sm 0.6/1kV Cu NXY 3x300sm 0.6/1kV In a transmission system there are fairly balanced loads, in contrast to distribution grids particularly low voltage grids, an obvious asymmetry exists between the phase loading, therefore a much more complex calculation is needed. The transmission grid is usually observed by an adequate number of measurement points on the network. At low and medium voltage networks, the network operator has usually no or very few measuring points compared with approximately 1,000-10,000 nodes. Distribution network operators are thus generally "blind". This can change in future with the use of a smart meter infrastructure. The number of nodes in the distribution may even be higher than the transmission system. However, the specific cost for measurement, information and communication infrastructure per unit of transmitted energy is much higher. In order to analyse the different behavior between transmission and distribution networks. The interdependencies and coupling between active and reactive power (P, Q) on one hand and voltage and it's phase angle (U, ) on the other hand will be analysed according to [5][6]. Fig.1. shows the analysed equivalent circuit with two voltage sources and which are connected via the network impedance z with its impedance angle φ N or the resistance to reactance ratio R/X = 1/atan(φ N ). The transfer capability of active power P and reactive power Q is analysed with regard to the voltage difference. Fig. 1. Equivalent circuit of two voltage sources connected via an impedance. With the following basic equations the interdependencies are analysed: Overhead lines in HV (110 kv) and EHV (750 kv) have a ratio of R/X << 1 and can therefore be considered as inductive. Therefore. From the root mean square (rms) voltages, the complex rms current can be calculated as: U = Ûe U = Û Z = Ze jω t U U I = Z = R + cos j sin e jϕ N N jω t sinδ COS N e jδ jx

3 With sinδ COS ² sin cos sin δ U² Q S U cosδ X N X N At LV level (1 kv) the cross section of cables is often below 150 mm² resulting in R/X > 1 and at MV (10-30 kv), the cross section is normally around 150 mm² resulting in R/X 1. Therefore From the rms voltages, the complex rms current can be calculated as: cos j sin cos sin With cos sin cos sin ² COS sin As seen from the equations, for small values of δ the clear functional dependency P(δ) and Q(U) is applicable in transmission networks resulting in simple control functions for frequency and voltage. Also for small values of δ there is no clear functional dependency for P(U, δ) and Q(U, δ) in distribution networks. These interdependencies result in more complex approaches for frequency and voltage control in distribution networks. Summarising these coupling dependencies following approximations can be stated: Transmission network: - Voltage (U) is mainly coupled with reactive power transfer. - Voltage phase angle is coupled with active power transfer. Distribution network: - Voltage (U) is based on both active and reactive power transfer. - Voltage phase angle is coupled with active power transfer. New distribution network management applications, which are only possible through the advanced control facilities together with the measurement data, are needed to keep pace with the current transaction from passive conventional to active smart grids behavior. Table.. shows these functions and their benefits [4]. Table.. Features and benefits of advanced distribution management applications through the use of state estimation. DMS Application Functionality Unbalanced load flow Determination of the line currents and node analysis voltages per phase for the entire distribution system, either online or offline in simulation mode Fault Location Identification of possible fault locations on system. Distribution Volt/Var control Remote switching and restoration Monitoring and control of line capacitors, voltage regulators, and load tap changers (LTCs) to reduce peak load and system losses. Automatic feeder reconfiguration considering network operating conditions Benefits Improved system awareness Higher assest utilisation Improved contingency planning Improved crew efficiencies in managing outages. Reduced customer average interruption duration index (CAIDI) and system average interruption index (SAIDI) Reduced customer demand at system peaks Lower system losses Improved voltage profiles. Reduced CAIDI and SAIDI. Lower system losses. However, the future planned use of smart meters, is now offering new types of measurements in the low voltage level, such as active power, reactive power, voltage and current measurements at each customer connection almost in real time. The availability of this information is an essential basis for state estimation in low voltage networks. With more accurate real-time model of the network through the distribution state estimation, other operational functions such as active and reactive power optimisation, network restoration, load balancing and optimal network configuration can be more reliably performed and controlled.

4 III. STATE ESTIMATION IN THEORY State estimation algorithm is commonly based on the weighted least square method [4], where the state variables (voltage magnitude and phase angle) are determined by the minimisation of the square of the error of all measurements. The basic equation which relates the measurements with the state variables is: Where z is the measurement vector x is the state variable vector h is the nonlinear power flow equations e is the measurement error vector The state vector x can be estimated using different estimator techniques [7]: 1- Weighted Least Squares Estimator. - Least Absolute Value Estimator. 3- Reweighted Least Squares Estimator. Weighted Least Squares Estimator will be used in this paper as it gives a consistent performance under Gaussian assumptions for known noise characteristics [9]. a. WLS estimator Weighted least square method deals with minimisation of the error e, which is done by minimising the cost function as follows : A lot of smart meters application can be found in the literature, the main focus of this paper is to obtain the technical feasibility of using smart meters and their instantaneous measurements (P, Q, U) for the low voltage network observability and controllability through state estimation techniques. An important factor in using the smart meter measurement data is the accuracy, the available smart meters in the market deliver measurements with different accuracy classes A, B and C given with different confidence levels specified from the manufacturer. The smart meter accuracy data are not linear and dependable on many factors, for example the ambient temperature, load, power factor and value of current [10]. In this paper, it will be assumed that the smart meter are operating under normal conditions. V. SIMULATION STUDY In order to evaluate the performance of state estimation in low voltage networks, WLS state estimation technique was applied on a real existing low voltage network in the region Freiamt. This network was kindly supplied by the company EnBW Regional AG. This small low voltage network consists of 8 loads with different sizes, two of the customers have integrated their PV system to the grid. More information about the network topology can be found in Fig.. Where W is the weighting matrix, which is chosen to be the inverse of the covariance matrix of the measurement error vector. b. Measurement function One fundamental requirement of WLS estimator is that the number of measurements m for an observable system has to be equal or larger than the number of the estimated variables (states). The state of the system in this case consists of the nodes voltage magnitudes and their phase angles. In this paper, the measurements data are collected from the smart meters and measurement data from the ancillary data at each DG unit in addition to the measurements collected from the substation at low voltage side. More details about the smart meter measurement sets and their accuracies will be explained in the next section. IV. SMART METER Smart Meter is an advanced energy meter which measures mainly the real time electrical energy consumption of the utility customers, in addition to measuring power quality and instantaneous values such as voltage and current at their connection points. Fig.. Example of low voltage network topology. Every customer is provided with a smart meter with accuracy class B (±1% accuracy for both active and reactive energy under normal conditions), voltage measurements in the smart meter have also ±1% accuracy for the same conditions with confidence level of 95% for all of them. Additional power (P, Q) and voltage measurement were also available from the substation at the transformer low voltage side with ±% accuracy for power measurements (P, Q) and ±1% for voltage measurements with confidence level of 95% for all of them. These smart meters provide the customers (P, Q, U) measurement which is considered as power injection measurements, power flow measurements are not available. It was assumed that errors associated with the measurements are independent and identically distributed. The smart meter measurements are represented as a Gaussian distribution,

5 their standard deviation is defined by the accuracy of the measurements. These smart meters have two way communication channels and are able to transmit their readings nearly in real time every 15 minutes. The state variables which will be considered here are the voltage and it's phase angle at nodes from to10.the PV systems are installed at nodes 3 and 6. In case that smart meter was not able to send the measurement set of data (P, Q, U), the state estimator automatically will use the measurement data from the same period of time in the same day from last week. This measurement data will be modeled as Pseudo with ±5%, ±0% measurement uncertainty for power (active, reactive) and voltage measurements respectively. Equality constraints were added to handle zero injection measurements from smart meters. VI. SIMULATION RESULTS AND DISCUSSIONS a. Observability Simulations Roughly speaking, network is called observable, if the measurements in the system provide enough information to estimate the state of that network. Four cases were simulated to check the observability of the network under the available measurement data combinations: Case 1 represents real time (P, U) measurements together from all nodes. Case represents real time (P, Q) measurements together from all nodes. Case 3 represents real time (P, Q, U) measurements from all nodes. Case 4 represents (Q, U) real time measurements from all nodes Observability simulations show that only in case 4 with (Q, U) measurements the network was not observable. The first three cases were observable with different accuracy for estimating voltage and its phase angle at every node by comparing the estimated values with the true measurement values as shown in Fig.3 and Fig.4. Fig. 3.Estimated Voltage error (%). Fig. 4.Estimated Phase Angle error (%). The following observations were noticed from the simulation results: (P, Q) and (P, Q, U) measurement combinations for voltage estimation have much better accuracies for both voltage and phase angle than in case of (P, U) measurement combination. (P, Q, U) measurement combination shows the best accuracy in estimating the angle among the other two measurement combinations. The accuracy of phase angle estimation in case of (P, U) is found to be only dependable on the voltage measurement accuracy and magnitude. b. Estimation Accuracy Simulations In order to evaluate the performance of this state estimator for (P, Q, U) measurement combination in case of failure in measurement transmission six cases were simulated as follow: Case 1 represents real time (P, Q, U) measurements from all nodes. Case represents real time (P, Q, U) measurements from all nodes except one (P, Q, U) pseudo measurement at node 4. Case 3 represents real time (P, Q, U) measurements from all nodes except one (P, Q, U) pseudo measurement at node 6. Case 4 represents real time (P, Q, U) measurements from all nodes except two (P, Q, U) pseudo measurements at nodes 3,6. Case 5 represents real time (P, Q, U) measurements from all nodes except three (P, Q, U) pseudo measurements at nodes 5,8,10. Case 6 represents real time (P, Q, U) measurements from all nodes except three (P, Q, U) pseudo measurements at nodes 8,9,10. Fig.5 and Fig.6 shows the state estimation accuracy for both voltage and angle respectively by comparing the estimated values with the true measurement values.

6 Fig. 5. Estimated Voltage error (%). Fig. 6. Estimated Phase Angle errorr (%). The following observations were noticed from the simulation results: Combining (P, Q, U) real time smart meter measurements from all nodes results always in the best state estimation accuracy for both voltage and phase angle estimations (less than 0.005% and % error for estimating voltage and phase angle respectively). Losing up to real time smart meter data does not affect the overall voltage and angle estimation accuracy in comparison to losing 3 real time smart meter measurements. In general, losing real time smart meter measurement for big loads will affect the state estimation accuracy distinctly in comparison to small loads. Losing real time smart meter data for nodes where photovoltaic systems is worse than losing 3 real time measurements from smart meters at the end of the feeder. Losing real time smart meter data for nodes where photovoltaic systems are installed will reduce the overall voltage estimation accuracy and more distinctly for phase angle estimation. On one feeder, it was noticed that high state estimation accuracy can be kept if the smart meter measurements were not lost for the first and last customer connected along this feeder. VII. CONCLUSION AND FUTURE WORK Accurate and reliable state estimation for low voltage network is the core stone for flexible operation and control in the current transaction from passive conventional to active smart grids. Therefore a high chronological and topological resolution of the system state estimation in the low voltage level information was presented in this paper using substation and the smart meter measurements real time data in addition to pseudo measurements generated from the historical data from the smart meters measurements. Simulation results for a realistic small low voltage network were shown and analysed. Simulation showed that the use of real time smart meters data combining P, Q and U measurements gives the best estimation accuracy for both voltage and phase angle. On the other hand, in case of losing measurement data from some smart meters, pseudo measurement from the smart meter's historical data were used to maintain the state estimation observability and to maintain the state estimation accuracy under an acceptable level compared to case 1. However, losing at least 3 smart meters (37.5% of the total number of smart meters) worsened the state estimation accuracy for both voltage and phase angle distinctly compared to cases 1, and 3. Losing real time smart meter data from nodes with big loads or installed PV systems also worsened the accuracy distinctly compared to the case of losing data from nodes with small loads. In Future, historical smart meter data can be used in load modeling techniques in orderr to improve their accuracy, which will also improve the output accuracy of state estimation in case of losing more real time measurements and reduce the communication infrastructure costs used to transmit the smart meter data. REFERENCES [1] Abdel-Majeed, Ahmad; Viereck, Robert; Oechsle, Fred; Braun, Martin; Tenbohlen, Stefan;, "Effects of distributed generators from renewable energy on the protection system in distribution networks," 46th International Power Engineering Universities Conference (UPEC). [] VDE-AR-N 4105 "Generators connected to the low-voltage distribution network". Technical requirements for the connection to and parallel operation with low-voltage distribution networks. [3] G. Kerber: Empfehlung zur Richtlinie zum Anschluss von Erzeugungsanlagen an das Niederspannungsnetz, TU-München, 15. Mai 009. [4] Schweppe, F.C.; Wildes, J.;, "Power system static-state estimation, Part I: Exact Model". [5] Martin Braun,"Provision of ancillary services by distributed generators", renewable energies and energy efficiency Band 10 / Vol. 10. [6] Jörg Jahn, "Energiekonditionierung in Niederspannungsnetzen unter besonderer Berücksichtigung der Integration verteilter Energieerzeuger in schwachen Netzausläufern", Renewable Energies and Energy Efficiency Band 10 / Vol. 10. [7] ABB Technik 3/009. [8] Kamireddy, S.; Schulz, N.N.; Srivastava, A.K.;, "Comparison of state estimation algorithms for extreme contingencies," 40th North American Power Symposium, 008. NAPS '08. [9] R. Sing, B. C. Pal and R. A. Jabr, "Choice of estimator for distribution system state estimation," IET Generation, Transmission and Distribution, No.7, Vol. 3, Jul 009. [10] EN Electricity metering equipment (AC) Part 3: Particular requirements Static meters for active energy (classes indexes A, B and C): Static meters for reactive energy, (classes and 3).