EFFECT OF SOLAR PV ON FREQUENCY MANAGEMENT IN NEW ZEALAND

Transcription

1 EFFECT OF SOLAR PV ON FREQUENCY MANAGEMENT IN NEW ZEALAND DECEMBE R 2017 TECHNICAL REPORT

2

3 48TTable of Contents Table of Contents EXECUTIVE SUMMARY... VII 1 INTRODUCTION Programme Overview The PV Generation Investigation Project The New Zealand power system Impacts on frequency regulation CURRENT FREQUENCY MANAGEMENT PRACTICES FREQUENCY IN THE NORMAL BAND STUDY Background Cloudy day study methodology approach Study approach FREQUENCY STABILITY STUDY Impact of PV generation on system inertia PV generation response to frequency deviations Study methodology Effects on System Inertia Effects on Under-Frequency management Effects on Over-Frequency management KEY FINDINGS AND CONCLUSIONS Managing frequency within the normal band Effect of low inertia system Managing an under-frequency event Managing an over-frequency event Effect of low inertia system on under- frequency mitigation measures Study limitations RECOMMENDATIONS Frequency management strategy Inverter standard Long term factors A1 CURRENT FREQUENCY MANAGEMENT PRACTICES A1.1 Contingent event classification A1.2 Managing frequency in the normal band A1.3 Managing Frequency Stability A2 INVERTER MODELS A2.1 Active power/frequency control A2.2 Reactive Power/Voltage control A2.3 Voltage/Frequency ride through A2.4 Reconnect characteristics A3 GENERATION DISPATCH A3.1 Generation dispatch for 630 MW PV generation scenario A3.2 Generation dispatch for 2500 MW PV generation scenario A3.3 Generation dispatch for 3250 MW PV generation scenario A4 SUMMER SUNDAY GENERATION AND GRID ZONES POWER-FLOW Effect of Solar PV on Frequency Management in New Zealand Transpower New Zealand Limited. All rights reserved. iii

4 48TTable of Contents A5 EMERGING ENERGY PROGRAMME: PLAN AND OUTCOME STRATEGY A5.1 Emerging Energy Technologies Outcome Strategy Map GLOSSARY OF TERMS AND ACRONYMS BIBLIOGRAPHY Effect of Solar PV on Frequency Management in New Zealand Transpower New Zealand Limited. All rights reserved. iv

5 Table of Figures Table of Figures Figure 1: GXPs and Weather Stations... 9 Figure 2: Weather Station/GXP Assignment example Figure 3: WVM applied to cloudy day PV Profile Figure 4: Simulated PV generation for GXPs around 100km apart at (a) Auckland using the Auckland Airport irradiance data, and (b) Nelson and Wellington using the Woodburn and Kelburn irradiance data Figure 5: (a) Total PV for three days around noon, scaled to a 3000 MW peak PV generation (b) Histogram of the change in total PV generation for each pair of consecutive minutes Figure 6: (a) PV output in 1min increments for the year. (b) PV output noise after taking the sun position into account Figure 7: Simplified model of the power system Figure 8: Bode plot for a simple frequency model of the system Figure 9: Model vs calculated Bode plots compared Figure 10: PV generation penetration and disturbance scenario Figure 11: Probability distribution for system frequency deviation Figure 12: Governor response, system inertia and system frequency in an under-frequency event Figure 13: PV generator response to Under-Frequency Event Figure 14: A typical PV generation response to an over-frequency event Figure 15: Frequency Management Barometer - Mitigations used to ensure statutory limits and PPOs are maintained during frequency events on the system Figure 16: System inertia plotted against PV generation for North Island and South Island Figure 17: RoCoF for 1 second after various system imbalances, at different levels of PV generation, with and without HVDC enabled (NI) Figure 18: RoCoF for 1 second after various system imbalances, at different levels of PV generation, with and without HVDC enabled (South Island) Figure 19: Frequency response with decreasing system inertia Figure 20: The response comparison of system frequency after a North Island CE when HVDC is not available Figure 21: The response comparison of system frequency after a South Island CE when HVDC is not available Figure 22: generation mix for Sunday in summer Figure 23: North Island System responds under a CE and ECE at different PV penetration levels Figure 24: South Island System responds under a CE and ECE at different PV penetration levels Figure 25: North Island contingent event frequency traces Figure 26: South Island contingent event frequency traces Figure 27: North Island under-frequency response after an ECE of 240 MW Figure 28: South Island under-frequency response after an ECE of 240 MW Figure 29: Likely configuration of AUFLS with embedded generation within the Low voltage network Figure 30: effect of reduced AUFLS in North Island for the 3250 MW PV generation scenario in managing a 240 MW ECE risk Figure 31: X: North Island under-frequency response with PV generation over-frequency support Figure 32: AUFLS operation timing diagram Figure 33 The South Island response with no PV generation support, at various PV penetration levels to a 400 MW demand trip Figure 34: North Island sensitivity to the level of PV generators with over-frequency response Figure 35: South Island sensitivity to the level of PV generators with over-frequency response Figure 36: The North Island over- frequency response for three levels of PV generation with no frequency response Figure 37: The North Island 3250 MW PV penetration response with levels of over- frequency response Figure 38: The South Island high PV over- frequency response for with adjusted OFA tripping times Figure 39 PV generation Over-Frequency Response Effect of Solar PV on Frequency Management in New Zealand Transpower New Zealand Limited. All rights reserved. v

6 Table of Figures Figure 40 Frequency Control component of Inverter C Dynamic Model Figure 41 PV generation Volt-Var Response to voltage step Figure 42 PV generation constant power factor response to over-frequency event Figure 43 Volt-Var component of Inverter Dynamic Model Figure 44 Constant Power Factor component of Inverter Dynamic Model Figure 45 Voltage and Frequency trip and reconnect component of Inverter C Dynamic Model Figure 46: Generation Mix and HVDC bipole transfer Figure 47: North Island inter-grid Zone electrical power-flow Figure 48: Grid Zone 1 generation, demand and inter-grid Zone power-flow Figure 49: Grid Zone 2 generation, demand and inter-grid Zone power-flow Figure 50: Grid Zone 3 generation, demand and inter-grid Zone power-flow Figure 51: Grid Zone 4 generation, demand and inter-grid Zone power-flow Figure 52: Grid Zone 5 generation, demand and inter-grid Zone power-flow Figure 53: Grid Zone 6 generation, demand and inter-grid Zone power-flow Figure 54: Grid Zone 7 generation, demand and inter-grid Zone power-flow Figure 55: Grid Zone 8 generation, demand and inter-grid Zone power-flow Figure 56: South Island inter-grid Zone electrical power-flow Figure 57: Grid Zone 9 generation, demand and inter-grid Zone power-flow Figure 58: Grid Zone 10 generation, demand and inter-grid Zone power-flow Figure 59: Grid Zone 11 generation, demand and inter-grid Zone power-flow Figure 60: Grid Zone 12 generation, demand and inter-grid Zone power-flow Figure 61: Grid Zone 13 generation, demand and inter-grid Zone power-flow Figure 62: Grid Zone 14 generation, demand and inter-grid Zone power-flow Effect of Solar PV on Frequency Management in New Zealand Transpower New Zealand Limited. All rights reserved. vi

7 Executive Summary EXECUTIVE SUMMARY Transpower has initiated a programme of work to investigate the impacts on the power system from an anticipated increase in distributed, non-dispatchable and renewable generation, and from other emerging technologies in New Zealand. The aim of Transpower s Emerging Energy Programme is to identify potential compromise to Transpower s ability to meet the system operator Principal Performance Obligations (PPOs) with the introduction of new generation technologies. The alternative would involve a revision of the PPOs to accommodate these emerging technologies. The first part of the programme features the PV Generation Investigation Project, which studies the effect of PV generation technology on four areas of the power system: generation dispatch, frequency management, transmission voltage management and transient stability. Study reports have been produced for each of these areas, with this report covering the study into frequency management under high penetration levels of solar PV in the power system. All the project studies used a scenario of 4 GW [1] of solar PV capacity installed, as discussed in the generation dispatch study (refer to Effect of Solar PV on Generation Dispatch in New Zealand). This study into frequency management under high penetration levels of solar PV concludes there is no immediate concern for the ability of the New Zealand power system to manage frequency through existing arrangements if PV generation increased to the level of 4 GW installed capacity. This assumes the current generation mix remains inplace; e.g to meet evening maximum demand. It is expected that current system operation and market design will manage frequency risk through procuring sufficient frequency reserve or by constraining down generation from risk plant. The study showed current frequency keeping arrangements are adequate to manage frequency within the normal band when national PV generation varies due to cloud effect. Although there will be some reduction in frequency quality compared to present system conditions, the power system could still be managed to keep frequency within the required +/- 0.2Hz band. Hydro generation is critical for power system solar PV limits, as it provides fast reserve and system inertia that quickly responds to deviations in system balance. When synchronous generators are displaced by PV generation the reduction in system inertia causes an increase in the Rate of Change of Frequency (RoCoF) following a contingent event. This does affect frequency response performance, but the system operator will be able to manage the system risks adequately without violating the frequency limits. Nevertheless, there is a need for ongoing research into fast reserve products to enhance the management of frequency. An interesting feature of the New Zealand power system with a high PV generation output is the reduction of frequency risk as larger thermal units are displaced. This lowering risk reduces the quantity of reserves required to cover the loss of a larger Effect of Solar PV on Frequency Management in New Zealand Transpower New Zealand Limited. All rights reserved. vii

8 Executive Summary thermal unit in the system. The studies also show inter-island HVDC flow was reduced during times of high PV generation output (as more demand is supplied locally by PV generation), thus reducing the impact of the loss of HVDC on frequency management in both islands. Under a high solar PV penetration scenario when the RoCoF is high following a credible loss of generation input to the power system, the current AUFLS system may not work effectively to arrest frequency fall following that event. High RoCoF may cause the AUFLS to over or under shed load, leading to unexpected post-event frequency response. It may prove challenging to design mitigation measures such as AUFLS to work effectively in both the extreme minimum demand (low inertia) in midday when PV generation is at its highest and also at maximum demand conditions in the evening. Transpower s current Over-Frequency Arming (OFA) strategy is adequate to manage the risk of frequency rising above the 52 Hz for North Island and 55 Hz for South Island. The ability of solar PV inverters to ramp down electrical output during an over-frequency event could positively contribute to the management of grid over-frequency events. The study highlighted the importance of the HVDC link operation to share system inertia and reserve between the two islands, even at low transfer levels. The HVDC link can operate in round power mode at low transfer levels to optimise this sharing. In summary, the key findings of this study are: The New Zealand power system can accommodate around 3000 MW of PV generation and maintain acceptable frequency performance within the normal band with the present frequency keeping arrangements. As PV generation levels increase, system inertia will decrease and the RoCoF during a large frequency event will increase. The way we manage large frequency events will need to change. The increase in RoCoF will require us to review the effectiveness of the current frequency mitigation measures, including OFA, IL and AUFLS. The increase in embedded PV generation in the distribution network can affect the effectiveness of the IL and AUFLS. Even with a reduction in the frequency risk, we may need faster reserves to mitigate the increase in RoCoF. The fast and flexible solar inverter control in response to a frequency event can enhance the flexibility and resilience of the future grid. Learnings from this initial investigation, and future work streams, will inform how Transpower needs to adapt to provide a more effective frequency management strategy to facilitate a reliable power system with a greater contribution from the variable renewable generation sources. The learnings gained in these studies will be useful in steering the future of the system operator service, electricity market design, industry regulations, policies and procedures for a period of increasingly decentralised supply and responsive consumer technologies. Effect of Solar PV on Frequency Management in New Zealand Transpower New Zealand Limited. All rights reserved. viii

9 Executive Summary Ultimately, this understanding will facilitate an evolving power system which can continue to meet the changing needs of New Zealanders. Effect of Solar PV on Frequency Management in New Zealand Transpower New Zealand Limited. All rights reserved. ix

10 Section 1 Introduction 1 INTRODUCTION 1.1 Programme Overview Transpower's Emerging Energy Programme investigates the potential impacts on the power system resulting from an anticipated increase in distributed, non-dispatchable, renewable generation and other emerging technologies in New Zealand. The programme outlines the strategy Transpower has adopted to develop its capability and business processes to enable a successful integration of such technologies in the New Zealand power system. See Appendix A5 for a scope of work flowchart that summarises Transpower's Emerging Energy Programme The growth of distributed, non-dispatchable renewable generation Distributed, non-dispatchable electricity generation, primarily PV generation, has grown rapidly in most regions around the world in recent years. The change in technology costs, consumer preferences, policies and environmental concerns, leads to this trend of growth [2]. PV generation uptake is still relatively low in New Zealand. However, the rate of growth is expected to increase for the foreseeable future, with PV generation projected to become a significant part of New Zealand's electricity supply mix. Other emerging technologies (such as energy storage devices, Home Energy Management Systems (HEMS), Electric Vehicles (EVs) and smart appliances) will also play a role in shaping the future of the New Zealand power system. New business models for energy trading and distributed generation ownership will facilitate consumer choice and change the way we produce and use electricity. Though the cumulative effect of these developments is highly interdependent and difficult to predict, the electricity industry will need to be proactive in meeting changing consumer expectations and a shifting market environment, to avoid significant business disruptions Assessing New Zealand's ability to adapt to new technologies The New Zealand power system has some unique features not the least of which is being an islanded system with a high proportion of electricity generated from hydro-power backed by storage. A 2008 study of the system's ability to accommodate wind generation indicated that hydro generation afforded a high degree of flexibility to accommodate variable generation. Transpower is assessing the possible future impacts of variable generation technologies on the power system and the policies it may need to adopt to continue to meet the PPOs in its role as system operator. These assessments will also provide useful context for the future development of the Electricity Industry Participation Code (the Code), including the PPOs The challenges due to the variability of PV Generation Electricity produced from solar irradiance depends on the position of the sun, which is predictable though variable. With consistently clear or overcast weather, PV generation output can be relatively steady; with output increasing in the early morning and Effect of Solar PV on Frequency Management in New Zealand Transpower New Zealand Limited. All rights reserved. 1

11 Section 1 Introduction decreasing during the late afternoon. However, PV generation output can be highly variable with changeable and fast moving cloud. The variability and intermittent effects of PV generation can cause operational issues for grid management. The increase in inverter-based generation in the power system (replacing conventional synchronous generators) can alter the dynamic behaviour of the power system. Inverters are highly programmable making their behaviour less predictable. Furthermore, PV generation will be more distributed compared to the present centralised generation topology. This form of distributed generation presents challenges in studying the dynamic behaviour and real-time operation of the power system. However, the studies are needed in order to understand the effect of PV generation variability and intermittency on the power system, and in forecasting the likely impact on the reliability of the ancillary services. 1.2 The PV Generation Investigation Project Transpower's PV Generation Investigation Project is part of the wider Emerging Energy Programme to ensure a smooth integration of these new technologies in New Zealand. The PV Generation Investigation Project provides studies into PV generation technologies and can be broadly separated into four main areas: generation dispatch, frequency management, transmission voltage management and transient stability. Effect of Solar PV on Frequency Management in New Zealand Transpower New Zealand Limited. All rights reserved. 2

12 Section 1 Introduction 1.3 The New Zealand power system Overview The New Zealand power system has several features which have the potential to impact the integration of distributed, non-dispatchable generation. The major factor is that ours is an isolated system with a high proportion of electricity generated from renewable sources which can vary in availability; namely hydro and wind generation. It is necessary to understand the impact to New Zealand's security of supply due to additional variable energy sources that are not highly correlated to either hydrology or wind resources. A significant increase in the share of PV generation in the generation mix may require changes to the existing transmission network equipment, operational processes, code and industry standards to: Secure adequate responsive generation (and possibly energy storage) capacity to manage the variable and intermittent nature of non-dispatchable PV generation. Introduce new equipment and operational measures to ensure adequate grid stability and control. Include distributed PV generation forecasting into scheduling processes. Ensure prices reflect economic costs. In reading the study reports produced for the PV Generation Investigation Project it is assumed the reader is familiar with the New Zealand power system, including the following key features: There is good generation mix with approximately 80% of electricity supply from variable renewable sources. There is existing thermal plant. There have been recent thermal plant retirements. There is existing wind penetration. It is a two island system; it is relatively small, with low inertia at times and large generating units present susceptibility to frequency disturbances. There is a mix of generation characteristics - fast ramping hydro, slower ramping thermal, constant geothermal, variable wind, etc. Transpower holds a classification of power system risks. Ancillary services are used to manage frequency - FIR, SIR, IL, FK, AUFLS (see the glossary for details of these) PV generation in New Zealand PV generation uptake in New Zealand has been relatively low. As of June 2017, New Zealand's installed PV generation capacity has grown to a total of 53.4 MW, with generation from residential, commercial and industrial sites [1]. This growth places total Effect of Solar PV on Frequency Management in New Zealand Transpower New Zealand Limited. All rights reserved. 3

13 Section 1 Introduction installed PV generation capacity at a level similar to the smaller, run-of-river hydro stations in New Zealand. However, at typical New Zealand solar capacity factors, this installed generation supplies only around 0.1% of total national energy consumption. This level of PV generation capacity has not compromised our ability to operate the power system securely and economically, with the existing tools and policies. However, the rate of growth is rapid, with a doubling time for installed PV generation capacity of approximately 18 months. PV generation installations are expected to continue to grow, as falling costs and an expanding market drive an increasing pace of PV generation uptake. Integration of high levels of PV generation into the power system will impact the frequency response to system imbalance, for the reasons outlined below: It is distributed and non-dispatchable, and therefore offsets load behind the GXP, with limited system operator visibility. It is highly stochastic, with rapid changes in output possible, depending on the relevant temporal and spatial scales, type of weather, season and level of uptake. The normal PV generation profile is negatively correlated with demand at times of maximum peak PV generation during mid-day, resulting in low system inertia that is susceptible to frequency disturbance. Inverter-based PV generation exhibits different frequency behaviour when subjected to system imbalance compared to conventional synchronous generation. 1.4 Impacts on frequency regulation Reliability of the power system depends on its ability to ride through a defined contingent (and credible) event and remain in a stable operating state. The power system is expected to provide continuous electricity supply with voltage and frequency within the statutory ranges. Roof-top PV generation represents another form of distributed and non-dispatchable generation. An increase in PV generation will displace grid connected synchronous generation, which will change the distribution of generation mix and alter the pattern of power flowing through the transmission network. This could result in significant changes to the way Transpower plans and operates the power system. As the New Zealand power system is relatively small in size and operates with no interconnection to other power systems, it is a relatively low inertia system susceptible to frequency disturbance events. Consequently, the power system is dynamic and constantly changing its active and reactive power demand. In addition, unplanned outages of equipment cause imbalances in the active and reactive power balance in the power system. These disturbances cause voltage and frequency to deviate from their nominal operating points. Effect of Solar PV on Frequency Management in New Zealand Transpower New Zealand Limited. All rights reserved. 4

14 Section 1 Introduction Historically, synchronous generators connected at the transmission level have provided ancillary/reliability services needed to securely operate the New Zealand power system. The ancillary/reliability services help maintain the balance to support reliable system operation and are an integral part of reliable power system operation. The services maintain system frequency by employing automatic functions which respond to loadgeneration imbalances. This study investigated the capability to manage frequency under high PV generation, where system inertia is low and has limited governor response units. There are three major effects on frequency regulation resulting from reduction of above mentioned elementary ancillary/reliability services: Reduction in the number of online synchronous generators reduces system inertia, resulting in a higher rate of change of frequency (RoCoF) during the credible loss of generation input (e.g., a contingent event). Reduction in the number of online synchronous generators reduces the ability of the system to provide governor response for frequency regulation. Variability of PV generation during cloudy days affects frequency management within the normal frequency band. An increase in PV generation penetration in the power system could compromise the ancillary/reliability services delivery and affect Transpower s ability to manage frequency to meet the PPOs. These changes in system dynamics will inevitably change the way the power system is planned, assets are utilised and the overall system is operated. Effect of Solar PV on Frequency Management in New Zealand Transpower New Zealand Limited. All rights reserved. 5

15 Section 2 Existing frequency management practices 2 EXISTING FREQUENCY MANAGEMENT PRACTICES The New Zealand power system encompasses two islands, connected by an HVDC link. The North Island power system serves an island maximum demand of 4500 MW, with an installed capacity of about 5,600 MW 1 from a mixture of fuel sources namely: hydro, wind, geothermal, gas and coal. The South Island serves a maximum demand of 2200 MW, with an installed capacity of about 3,700 MW of hydro and wind. Most of the time excess hydro generation from the South Island is exported through an HVDC link to meet load demand in the North Island. The New Zealand power system is an isolated and small power system making it very susceptible to frequency deviation, due to small imbalances in supply and demand caused by generator or load ramps or trips. Consequently, frequency management is a high priority in New Zealand to ensure the power system is operated securely and reliably. The Transpower system operator has an obligation to maintain frequency within bounds stipulated in Code 7.2, which stipulates that frequency is to be maintained: Within the normal band between 49.8 Hz and 50.2 Hz (both inclusive) arising from a supply and demand imbalance At or above 48 Hz for both islands during a contingent event At or above 47 Hz in North Island and 45 Hz in South Island during an extended contingent event In addition, the Code stipulates a generator must remain connected to support frequency when system frequency is within the statutory limits. The Transpower system operator manages the system frequency at or below 52 Hz in North Island and 55 Hz in South Island following a credible contingent event or an extended contingent event to prevent generator disconnected at high frequency. See Appendix A1 for more details on current frequency management practices, including contingent event classification and managing frequency stability. 1 Asset Capability Statement (ACS) as of December Effect of Solar PV on Frequency Management in New Zealand Transpower New Zealand Limited. All rights reserved. 6

16 Section 3 Frequency in the normal band study 3 FREQUENCY IN THE NORMAL BAND STUDY 3.1 Background Impact of PV generation on cloudy day PV generation output is impacted at various levels by atmospheric and physical obstructions that may complicate traditional power system operation measures. This is generally referred to as the variability of PV generation. The most common factor is sudden changes in cloud cover. Shadowing from cloud movement causes an unpredictable variability in the amount of solar irradiance reaching the solar panels, resulting in an equally unpredictable variability in the output from the PV system. Within the scope of Transpower s Emerging Energy programme this section investigates the impact of variable PV generation on managing frequency within the normal band. 3.2 Cloudy day study methodology approach Introduction This subsection describes the study methodology used to calculate the national cloudy day photovoltaic (PV) profile to investigate the impact on the PV generation when operating the New Zealand power system. Investigating the impacts of PV generation on power system operations requires accurate simulation of the high-resolution solar irradiance data. This is typically performed using weather stations and census data. The task is difficult because the weather station data requires a resolution that enables the study of both steady-state and dynamic responses of weather conditions on PV generation. The cloudy day method described here involves sourcing weather station data for each council region and statistically disaggregating this to the system grid exit points (GXP). The method considers geographical diversity, as the performance of individual PV generators is impacted differently at their associated GXPs due to variations in the ambient speed and location of the clouds; that is, consideration is given to wind speed together with potential sun energy (measure of solar irradiance) at specific geographical locations. In short, the method simulates the aggregated power system level smoothing effect caused by the passage of clouds across the country that leads to variations in individual PV generator outputs to their associated GXP. The national PV generation generated is used to assess the performance of the power system to maintain frequency within the normal band. This type of study involves performance running simulation for a long duration to generate enough data for statistical analysis. Traditionally, power system simulation software such as DSATool or PowerFactory is used to perform a dynamic study to assess power system frequency performance. However, these simulation tools are not suitable as they need significant Effect of Solar PV on Frequency Management in New Zealand Transpower New Zealand Limited. All rights reserved. 7

17 Section 3 Frequency in the normal band study model improvement. To use them in standard form would likely result in an accumulation of numerical errors, resulting in solution convergence issues. To undertake this study, a simplified linear power system model was developed to represent the New Zealand power system to undertake this study. The model allowed streams of PV generation data as input to estimate the system frequency response due to the variability of PV generation. Statistical analysis was then carried out to determine the frequency performance to stay within the +/- 0.2 Hz band Weather station data High-resolution weather station data is provided by the National Institute of Water and Atmospheric Research (NIWA) and provides solar irradiance data, as follows: Units of Watts per a meter squared (MMMM/mm 2 ) One minute resolution for one year (2012) 15 weather station locations shown in the table below There is one location for every council region Weather Station Locations Tauranga Airport Mahia Gisborne Airport Palmerston North Airport Kelburn Wellington Dunedin Airport Paeroa Woodburn Airport (Marlborough) Christchurch Airport New Plymouth Airport Invercargill Airport Auckland Airport Nelson Airport Hokitika Airpot Kaikohe Weather station data into GXP data Table 1: Weather station locations For a cloudy day scenario, it is highly likely that each PV generator is uniquely influenced, depending on location and speed of cloud cover. Therefore, to simulate the impact of cloud cover on PV generation, the method described here disaggregates the weather station data to the GXP level. Figure 1 shows weather stations and GXPs across the country. Effect of Solar PV on Frequency Management in New Zealand Transpower New Zealand Limited. All rights reserved. 8

18 Section 3 Frequency in the normal band study Figure 1: GXPs and Weather Stations To match each GXP to the nearest weather station the following distance formula is used. DDDDDDDDDDDDDDDD(xx) GGGGGG(xx),WWWWWWWW(xx) = (LLLLLLLLLLLLLLLL GGGGGG(xx) LLLLLLLLLLLLLLLL WWWWWWWW(xx) ) 2 + (LLLLLLLLLLLLLLLLLL GGGGGG(xx) LLLLLLLLLLLLLLLLLL WWWWWWWW(xx) ) 2 The calculated distance for every weather station and GXP in the country was computed and analysed; the weather stations having been assigned to GXPs by the least distance. Effect of Solar PV on Frequency Management in New Zealand Transpower New Zealand Limited. All rights reserved. 9

19 Section 3 Frequency in the normal band study This least distance approach is more geospatially accurate than that applied in method one of Transpower s PV Generation Investigation Project. This level of accuracy is required for the model two studies to simulate an accurate time response of PV generation due to cloud movement. Distance 1,2 Distance 2,3 Figure 2: Weather Station/GXP Assignment example The distance formula process used in this methodology does not take topography into account. This means there could be a mountain between a weather station and its assigned GXP. However, the method is simplified by not considering topography and no apparent concern was encountered with the results in the analysis of the final assignments Wavelet model The difficulties encountered in simulating cloud cover and the aggregated impact on correlated PV generation sites is not new to those investigating PV generation effects. Transpower selected an existing and proven wavelet variability model (WVM) provided by the University of California, San Diego, for development of cloudy day PV profiles for the New Zealand study. As shown in the following Figure 3, WVM applies a correlation scaling coefficient that adjusts the generation output at a weather station as a function of distance and time scale that varies by day and the geographic location of the associated GXP. [3] Figure 3: WVM applied to cloudy day PV Profile Effect of Solar PV on Frequency Management in New Zealand Transpower New Zealand Limited. All rights reserved. 10

20 Section 3 Frequency in the normal band study It is assumed that the PV plant: Has a constant wind speed of 10m/s for the initial investigation Is configured north facing at a 37-degree tilt Consists of discrete generating sites for simplicity The WVM takes the raw weather station data and the location of the associated GXPs to produce a geospatially adjusted time series for each GXP. The sum of the geospatially adjusted time series reflects a national solar irradiance profile which shows less variability than the raw data summed together GXP PV generation profiles As previously explained, the Wavelet Variability Model was used to de-correlate solar irradiance potential at each weather station down to the closest GXP and summed to make a national PV profile for this study. The estimate at each substation used the nearest irradiance sensor and took the number of dwellings and area served by the substation into account when producing the estimate. As the following analysis shows, the PV profiles are quite conservative in representing variability of PV generation at each GXP. Note in particular that different GXPs using the same irradiance data have very similar simulated PV generation. An example of this is shown in Figure 4 where: The first case shows the two substations are 100km apart and use the same Auckland Airport raw irradiance data. The second case also shows the two substations are around 100km apart, but uses irradiance data from different sensors at Woodburn (Nelson) and Kelburn (Wellington). The signals should be approximately equally de-correlated, but it can be seen from Figure 4a, the signals are significantly more correlated Wellsford Bombay Stoke (Nelson) Central Park (Wellington) Normalized PV Gen (1 pu) Normalized PV Gen (1 pu) Time (min) Time (min) Figure 4: Simulated PV generation for GXPs around 100km apart at (a) Auckland using the Auckland Airport irradiance data, and (b) Nelson and Wellington using the Woodburn and Kelburn irradiance data Summing the Nelson/Wellington profiles in Figure 4b and taking the minute-by-minute difference gives a standard deviation of 0.048pu, while the Auckland sites have a standard deviation of pu, which is 1.37 higher than the Wellington profiles. This Effect of Solar PV on Frequency Management in New Zealand Transpower New Zealand Limited. All rights reserved. 11

21 Section 3 Frequency in the normal band study would result in smoother PV generation profile if the sites were statistically identical but completely de-correlated. Thus, by using the data from 200 measurement points instead of 15, a reduction in the standard deviation of the order of about 3.6-fold could be expected. Therefore for the large number of PV sites that would occur in practice, the results would be significantly smoother. Put another way, the synthetic smoothing applied to each site is extremely conservative, resulting in a similar conservativism in the summed total PV generation signal. Therefore, statistical calculation methods would be preferable, since small scale simulation methods are dominated by the small scale variability, which is too conservative System wide PV generation profiles This study focused on analysing power system frequency. In short this means the GXP level PV generation profiles needed to be summed to make a national PV generation profile for this study. The magnitude of the national PV generation was scaled such that it would provide full generation for the max PV generation for the entire year. This is unrealistic, but the results can be scaled linearly to estimate the results for more realistic levels of PV penetration. This assumes a 4 GW PV total installed capacity in the system, a scenario which would generate 3000 MW to 3500 MW of PV generation (depending on the solar irradiance received by panels). Figure 5a plots the simulated total PV output around noon for three different days in summer. The PV generation for the selected three days is that which is most affected by the cloud NZ Load Simulated PV MW (3000MW) Time - Min Histogram (scaled to peak) MW Figure 5: (a) Total PV for three days around noon, scaled to a 3000 MW peak PV generation (b) Histogram of the change in total PV generation for each pair of consecutive minutes Figure 5b shows a histogram of the change in simulated PV output on a minute-by-minute basis for the same summer period. This is compared to the minute-by-minute variations of the current entire NZ load. The graph showed that with 4 GW PV installed in the system, the same magnitude of random load variability is generated in the power system as the current system. Effect of Solar PV on Frequency Management in New Zealand Transpower New Zealand Limited. All rights reserved. 12

22 Section 3 Frequency in the normal band study Data preparation If PV penetration is sufficiently high, the system operator would need to take the weather forecast and time into account when re-dispatching plant during the 5-minute dispatch interval. Sufficient ramping generation would be available to manage the solar ramping caused by the predictable movement of the sun. However, the system operator is not able to account for random variations caused by the movement of the cloud. The solar irradiance data accounted for variations caused by the movement of the sun and the cloud. A simple correction was used to remove the first-order variation caused by the predictable movement of the sun. The result is solar irradiance data containing only the effect of cloud movement. The process to remove the effect of sun movement from the raw solar irradiance data starts by assuming PV generation is proportional to the cosine of angle of the sun above the horizon. Where: Where: L is the latitude of -44 h is the hour angle δ is the declination of the sun given by N is the day number of the year. cccccccc = sin LL sin δδ + cos LL cos δδ cos h δδ = cos 360 (NN + 10) 365 Note that the angle calculation correction is for a flat horizontal panel which is appropriate for an irradiance recorder. PV panels generally have an approximate 40 degree north-facing angle to maximize energy collection. This should not have a significant bearing on the efficacy to dispatch enough ramping generation to correct dispatch error due to sun movement. In Figure 6a, total system PV generation for the entire year is normalised to the base PV generation of 3000 MW. Five days in February were chosen to illustrate the use of the simplistic method used to prepare the system PV generation for the studies. However, the entire year data was used in the study to assess the impact of PV generation variability from cloud movement. The three curves shown in Figure 6b are: Blue: the PV generation calculated from the RAW solar irradiance data containing the effect of sun and cloud movement Red: using the simplistic method to calculate the PV generation due solely to sun movement Black: the PV generation with the effect of cloud movement only. Effect of Solar PV on Frequency Management in New Zealand Transpower New Zealand Limited. All rights reserved. 13

23 Section 3 Frequency in the normal band study The studies were carried out by introducing a disturbance signal to the equivalent power system. The disturbance signals used were the PV generation calculated from the raw irradiance data and the irradiance data with the cloud movement effect only. Figure 6: (a) PV output in 1min increments for the year. (b) PV output noise after taking the sun position into account. Effect of Solar PV on Frequency Management in New Zealand Transpower New Zealand Limited. All rights reserved. 14

24 Section 3 Frequency in the normal band study 3.3 Study approach Introduction This subsection describes the model and analysis used to study the impacts of variable generation on managing frequency within the normal band of ±0.2 HZ nominal system frequency, as defined in the Code [4]. This type of study involves running a simulation for an entire year to assess the fluctuation of system frequency resulting from the continuous slow variation in the supply and demand. This slow perturbation is generally due to ramping of either load or generation. In this study, this perturbation was from PV generation variation from cloud moving over solar panels. Power system simulation software such as DSATool TM and PowerFactory could have been used for this type of study. However, components are currently modelled that respond within a period of one second to a few minutes. To run simulations with model components with slower response times, ie up to tens of minutes, significant model improvement is needed to capture the long term dynamic. The short term dynamic has little effect in a study of this nature. Furthermore, with this simulation, numerical errors can accumulate causing convergence issues, and computation time can be excessive. Instead, a simple model was created to represent the New Zealand power system frequency response behaviour. The model mimics the collective action of governors, frequency keeping and system inertia. This simplified model is adequate to assess the cloudy day PV generation variability effect and is more efficient to avoid excessive computation time and convergence issues. The model and its system parameters were verified using historical data in the spectral domain (Bode plot) prior to running the study. In subsection 3.3.3, the spectral domain technique shows that the model is not sensitive to changes in system parameters. To represent the current system behaviour, the study introduced a fictitious load that switches at four times per hour at a magnitude of 30 MW peak-to-peak. This forms a benchmark performance level for comparison with the PV generation cloudy day performance. A perturbation resulting from cloudy day PV generation was injected into the model to determine system frequency performance. The performance was compared to the benchmark scenario to determine the effect of the PV generation on managing frequency within the normal band Equivalent power system model For this study, a simplified linear model was created in Matlab/Simulink to represent the New Zealand power system. This model provides an acceptable representation of the frequency and control characteristics of the system, as shown in the following depiction of the power system (Figure 7). The simplified model represents the entire New Zealand Effect of Solar PV on Frequency Management in New Zealand Transpower New Zealand Limited. All rights reserved. 15

25 Section 3 Frequency in the normal band study power system, assuming that the HVDC link has full capability in regulating frequency in both islands. Figure 7: Simplified model of the power system The model has four major system wide control blocks that influence the performance of system frequency control. The four blocks are: System frequency keeper control: represents the combination of a traditional frequency keeper with a sufficient bandwidth (or re-dispatched sufficiently quickly not to hit its limits) and a re-dispatch action which simply resets the frequency keeper. System governor control: represents the governor droop present on all synchronous generators. System inertia: represents the dampening effect in change of system frequency available from all rotating machines on the system. Disturbance: represents the disturbance from the load or PV generation. Since this is a linear model, superposition of the disturbance can be applied. (The uncorrected and corrected PV generation data are used as the disturbances signals.) The PV generation signal was input as Disturbance to the simplified system model. Without any disturbance (Disturbance=0) in the simplified control system, the frequency deviation trends at 0 pu, which represents an equal balance of system generation and demand. Due to the dynamic nature of power system demand, the balance with generation output is continually moving. To keep the system inside the normal frequency band of ± 0.2 Hz nominal, machines on the system need to control their output with governor action accordingly. The system operator procures a single or group of generators known as frequency keepers. These keep a certain amount of generation available to be put on or taken off the system. In addition to the machine governor action, these actions help control the system frequency in the normal band. Effect of Solar PV on Frequency Management in New Zealand Transpower New Zealand Limited. All rights reserved. 16

26 Section 3 Frequency in the normal band study Model and parameter validation The model was then tested for sensitivity to changes in inertia, with the Frequency Keeper gain (FKGain), Governor Gain (GovGain) and Governor Time (GovTime) control parameters held constant. An inertia constant of 10s was used to represent the current system. The study assumed that grid connected synchronous generators would be displaced by PV generation, thus reducing system inertia and governor response capability. By reducing the inertia constant to 5s and the governor response reduced by half, this represented the power system with 3000 MW of PV generation. Figure 8 below shows the Bode plot of the simplified model with several different inertia parameters. The model has low sensitivity to changes in the inertia constant and GovGain, with the system behaviour remaining predictable while the disturbance time constant (shown in the x-axis) can change. By Parseval s theorem, the final result is the integration of the multiplication between the Bode plot and the frequency spectrum of the disturbance. This meant a large amount of work was not required in determining and tuning the exact inertia constant and GovGain, since it would make little difference to the final result. The studies were carried out with different values of inertia and GovGain to confirm the lack of sensitivity of these parameters upon the final result Inertia = 10, GovGain = Inertia = 15, GovGain = Inertia = 5, GovGain = Inertia = 5, GovGain = s 10s 1min 10min 1hr 6hr 1d Time Figure 8: Bode plot for a simple frequency model of the system To validate the Bode plot, two types of data were used. Using data at two timescales allowed the spectral domain to be explored more fully. The first set of data was a 2 second timestep frequency and New Zealand load data (from SCADA). This data was collected in weekly increments from September 2015 to Effect of Solar PV on Frequency Management in New Zealand Transpower New Zealand Limited. All rights reserved. 17

27 Section 3 Frequency in the normal band study September The 2 second frequency data from each week was Fourier transformed and used to calculate a Bode plot for each week - see Figure 9. The second set of data, created by using 1 minute data for an entire year, was then used to create a single Bode plot. Although there was a shift at the peak of the spectral, the shape of the Bode plot did not change significantly to affect the final results. The Bode plots show a peak in the frequency sensitivity to load variations of around the 1 minute mark. This is expected because for higher frequencies (shorter timescales), the inertia of the system stabilizes the system while for lower frequencies the control actions of the grid drive the frequency back to nominal wk 2 sec data Mean 1wk 2sec data 1 year 1 min data Model Inertia = 10, GovGain = Model Inertia = 15, GovGain = Model Inertia = 5, GovGain = Model Inertia = 5, GovGain = s 10s 1min 10min 1hr 6hr 1d 1wk 3mon Time Figure 9: Model vs calculated Bode plots compared Figure 9 shows that the selected equivalent power system model and values produce similar Bode plot spectra as the Bode plot obtained from historical data. Based on the validation study, the parameter setting of inertia = 10 and GovGain= were selected to represent the current system behaviour. With high PV generation penetration, most of the synchronous generators in the power system would be dispatched off. With that assumption, the inertia constant and the GovGain will drop. In one of the study cases, it was assumed that the inertia of the system dropped by half and that governor control ability was reduced by half. Effect of Solar PV on Frequency Management in New Zealand Transpower New Zealand Limited. All rights reserved. 18

28 Section 3 Frequency in the normal band study Study results Two sets of PV generation data were prepared and studied. The set of data based on the effect from cloud movement is termed corrected data, whereas data containing the effect from both sun and cloud movement is termed uncorrected data. The corrected and uncorrected PV signals were run through the model. Scenarios with differing loads were not considered as it would make little difference in the eventual ballpark estimate, as explained in subsection For comparison, a fictitious load of 0.01pu (e.g. 30 MW for a 3000 MW load) was also used as an alternative disturbance. This switches randomly around 4 times an hour, which is somewhat comparable to the system behavior experienced on the New Zealand power system. The frequency disturbance for PV generation scaled to 1 pu generation (equivalent to 3000 MW of PV generation) at peak summer was introduced into the simplified power system model to calculate the frequency deviation resulting from sun and cloud movement. The result was compared to that generated by the fictitious load introduced in the benchmark case. Assuming that PV generation has displaced grid connected synchronous generations, reducing the system inertia to half the benchmark case value, the New Zealand power system will experience a similar frequency performance to the benchmark fictitious load switching case, with a solar PV penetration of around 1748 MW. The frequency disturbance in this case is caused solely by the switching of 30 MW fictitious load at 4 times and hour. The degradation of frequency performance is mainly due to the reduction of system inertia, as grid connected synchronous generation is displaced by the PV generation. Cloud movement that intermittently shades solar panels can result in rapid variations in PV generation. In addition, sun movement will ramp up PV generation during the morning with decreasing generation in the afternoon. The combined effects will cause PV generation to vary considerably, resulting in fluctuations of system frequency. The study results showed that standard frequency deviation of mhz is experienced by the power system under these operating conditions. The frequency performance of 1050 MW of PV generation was comparable to the current system. If the slow ramping of PV generation is balanced by re-dispatch of generation, the PV generation penetration level could increase to 1260 MW and still maintain the same frequency performance as the benchmark system. In such case, the frequency deviation is caused solely by the movement of cloud. Figure 10 below shows the likely effect on frequency produced by various levels of PV load and cloud movement. Effect of Solar PV on Frequency Management in New Zealand Transpower New Zealand Limited. All rights reserved. 19

29 Section 3 Frequency in the normal band study Figure 10: PV generation penetration and disturbance scenario Thus, with 3000 MW of PV generation in the New Zealand power system it can result in a deterioration of frequency performance compared to the current system. However, the magnitude of the frequency deviation is small, in the region of +/- 32 mhz standard deviation. The ratio of the amount of governor control effort for the PV generation compared to the benchmark case was calculated to estimate the additional governor action required to maintain the same performance level as the benchmark cases. With 3000 MW PV generation penetration, the analysis shows that 1.5 to 2 times of additional governing action is required to maintain a comparable frequency performance standard to the current system Key findings The ratio of the frequency disturbance and governor effort between the current system and the 3000 MW PV generation scenario was estimated to be around 1.5 to 2. Discounting the effect of system inertia, the expected frequency disturbance of a 3000 MW PV generation penetration resulted in a frequency performance approximately 2 to 4 times worse than the current system. Alternatively, if about 30% of total generation were PV (estimated to be around 1000 MW) then expected frequency disturbance would likely be equal to that experienced currently on the system. Through the analysis of historical data, the standard deviation in system frequency noise is determined to be around Hz for a combined North and South Island power system. This frequency deviation is caused purely by changes in system load demand and generation. The frequency deviation determined in this study excluded this frequency noise. System frequency noise was added to the study results to assess the percentage of time that frequency will lie within the +/- 0.2 Hz normal band frequency limit. Effect of Solar PV on Frequency Management in New Zealand Transpower New Zealand Limited. All rights reserved. 20

30 Section 3 Frequency in the normal band study Computation of the probability distribution for system frequency deviation is shown in Figure 11 below. Figure 11: Probability distribution for system frequency deviation The study indicated that for % of the time, system frequency will deviate more than 0.2 Hz for the 3000 MW PV generation scenario. This equated to 33 minutes of the time (annually) that frequency exceeded the +/- 0.2 Hz limit. The probability of system frequency exceeding the +/- 0.2 Hz limit is, therefore small for a 3000 MW PV generation penetration scenario Recommendations The national PV generation scenario was derived from solar irradiance data from 15 weather stations. The data was de-correlated down to the closest GXP and summed to make a national PV generation profile for this study. The solar irradiance data from limited number weather stations made the solar generation profile at each GXP much more correlated than it should have been. In addition, the assumption of using constant wind speed adds a degree of conservatism to the estimation of the national PV generation. In reality, the wide spread of PV generation panels will significantly smooth out the effects of variable cloud movement at the national PV generation level. The results obtained from this study are therefore conservative and it is expected that the cloud movement effect will have little impact on the way we manage the frequency within the normal band. It is however prudent to re-check the study assumptions when national PV generation penetration reaches 1000 MW and when solar irradiance data from more weather stations is available to validate the study results. Effect of Solar PV on Frequency Management in New Zealand Transpower New Zealand Limited. All rights reserved. 21

31 Section 4 Frequency stability study 4 FREQUENCY STABILITY STUDY This section details frequency stability and associated impacts of high PV generation on the system operator's ability to continue to meet the current PPOs. The conventional power system uses synchronous generators that are designed to operate at nominal system frequency when generation is balanced with demand plus system losses. In an event where there is a sudden loss of generation or load, an imbalance in supply or demand (known as system imbalance) is created in the power system and system frequency deviates from nominal. The synchronous generators still connected to the system would conventionally increase or decrease supply accordingly (within their capability) to bring system frequency back to nominal. The capability of the power system to respond to changes in system frequency is referred to in this report as frequency stability and can be analysed in two time frames, as demonstrated in Figure 12 below, in which: tt aa to tt bb, is where generator inertia and load response are most dominant (further detailed in subsection 4.1). tt bb to tt cc, is when synchronous generators respond to stabilise the system frequency close to nominal, known as governor response (detailed further in subsection 4.1). Figure 12: Governor response, system inertia and system frequency in an under-frequency event To study system inertia (green area in Figure 12 above) this study used a RoCoF (Rate of Change of frequency) base measure to determine the level of impact that high PV penetration will have on the system. RoCoF is linearly related to the level of generation and demand imbalance on the system and inversely proportional to the level of system inertia. The following equation shows the relationship between system inertia and system imbalance (generation and demand plus losses) on the system. [5] RRRRRRRRRR HHHH/ss = dddd dddd = 1 2 ff SSSSSSSSSSSS IIIIIIIIIIIIIIIIII MMMM 2 0 SSSSSSSSSSSS IIIIIIIIIIIIII MMMM ss 2 This topic is further explained in the analysis section of this report. Effect of Solar PV on Frequency Management in New Zealand Transpower New Zealand Limited. All rights reserved. 22

32 Section 4 Frequency stability study As PV generation begins to offset large quantities of system demand, a reduced number of generators would be dispatched, causing system inertia and governor response to reduce and RoCoF to raise. The reduction in system inertia would cause frequency to fall faster in the green area of Figure 12 (above) in an under-frequency event and may push the system to a point where it cannot recover. A reduction in governor response on the system would cause system frequency (in the red area in Figure 12, above) to take longer to recover and come back to nominal in an under-frequency event. The reduction of inertia and governor response after a loss of substantial amounts of demand on the system would have an inverse system frequency response (overfrequency event) to that explained for the loss of generation (under-frequency event); essentially, lost demand and high PV penetration will increase the potential for overfrequency on the system. The reduction in inertia and governor response on the system would have a knock-on effect to the quantity of under-frequency and over-frequency reserve needed to manage the system within the Code limits following a sudden loss of generation or demand on the system. The following subsections will discuss the study methodologies for impacts of high PV penetration on system inertia and governor response in the New Zealand system. 4.1 Impact of PV generation on system inertia This subsection summarizes the methodology used in studying future impacts of PV generation on system inertia and the ability for the system operator to continue to meet the PPOs. System inertia is proportional to the sum of stored kinetic energy in synchronous machines (generators and motors) that is absorbed or released to arrest changes in system frequency on the power system. During high inertia scenarios (i.e. many synchronous machines), the power system experiences reduced frequency deviations compared with low inertia periods (few synchronous machines), given similar events (loss of supply or demand). The amount of inertia generators contributing to the system depends on the size and type of the generators, and is not directly coupled to how much power is being produced. In New Zealand, conventional synchronous generators have inertia constants in a range of 1.5 to 7 seconds. 3 The total system inertia constant, commonly referred to as H, is the sum of all inertia constants for machines connected to the power system and SS ni is the rated power of the generator as shown in the following equation: TTTTTTTTTT SSSSSSSSSSSS IIIIIIIIIIIIII = SS ni HH ii [MWs} In high PV generation scenarios, conventional synchronous generators would be displaced by PV generation, which does not provide inertia to the power system, effectively N i=1 3 The common inertia constants calculated using Asset Capability Statements (ACS) supplied by asset owners as required by the Code. Effect of Solar PV on Frequency Management in New Zealand Transpower New Zealand Limited. All rights reserved. 23

33 Section 4 Frequency stability study reducing total available system inertia. Power system frequency response becomes more sensitive to changes in system imbalance in low system inertia scenarios. Low system inertia causes the frequency to fall or rise more quickly following the loss of generation or load on the system. System Inertia is a critical factor in determining the level of mitigations used for frequency management to ensure the system remains within statutory limits; e.g., procuring ancillary services such as Over-Frequency Arming, Fast and slow Instantaneous reserve (IR) as well as mandated AUFLS arrangements. With high penetration of solar PV, system inertia can change significantly over a day, thus making the design of mitigations for frequency management more challenging. Poorly designed mitigations will reduce the discrimination ability of protection equipment during high RoCoF conditions, resulting in an increased shedding too much load or tripping too much generation. 4.2 PV generation response to frequency deviations High PV generation levels during sunny hours of the day will reduce the number of conventional synchronous generators to be deployed to meet lower net system demand. In a situation where the generation mix includes a high proportion of PV generation, the behaviour and characteristics of the PV generation is important when analysing the frequency performance of the power system caused by system imbalance. PV generation dynamic behaviour in responding to frequency deviation is thus critical to this study and must be modelled reasonably accurately to assess the impacts. Three commercial roof-top single phase solar PV inverters were tested to record the active power dynamic behaviour when subjected to frequency deviation. Three inverters from different vendors with different capacity ratings, operating characteristics and dynamic behaviour were subjected to various bench tests to understand the frequency response behaviour of the inverters. The test results indicated: Three forms of PV generation did not respond to an under-frequency event. Two forms of PV generation responded to an over-frequency event by reducing their electrical output. The response of the two forms of PV generation to over-frequency is compliant with the AS/NZS Computer models were built to represent the three forms of PV generation. The recorded information was used to validate the computer models to ensure they could predict the dynamic behaviour reasonably accurately during a frequency event. The responses of the inverters are in Figure 13 and Figure 14. Details of the PV generation characteristics and the model representing each type of PV generation are described in Appendix A2. Effect of Solar PV on Frequency Management in New Zealand Transpower New Zealand Limited. All rights reserved. 24

34 Section 4 Frequency stability study Figure 13: PV generator response to Under-Frequency Event Figure 14: A typical PV generation response to an over-frequency event For simplicity, the different forms of PV generation were modelled with an equal split for this study, meaning each type had an equal chance of being installed, thereby making each form of PV generation about 33% of the national PV penetration for this study 4. This assumption should not greatly affect the assessment on the effect of under-frequency. A sensitivity study was carried out to consider a scenario in which the over-frequency 4 More information on the investigation of PV generation inverters is in Appendix A of this report. Effect of Solar PV on Frequency Management in New Zealand Transpower New Zealand Limited. All rights reserved. 25

35 Section 4 Frequency stability study response of different forms of PV generation was reduced to 17% to illustrate the effect on under-frequency. 4.3 Study methodology This subsection covers the study scenarios and assumptions undertaken in the analysis of high PV penetration on frequency stability and its impacts on the ability of the system operator to continue to meet the PPOs. The general policies used by the system operator to meet the PPOs for frequency events, are classified by risk. 5 This study focused on identifying the impacts of high PV penetration on frequency stability in two types of power system events, a contingent event (CE) and extended contingent event (ECE) 6. Powertech s Transient Simulation Analysis Tool (TSAT) was used for this analysis. Mitigations currently used to ensure statutory limits and PPOs are maintained during frequency events on the system are shown in Figure Mitigations Over frequency reserves Frequency keeping, reserves and governor response Reserves and interruptible load AUFLS blocks This way to cascade failure and black start Over-frequency event Stay at or below 55 Hz in SI Stay at or below 52 Hz in NI Restore to 50 Hz Within 60s CE at or above 48 Hz Under-frequency event ECE above 47 Hz in NI ECE - Stay at or above 45 Hz in SI Normal band This way to cascade failure and black start System Frequency Figure 15: Frequency Management Barometer - Mitigations used to ensure statutory limits and PPOs are maintained during frequency events on the system The following subsection details how the study cases, frequency management mitigations and events were represented in this investigation Study scenario The study cases used were developed during the work on the first report produced for the PV Generation Investigation Project - Effect of Solar PV on Generation Dispatch in New Zealand and are summarised in the following subsection Twelve sample study days were selected to provide representation of: All four seasons 5 The policies used by Transpower system operator are set out at part of the Policy Statement. This can be found on the Electricity Authority webpage (under documents incorporated into the code by reference) at 6 See the Glossary Effect of Solar PV on Frequency Management in New Zealand Transpower New Zealand Limited. All rights reserved. 26

36 Section 4 Frequency stability study Examples of different types of daily profile (weekday, Saturday, Sunday) A wide range of demand profiles Different system conditions that can be expected on the power system The table below shows the selected study days for the sunny-day PV generation investigation. T Daily profile Summer Autumn Winter Spring Weekday 5 Jan Apr Jul Oct 2015 Saturday 9 Jan Apr Aug Oct 2015 a b l e Sunday 10 Jan Apr Aug Nov 2015 Table 2: Selected study days for sunny-day PV generation investigation Each study case considered all necessary input data to carry out a dynamic simulation. The power-flow cases were screened to identify the worst system condition for frequency management and the screened cases were analysed in detail (see subsection 4.5.1). The worst system condition was identified as the Sunday during summer when PV generation was at its maximum output and system demand lowest. This created a worst system scenario in which the least synchronous generation was dispatched Study generation assumptions Generation dispatch used in this study was produced using SPD and based on a wet year scenario. A dry year scenario is analysed in the initial screening but is not considered in this study. It was envisaged the risk quantity would be similar and the wet year scenario would reflect similar study conditions to assess the frequency management capability of the New Zealand power system. In addition, the wet year scenario will ensure that some hydro generating units in South Island will be dispatched to maintain steady state stability. The generation inputs are modified to represent a conservative situation, as outlined below. Hydro generation was used for base loading (i.e. dispatched first), leaving this technology less available to respond to marginal changes in demand. All existing hydro plant in both islands was considered available in the study, subject to transmission system capacity and outages. No plant was modelled operating in TWD 7 mode. Large thermal generation can have significant start-up and shutdown costs, and the amount of this generation modelled as available in the cases was checked to ensure 7 Hydro generators can be dispatched in Tail Water Depressed mode (TWD) to dispatch reactive power only with no active power output. Effect of Solar PV on Frequency Management in New Zealand Transpower New Zealand Limited. All rights reserved. 27

37 Section 4 Frequency stability study realistic outcomes 8. This reduced the amount of inertia and governor response available on the power system. IR was scheduled as per reserve requirements (this is summarised in the next subsection). Generators were modelled as dispatched in both islands to ensure system voltage and stability could be maintained. The lightly loaded power system during high PV generation required a minimum number of online generators to absorb excessive reactive power generated by the transmission network to maintain steady state stability Other study assumptions Other study assumptions based on power-flows during a high PV generation scenario are: The HVDC remains in operation, even at low transfer level, to offer frequency support in both islands. Some study cases have the HVDC operated in round power mode to offer full frequency sharing capability. Generator CE risk for the North and South Islands was set at 120 MW whereas the ECE risk was set at 240 MW for both islands. The thermal generators forced offline and the reserve requirement would likely reduce system CE risk (see the following subsection). HVDC did not pose an ECE risk as transfer would be low. PV generation would ride through all frequency and voltage events Instantaneous reserves As PV generation starts to offset large amounts of conventional synchronous generation and system inertia reduces, the instantaneous reserve (IR) requirements will change. This impact needed to be simulated appropriately for the study. With large thermal generation units being displaced by PV generation, the quantum of the system risk usually associated with these units is expected to reduce or be constrained to meet IR requirements. The scheduling of IR is not covered in this report Interruptible Load and AUFLS modelling When studying the impacts of frequency management, it is essential to ensure the mitigations used on the system are modelled correctly. In an under-frequency scenario where a CE has occurred on the system, market procured interruptible load (IL) and IR acts simultaneously to arrest the frequency fall. The IL providers are mandated by the Code to drop load within one second of the grid system frequency falling to or below 49.2 Hz and sustained for at least 60 secs. 10 Practically, the IL trip times vary by the 8 Study cases considered closed-cycle gas turbines, open-cycle gas turbines, coal-fired units and diesel units offered in the market. The Otahuhu B, Southdown and Huntly Rankine stations were not included in the study. 9 Readers can refer to report Effect of PV generation on generation dispatch in New Zealand Appendix A2.2 for technical details of the NFR estimation model used in this study. 10 The IL requirement is paraphrased from the Electricity Industry Participation Code as defined in the Schedule 1.1, Interpretations. Effect of Solar PV on Frequency Management in New Zealand Transpower New Zealand Limited. All rights reserved. 28

38 Section 4 Frequency stability study technology used to detect system frequency and the circuit breaker operation time. To standardise the modelling of IL in this study, it was assumed that all modelled IL was tripped at exactly 1 second after the system frequency fell below 49.2 Hz. An ECE event generally involves the disconnection of multiple generation sources from the power system, such as the HVDC bipole or generator bus section that resulted in the loss of multiple generators connected to that bus section. As defined in the Policy Statement, the ECE risk can be mitigated by utilising IL, IR and in addition, AUFLS. An Automatic Under-Frequency Load Shedding (AUFLS) scheme has been designed in the model to disconnect 32% pre event demand as mandated in the Code and detailed below. It should be noted that there are different AUFLS tripping requirements in place for the South and North Island but both aim to automatically disconnect to recover from large frequency deviations caused by an ECE event. The allocation of load to be shed in different AUFLS stages are presented in Table 3 North Island South Island Block Trip setting ( Hz) Time delay (s) Demand (%) Trip setting ( Hz) Time delay (s) Demand (%) * * (ii) * Tiwai 3(ii) * (ii) (iii) RoCoF >1.2 Hz/s N/A 6 Table 3: Allocation of load to be shed in different AUFLS stages * Backup AUFLS stage to shed additional load to recover the system frequency after a time delay of 14 seconds Over-Frequency Arming modelling Over-frequency reserve, also known as Over-Frequency Arming (OFA) is provided by generating units as frequency management mitigation, when required, to automatically disconnect from the power system in the event of a sudden rise in system frequency. Over-frequency arming can be required due to an unplanned loss of a large industrial load or the tripping of the HVDC link. Certain generating units may also trip during an overfrequency event if their protection limits are violated. Effect of Solar PV on Frequency Management in New Zealand Transpower New Zealand Limited. All rights reserved. 29

39 Section 4 Frequency stability study Over-frequency reserves are procured under ancillary service contracts to be deployed as required by the Code and associated terms and conditions. Transpower system operator uses an automated real-time system called the Over-Frequency Arming System (OFAS) to determine whether arming of generators is required, (and if so, how much arming is required), due to the current HVDC transfer levels and grid conditions. The study cases have the OFAS modelled to simulate the OFA amounts required for investigation. 4.4 Effects on System Inertia The effect on system inertia is investigated in this subsection. As PV generation in New Zealand increases it will significantly reduce system inertia. System inertia is proportional to the sum of stored kinetic energy in synchronous machines (generators and motors) that is absorbed or released, responding to the changes in system frequency on the power system. During high system inertia scenarios (i.e. many conventional synchronous generators) the power system experiences low RoCoF when compared to low inertia periods given similar system imbalance. In this study, generation and load rejection disturbances were introduced to demonstrate a relationship between PV generation and system inertia by using a RoCoF comparison. Rejection tests of 50 MW, 100 MW and 130 MW were performed with available HVDC response. The HVDC fast response to frequency deviation is akin to the inertia response of the synchronous generator. This relationship and its impact on system inertia was determined with a similar set of studies carried out with HVDC response disabled in the model to determine the impact of HVDC on system inertia by comparing RoCoF calculations. The following subsection describes the detailed results and analysis of these studies Study results The study results show that the system inertia of both islands reduces quite rapidly as PV generation increases, see Figure 16 below. The North Island has greater system inertia than the South Island. The important note here is that the system inertia is trending lower with high PV generation. Effect of Solar PV on Frequency Management in New Zealand Transpower New Zealand Limited. All rights reserved. 30

40 Section 4 Frequency stability study Figure 16: System inertia plotted against PV generation for North Island and South Island The HVDC link is designed to share frequency reserve to help maintain frequency in both islands by regulating the transfer of power between the two islands. As mentioned above, the fast response of HVDC is comparable to the inertia response of the synchronous generator which helps to reduce the RoCoF during a frequency event. To assess the change in system inertia when HVDC is enabled versus disabled, system imbalance disturbances of 50 MW, 100 MW and the maximum of 130 MW were introduced in both study scenarios. RoCoF at 1 second after each study case at different levels of PV generation was recorded and analysed, as shown in Figure 17 and Figure RoCoF (Hz/s) CE Risk (MW), HVDC Status 50,enabled 50,disabled 100,enabled 100,disabled 130,enabled 130,disabled NI PV generation (MW) Figure 17: RoCoF for 1 second after various system imbalances, at different levels of PV generation, with and without HVDC enabled (North Island) Effect of Solar PV on Frequency Management in New Zealand Transpower New Zealand Limited. All rights reserved. 31

41 Section 4 Frequency stability study Figure 18: RoCoF for 1 second after various system imbalances, at different levels of PV generation, with and without HVDC enabled (South Island) The figures show that the availability of HVDC to support system frequency had significant impact on the system inertia in the affected island. The availability of additional system inertia contributed by the HVDC link resulted in lower RoCoF for all test cases. The next subsection provides analysis of the impacts on frequency management due to high PV generation which leads to a reduction in system inertia Analysis The time domain frequency plot in Figure 19 shows that the system frequency decreases at a faster rate when the system inertia is reduced. As discussed in the previous subsection the decrease in system inertia is caused solely by an increase in PV generation displacing the grid connected synchronous generators. As the RoCoF increases due to a reduction in inertia on the system, the minimum frequency (Nadir dot in Figure 19) is reached much quicker. The rapid reaction of high RoCoF does not allow time for governor response and other mitigation measures such as IL and AUFLS to act to correct the system imbalance. This phenomenon will be analysed in more detail in the underfrequency management subsection 4.5. Effect of Solar PV on Frequency Management in New Zealand Transpower New Zealand Limited. All rights reserved. 32

42 Section 4 Frequency stability study Figure 19: Frequency response with decreasing system inertia The HVDC contribution is significant in reducing the RoCoF, provided there is time for governor, IL, and AUFLS to act to restore the system imbalance. The sensitivity studies for each island were conducted at different levels of PV generation and subjected to several levels of system imbalance disturbances with HVDC enabled and disabled. For each of the cases, the rate of frequency fall after the disturbance event was initiated was calculated as a RoCoF to compare the contribution of the HVDC and the result for the North Island study and South Island are shown in Figure 20 and Figure 21 below. Figure 20: The response comparison of system frequency after a North Island CE when HVDC is not available Effect of Solar PV on Frequency Management in New Zealand Transpower New Zealand Limited. All rights reserved. 33

43 Section 4 Frequency stability study Figure 21: The response comparison of system frequency after a South Island CE when HVDC is not available The RoCoF is significantly higher in all cases where the HVDC response is disabled. The average inertial constant contributed by the HVDC was approximately 0.6 seconds for both the South Island and North Island sensitivity studies. The study confirms that an increase in PV generation leads to a reduced level of system inertia, when less synchronous generators are operating on the system during sunlight hours. The RoCoF for the 3250 MW PV generation case is below 0.9 Hz/s without any contribution from the HVDC. It reduced to about 0.7 Hz/s when the HVDC was operating and contributing to the inertia response. The effect of high RoCoF on frequency mitigation measures will be discussed further in the later sections of the report. 4.5 Effects on Under-Frequency management Managing frequency in a low inertia system is a challenging task given that the system frequency will deviate faster and require more effort from mitigation measures to restore it. IR, IL and AUFLS are scheduled to mitigate various identified frequency risks to ensure that Transpower system operator can maintain the system frequency within the statutory limits. A time domain simulation study was performed to analyse the interaction of system inertia and the effectiveness of the frequency mitigation measures to identify any operational limitations in low system inertia conditions. Simulations were done for varying levels of PV generation to investigate the effect of a CE and ECE event. The main assumption in this study is that PV generation will ride through the frequency event without posing a secondary risk Study results Thermal generators are the single largest generating units in the New Zealand power system, and therefore likely to set the CE risk when operating. Since thermal generators have a relatively higher operating cost compared to other conventional generation types, Effect of Solar PV on Frequency Management in New Zealand Transpower New Zealand Limited. All rights reserved. 34

44 Section 4 Frequency stability study this study assumes thermal generators are dispatched off first. This is especially true during the wet year scenario, which is what the generation dispatch by the SPD is based on. In the study on the Effect of PV generation on Generation Dispatch in New Zealand, a modified market tool was used to determine the optimal generation mix in a high PV generation study cases. An example of the generation mix for Sunday in summer is shown in Figure 22 below. Figure 22: generation mix for Sunday in summer As the sun moves to its highest point in the sky, PV generation is operating at its peak and thermal generation is significantly reduced (dark blue bars in the chart) in the market mix. Hydro (yellow bars) is the base generation. These results confirm that thermal generation will be dispatched off first and make up a small amount of the generation mix in a high PV generation scenario. Dispatching off thermal generation during high PV generation periods will significantly reduce the system risk size, hence the amount of reserves required to cover the largest CE risk. The North Island and South Island CE risk are identified to be around 120 MW. Conventional synchronous generators provide inertia as well as IR to manage frequency in the power system. As synchronous generators are being replaced during a high PV generation scenario, the reduction in the amount of IR available to Transpower system operator to manage system frequency may lead to other market impacts. The usual risk setting component such as the large thermal generator and HVDC may have to constrain down the electrical output or transfer due to insufficient IR to maintain frequency obligations on the power system. This supports the assumption to use a lower CE and ECE risk for the study. Effect of Solar PV on Frequency Management in New Zealand Transpower New Zealand Limited. All rights reserved. 35

45 Section 4 Frequency stability study The following two graphs in Figure 23 and Figure 24 provide summary results of the under-frequency management study. As solar PV penetration increases on the system, and subjected to a CE or ECE event, the system frequency behaves consistently with the system inertia analysis. When subjected to a CE event, the Nadir reduced as the PV generation increased, but there was adequate IR to maintain the system frequency above 48 Hz following the event. The frequency responses of both islands are characterised by the same trend. Detailed analysis is carried out in subsection Interestingly, when the 3250 MW PV generation case was subjected to an ECE event, the North Island experienced a high over-frequency, in which system frequency registered close to 53 Hz. The effect was the result of shedding too much load in the AUFLS to correct the preceding under-frequency event. The impacts of ECE on North Island frequency management are covered in subsection When analysing the ECE study results in the South Island, the lowest frequency reached is lower than that in the North Island. The study indicated a Nadir close to 47 Hz when the South Island power system was subjected to an ECE event. One of the inverters exhibited a disconnection frequency of 47 Hz with a time delay of two seconds. This shows that there is a risk that inverters may disconnect when the South Island frequency falls below this level for more than two seconds. This risk is analysed in detailed in subsection The South Island studies revealed similar results to the North Island at various levels of PV penetration, as shown in the figures below. Figure 23: North Island System responds under a CE and ECE at different PV penetration levels. Effect of Solar PV on Frequency Management in New Zealand Transpower New Zealand Limited. All rights reserved. 36

46 Section 4 Frequency stability study Figure 24: South Island System responds under a CE and ECE at different PV penetration levels Analysis The analysis of high PV penetration and its impacts on Transpower system operator to manage under-frequency events is broken down into: North and South Island CE risk North Island ECE Risk South Island ECE Risk Effect of RoCoF on AUFLS North and South Island CE risk To simulate the most severe impacts of high PV generation when subjected to a CE, the summer midday study case was chosen for this investigation. Summer is when demand is expected to be lowest at midday and when PV generation is at its highest output, ensuring the least amount of conventional generation will be dispatched. In the summer study case, the highest system risk was identified as 120 MW in both Islands and rejected as a CE in the study scenarios to simulate an under-frequency event. The rejection of this risk was applied to the PV generation scenarios ranging from 600 MW to 3000 MW, as shown for the North Island in Figure 25 below. The highest system CE risk was identified as 120 MW in both Islands. The CE risk was applied to the PV generation scenarios ranging from 630 MW to 3250 MW. Figure 25 shows the frequency traces for North Island. Effect of Solar PV on Frequency Management in New Zealand Transpower New Zealand Limited. All rights reserved. 37

47 Section 4 Frequency stability study Figure 25: North Island contingent event frequency traces The HVDC bipole is operated in these scenarios to provide sharing of IR between the two islands; this assumption is the most likely scenario with high PV generation on the system. Sensitivity to the operation of HVDC and frequency management is discussed in subsection 4.4. In the North Island, with 630 MW of national PV generation (yellow curve above), the system is subjected to a 120 MW generator rejection that did not trigger IL to arrest the frequency fall. This outcome was due to two effects: A lower PV penetration equating to relatively higher system inertia when compared to other study scenarios More IR provided by the synchronous generators that can react to frequency disturbance before the frequency falls below the IL frequency threshold; that is 49.2 Hz This combination of the system effects resulted in system frequency falling at a slow enough rate to allow the governor to respond and recover the frequency before the IL threshold was reached. As PV generation increases (green and red curves above), more synchronous generators are constrained off the grid and in turn reducing the system inertia. The lower amount of system inertia causes system frequency to fall faster and further, making it harder for governor response to arrest the fall. The frequency falls to the point where IL is triggered. After IL triggers (at 1 second), the frequency recovers sharply. The simulation reconfirms IL as being an effective measure for restoring the balance of supply versus demand and ultimately returning frequency to nominal. It is evident from the simulation results that with large amounts of PV generation, frequency will fall faster, leading to the frequency nadir being reached faster. A similar analysis was conducted on the South Island. With lower system inertia in the South Island, the frequency fall is faster compared to the North Island in all cases, as shown in Figure 26 below. Effect of Solar PV on Frequency Management in New Zealand Transpower New Zealand Limited. All rights reserved. 38

48 Section 4 Frequency stability study Figure 26: South Island contingent event frequency traces In the South Island, there are currently very small amounts of IL available and these are assumed to be zero in this study. This means only IL in the North Island is available to arrest the frequency fall before AUFLS is needed. All of the studied PV penetration levels show that the frequency falls below the IL trigger value (49.2 HZ) in the South Island. In reality the frequency in the North Island (dotted brown line above), which is slightly different than the South Island, fell below the IL trigger value for only the 2500 MW and 3250 MW scenarios. With 630 MW of PV penetration, the governor action recovered the system frequency with a Nadir above the IL frequency threshold in the North Island; and no IL was lost on the system. The difference in frequency between the two islands during a frequency event is due to the frequency keeping modulation of the HVDC link. With the current frequency management strategy, the schedule of IR is adequate to mitigate a CE risk in both islands to manage post contingency system frequency above the statutory limit of 48 Hz for PV generation of around 3000 MW. The performance of the IL and IR need to be reviewed when system inertia is lower enough to cause a RoCoF close to 1 Hz/s to ensure its effectiveness North Island ECE risk The HVDC transfer between the islands reduced significantly during a high PV generation scenario. As the demand is supplied by the PV generation locally, the reliance of South Island hydro generation to supply the North Island demand reduced. In addition, the North Island power system is required to maintain a minimum number of synchronous generators in the system manage system voltage and to maintain system stability. A combination of these factors will significantly reduce the inter-island HVDC transfer. For all cases produced by the SPD, the inter-island HVDC transfer during high PV generation scenario did not exceed 200 MW. The reduction in HVDC flow equates to a reduced level of HVDC ECE risk. With the larger thermal units dispatched off, the study results show it to be unfeasible to identify a credible ECE risk in North Island that can cause the system frequency to drop below the 48 Hz code limit. As such, a hypothetical ECE contingency was created, where Effect of Solar PV on Frequency Management in New Zealand Transpower New Zealand Limited. All rights reserved. 39

49 Section 4 Frequency stability study two generators with a combined generation loss of 240 MW in North Island was used to study the impact on ECE frequency management during a high PV generation scenario. In the South Island, a disconnection of a Manapouri bus section can potentially cause two or more Manapouri units to disconnect from the system. In this study, two Manapouri units were assumed to be connected to the same bus section, allowing the loss of this bus to be used as a credible ECE event to assess frequency management in the South Island. In the North Island, 240 MW of generation was disconnected from the system with 630 MW, 2500 MW and 3250 MW of PV generation levels. The results of these three study scenarios are shown on Figure 27 below. Figure 27: North Island under-frequency response after an ECE of 240 MW The IL and IR scheduled in the system is sufficient to recover system frequency above 48 Hz for the PV generation of 630 MW and 2500 MW scenario. The lower RoCoF is likely to be the contributing effect allowing enough time for IL to be shed reducing the system imbalance. The governor then responded to a smaller imbalance to recover the system frequency at a higher Nadir. The IL disconnects at a lower frequency for a higher RoCoF case presented by the 3250 MW PV generation scenario, assuming the IL is configured to shed its load one second after the frequency dropped below the threshold of 49.2 Hz. With a simple calculation, the IL disconnected its load at Hz when the frequency RoCoF was Hz/s as observed by the 3250 MW PV generation scenario. In the red curve above shows the RoCoF changed at around 48.4 Hz, highlighted by the red dot. There was not enough governor action to arrest the continued frequency fall to below the AUFLS first stage frequency threshold of Hz. After the expiry of the intentional time delay, the AUFLS first stage operated, causing the frequency to upturn sharply. The study focussed on the future operation scenario. Thus, assuming that the Extended Reserve 11 initiatives are defined in the Code, they will be fully implemented as scheduled 11 The extended reserve initiative is defined in the Electricity Industry Participation Code in Part 8 subpart 5 - Extended reserve. Effect of Solar PV on Frequency Management in New Zealand Transpower New Zealand Limited. All rights reserved. 40

50 Section 4 Frequency stability study for the end of The AUFLS scheme is the only form of extended reserve product that is currently identified to prevent cascade failure following an ECE event. The first stage of AUFLS under the Extended Reserve initiatives is required by the Code to disconnect approximately 10% of North Island system load. The sharp upturn of frequency resulting in an overshoot to a maximum frequency of 53 Hz is a clear indication that excessive load shedding occurred, creating a positive system supply/demand imbalance. This is partially due to low inertia system as well. The activation of AUFLS first stage is a correct and expected action resulting from an ECE event. However, over shedding of load by the mitigation measures can be catastrophic, as over-frequency can cause an uncontrolled tripping of generators, causing a subsequent second negative system supply/demand imbalance. The second frequency downturn will have a higher RoCoF, potentially unable to be arrested by the remaining AUFLS resulting in system collapse. In reality, OFA will be activated to prevent the system frequency from rising above the 52 Hz limit. The above simulation did not include the OFA model to illustrate the effect of poorly coordinated mitigation measures. The above simulation results illustrated that the design of AUFLS is critical to the power system with low system inertia. The fast falling rate of system frequency (high RoCoF), combined with limited governor response can jeopardise the effectiveness of AUFLS by over shedding of load and causing a subsequent over-frequency event. The matching of an ECE risk to the AUFLS load percentage allocation is important making the subsequent overshoot in frequency respond to AUFLS. The mitigation measures have to work in both extreme trough demand periods when PV generation is at its peak, and periods of high maximum demand in the evening when there is little PV generation South Island ECE risk The frequency traces showing the responses of an ECE event with different PV generation are shown in Figure 28. An ECE event did not cause the South Island frequency to fall below 48 Hz to activate the AUFLS. When PV generation exceeded 2500 MW level, an ECE event causes the frequency to fall below the first stage threshold, triggering load shedding in South Island to recover the frequency. The first stage of the South Island AUFLS sheds approximately 16% of South Island load. Effect of Solar PV on Frequency Management in New Zealand Transpower New Zealand Limited. All rights reserved. 41

51 Section 4 Frequency stability study Figure 28: South Island under-frequency response after an ECE of 240 MW The ECE event did not lead to an over-frequency event because the amount of generation lost on the system matched reasonably well with the amount of AUFLS that was shed. The South Island demand is lower, hence the 16% of system load allocated to the first stage of South Island AUFLS represents a smaller quantum than the 10% allocation in North island. A point to note here is that the South Island Nadir is very close to 47 Hz. The type C inverter exhibited a minimum disconnection frequency threshold of 47 Hz with a two second time delay. The feature is not compliant with As/NZS If an ECE event in South Island causes the frequency to drop below 47 Hz, it will be a contest between the type C inverters being disconnected and the AUFLS second stage to recover the frequency above 47 Hz. The South Island AUFLS second stage is configured to disconnect load at 46.5 Hz with maximum operation time of 0.4 second. The trigger of the second stage of AUFLS is likely to arrest frequency fall and restore the system back to nominal should the inverters stay on the system. In the scenario where the AUFLS second stage does not recover the frequency above 47 Hz within two second, the type C inverter will drop off the system inevitably collapsing the system. The studies showed that in the current frequency management strategy, the combination of IR and AUFLS is adequate to mitigate ECE risk posed to the power system in a high PV generation scenario. It is of utmost importance to have all the inverters compliant with the AS/NZS standard so as not to pose a secondary risk to the system operation for not riding through a system fault AUFLS configuration In this study it is assumed that all the PV generation is the roof-top type that is embedded within the distribution network. It is likely that both the loads and solar PV inverters will be disconnected by the under-frequency relay if they are connected to the same zone substation feeders. The net effect is that there will be less AUFLS (demand) available to Transpower system operator to manage an ECE risk. The simplified diagram in Figure 29 Effect of Solar PV on Frequency Management in New Zealand Transpower New Zealand Limited. All rights reserved. 42

52 Section 4 Frequency stability study shows the likely configuration of AUFLS with embedded generation within the Low voltage network. G The Grid LV network M Under-Frequency relay system Figure 29: Likely configuration of AUFLS with embedded generation within the Low voltage network In the real life scenario, at 2250 MW of PV penetration, an AUFLS feeder tripping will result in a certain percentage of PV generation tripping as well (or generating in an islanded mode). Therefore, the net system would see a smaller amount of load shed on the system. This raises the question around whether there will be enough load available on the system to be shed to arrest system frequency after an ECE event. To investigate this question, the AUFLS modelled in the study cases were adjusted so that approximately 4% of North Island load is shed in each AUFLS block. Figure 30 illustrates the effect of reduced AUFLS in North Island for the 3250 MW PV generation scenario in managing a 240 MW ECE risk. Figure 30: effect of reduced AUFLS in North Island for the 3250 MW PV generation scenario in managing a 240 MW ECE risk The example shows that AUFLS allocation percentage has a profound effect on the postevent system frequency response. In this case, it is a positive effect preventing the system frequency from rising above the 52 Hz limit. It is imperative that the configuration of the AUFLS should be clarified and agreed so that a right quantum of AUFLS load can be assigned to each stage to study and optimise, thus reducing the chance of undesirable effects caused by under shedding or over shedding of AUFLS load. The challenge is that in the future, AUFLS will need to operate in an extreme trough demand period where PV generation is in abundance and also in an extreme Effect of Solar PV on Frequency Management in New Zealand Transpower New Zealand Limited. All rights reserved. 43

53 Section 4 Frequency stability study maximum demand period in the evening when there is no PV generation embedded in the LV network Managing post event over-frequency by PV generation over-frequency response It has been identified that PV generation has the capability to decrease its active power to support over-frequency management. This means that during an over-frequency event PV generation will help mitigate the level over-frequency reached if load is over-shedding. The green curve in Figure 31 below shows the response to system frequency when the over-frequency ramp down function of PV generation is enabled. Figure 31: North Island under-frequency response with PV generation over-frequency support. With 3250 MW of PV penetration on the system, the PV generation s capability to support the system in an over-frequency event helped to manage system frequency such that it remained below the 52 Hz code requirement. The ability of PV generation to support the system in an over-frequency scenario required by the AS/NZS may help mitigate the effects of over-shedding of load. The effect of PV generation to manage over-frequency in an ECE event in the South Island is detailed in the subsection Effect of RoCoF on AUFLS AUFLS is the last defence against extreme events such as the disconnection of multiple generators or an HVDC bipole. The AUFLS is a protection system that detects underfrequency conditions and disconnects load to maintain the supply-demand balance. An intentional time delay is incorporated into the under-frequency relay to eliminate erroneous operations due to spurious frequency signals and to provide discrimination between different stages in the AUFLS. An intentional time delay of 0.2 second is incorporated in the AUFLS stage in both North Island and South Island AUFLS schemes. In addition, the distribution circuit breaker takes a finite time to operate. It is reasonably conservative to assume that the breaker will start to operate to disconnect load after 0.2 second. In essence, the AUFLS will only start to shed load after 0.4 second (intentional Effect of Solar PV on Frequency Management in New Zealand Transpower New Zealand Limited. All rights reserved. 44

54 Section 4 Frequency stability study time delay + breaker operating time). The minimum time to disconnect AUFLS load is 0.4 second, including the intentional time delay. For a RoCoF of 1 Hz/s scenario, the frequency will drop from 47.9 Hz (first stage in North Island) to 47.7 Hz (second stage in North Island) in 0.2 second. By the time the second stage AUFLS is activated, the first stage AUFLS is just beginning to shed load in an effort to arrest the frequency. The first stage AUFLS will have to shed enough load within the next 0.1 second to recover the frequency above 47.7 Hz in order to prevent second AUFLS stage to shed load. This is illustrated by the solid blue curve in the diagram below. If the frequency does not recover above 47.7 Hz, the second stage AUFLS will start sending trip signal to the breaker to shed load. At this point, the second stage AUFLS will shed its load even if the frequency recovers above its threshold setting. This is illustrated by the dotted blue curve in the diagram below. As RoCoF increases, the North Island AUFLS will start losing the discrimination between each stage. At higher RoCoF, the later stage AUFLS will start sending a trip signal to the breaker even before the early stage can disconnect load to arrest the frequency. At RoCoF higher than 1.2 Hz/s, the df/dt stage will shed load at 48.5 Hz without any intentional time delay in an effort to reduce the RoCoF to allow the following AUFLS stages to response adequately to recover the frequency. Figure 32: AUFLS operation timing diagram This simple analysis is based on AUFLS operating within the desired timeframe. When RoCoF is increased (frequency falls faster), if AUFLS does not operate as expected, over shedding (lose discrimination) will cause over-frequency, whereas under shedding (AUFLS does not have time to act) will violate the lower frequency limit of 47 Hz. Effect of Solar PV on Frequency Management in New Zealand Transpower New Zealand Limited. All rights reserved. 45

55 Section 4 Frequency stability study 4.6 Effects on Over-Frequency management The reduced number of synchronous generators in the power system in the high PV generation scenario will effectively reduce the ability to manage over-frequency when a large quantum of load is disconnected from the power system. Transpower system operator is required to keep the frequency at or below 55 Hz in the South Island and at or below 52 Hz in the North Island, and relies on governor response and over-frequency arming generator to arrest frequency rise. The HVDC posed an over-frequency risk at the sending island. High HVDC transfer will cause frequency to rise in the sending end island following a bipole HVDC trip. As the HVDC transfer is low under the high PV generation scenarios (subsection 4.6.1), this risk is considered as not credible and is not considered in this study. In order to study the impacts of high PV penetration on over-frequency management, 400 MW load rejection simulations were conducted for varying levels of PV generation to investigate: The effect of RoCoF on the present Over-Frequency Arming scheme The effect of RoCoF for the different inverter types The results and analysis of these studies are detailed in following subsections. In the South Island, a 400 MW demand disconnection is comparable to two Tiwai potline trips. It is considered as a worst contingency to test the system with high PV generation. In the North Island, the electrical power transfer across the power system is low, implying that it is very unlikely a large quantum of load will be disconnected by a single contingent event. A hypothetical contingent event (multiple contingency) was created to disconnect 400 MW of North Island load to study the North Island over-frequency performance Study results Transpower system operator relies on governor response and over-frequency arming (OFA) as mitigation to over-frequency events. The reduction in governor response indicates the use of OFA in high PV penetration scenarios may be relied upon by Transpower system operator to a greater extent, to arrest a frequency rise following a large quantum of demand suddenly disconnected from the system. Two types of PV generation have the capability to reduce their outputs during an over-frequency event to help to arrest the frequency rise. These studies focus on the impacts of PV generation on the present OFA mitigation measures and over-frequency response of the solar PV generator to support the system in over-frequency scenarios. The North Island power system was subjected to a loss of 400 MW of demand for various levels of PV generation in our simulations. As the PV generation increases, the RoCoF observed in North Island and South Island increases. This observation is consistent with the analysis carried out in subsection 4.4. The maximum frequency observed in North Effect of Solar PV on Frequency Management in New Zealand Transpower New Zealand Limited. All rights reserved. 46

56 Section 4 Frequency stability study Island increased, which was partly due to the OFA generator being dispatched off in the high PV generation scenario. The high RoCoF can also contribute to this behaviour. The South Island responses showed similar trends but the maximum frequency observed was about the same for all three generation scenarios. The South Island has enough OFA generators available for all three generation scenarios. The study results are summarised in Figure 33. Figure 33 The South Island response with no PV generation support, at various PV penetration levels to a 400 MW demand trip. This study used the frequency performance test results of three different types of PV inverters as detailed in Appendix A2. The assumption for this study was that the three types of tested inverters were expected to be adopted with an equal probability (33% each) on the New Zealand system. Out of the three PV generators that were tested only two met the standard to support the power system in an over-frequency events, termed over-frequency response in this report. In order to test the sensitivity to PV generation Effect of Solar PV on Frequency Management in New Zealand Transpower New Zealand Limited. All rights reserved. 47

57 Section 4 Frequency stability study frequency response on managing the power system in high PV penetration scenarios, the 3000 MW PV penetration study cases were rerun with several adoption levels. The level of adoption for PV generators with over-frequency response was reduced from 33% to approximately 17% in the 3000 MW study cases, and the response in the North Island was observed as shown in Figure 34 below Hz/s 260 MW Hz/s Hz/s Hz Hz Hz 0MW 0MW No PV Response 17% Responsive PV 33% Responsive PV Max Frequency (Hz) Total OFA triggered (MW) RoCoF (Hz/s) Figure 34: North Island sensitivity to the level of PV generators with over-frequency response. It is apparent that as PV penetration increases on the system, the capability for PV generation to support the system in an over-frequency event will help mitigate the level of OFA generation tripped. The North Island studies show that with 17% of PV generation having over-frequency response, there are no OFA generators tripped to arrest the frequency rise on the system. When the South Island is subjected to a 400 MW trip in demand, the sensitivity to PV generation having over-frequency response is shown in Figure 35 below Hz/s 298 MW Hz/s Hz/s 69 MW Hz Hz Hz 32 MW No PV Response 17% Responsive PV 33% Responsive PV Max Frequency (Hz) Total OFA triggered (MW) RoCoF (Hz/s) Effect of Solar PV on Frequency Management in New Zealand Transpower New Zealand Limited. All rights reserved. 48

58 Section 4 Frequency stability study Figure 35: South Island sensitivity to the level of PV generators with over-frequency response. The South Island shows a similar reduction in the level of OFA tripped as large amounts of PV generation on the system have PV generators with over-frequency response. The detailed analysis of these results is presented in the following subsection Analysis The analysis of high PV penetration and its impacts on Transpower system operator to manage over-frequency events is broken down into the North Island and South Island. The study results explained in the previous subsection showed individual responses in each island to the increase in PV penetration. The same individual responses per an island were present for high PV penetration levels with various adoption levels of PV generation with over-frequency response. The following subsections analyse these studies in detail on an island basis North Island Compared with the South Island, the North Island has fewer generators with Over- Frequency Arming (OFA) ancillary service contracts used to mitigate frequency rise. Transpower system operator relies on sufficient levels of inertia, governor response and OFA to be available on the system in an over-frequency event to maintain statutory requirements of the Code. In the study, the North Island power system was subjected to the loss of 400 MW of demand at various levels of PV penetration to investigate the amount OFA that is tripped to arrest frequency fall, as shown in Figure 36 below. Figure 36: The North Island over- frequency response for three levels of PV generation with no frequency response. With PV generation offsetting large amounts of demand, there are less conventional generators available, causing a decrease in system inertia and governor response to help arrest frequency rise on the system. In all three PV penetration scenarios shown in Figure 36 above, OFA generation is tripped to arrest teh frequency rise. Effect of Solar PV on Frequency Management in New Zealand Transpower New Zealand Limited. All rights reserved. 49

59 Section 4 Frequency stability study When PV penetration reaches 2500 MW (orange curve) the system frequency is quickly arrested by governor action and the tripping of OFA generation, and as the frequency returns close to nominal it is over shot before stabilizing. Once PV penetration reaches 3250 MW (grey curve) the frequency response violates the 52 Hz North Island frequency limit requirement before recovers. The frequency violation results from reduced inertia causing the frequency to rise so quickly that the governor has limited time to respond, and preventing enough OFA generation to be tripped to arrest frequency rise. The consequences of frequency rising above 52 Hz in the North Island would be catastrophic, as most of the North Island thermal units may trip causing frequency to fall, resulting in an under-frequency event. Figure 37: The North Island 3250 MW PV penetration response with levels of over- frequency response It is apparent that the ability for the PV generation to response to over-frequency will enable the power system to handle over-frequency events in a high PV generation scenario without violating the 52 Hz limit. When PV generation reaches a certain penetration level, the collective response of the PV generation can have a positive impact on power system operation. This is provided that the PV generation response is coordinated properly with the actions from the other power system components South Island In the South Island there are currently more generators with Over-Frequency Arming (OFA) ancillary service contracts used to mitigate frequency rise when compared to the North Island. Transpower system operator is required to keep the system frequency at or below 55 Hz in the South Island. In all three over-frequency study cases, the 55 Hz limit was never violated due to there being a wider frequency band and because OFA generation trips earlier when compared to the North Island. The South Island power system was subjected to the loss of 400 MW of demand at various levels of PV penetration to investigate the amount OFA that is tripped to arrest frequency fall, as shown in Figure 38 below. Effect of Solar PV on Frequency Management in New Zealand Transpower New Zealand Limited. All rights reserved. 50

60 Section 4 Frequency stability study Figure 38: The South Island high PV over- frequency response for with adjusted OFA tripping times. With PV generation at 630 MW (blue curve) and 2500 MW (orange curve), approximately 317 MW and 210 MW of OFA generation was disconnected respectively when the frequency rose above 53 Hz. The tripping of OFA generating led to a sharp frequency turnaround at 53 Hz in both scenarios, resulting in the frequency falling down to approximately 49.7 Hz before stabilising. In a 3250 MW PV scenario, the current overfrequency mitigations prove to be sufficient in the ability to arrest frequency rise and bring the system frequency back to the normal band when the South Island is subjected to a 400 MW of load disconnection. If PV penetration has an over-frequency response, the capability of the South Island system to arrest the frequency rise due the loss of large amounts of demand will also increase, and the amount of OFA that needs to be tripped will be reduced. Effect of Solar PV on Frequency Management in New Zealand Transpower New Zealand Limited. All rights reserved. 51

61 Section 5 Key findings and conclusions 5 KEY FINDINGS AND CONCLUSIONS 5.1 Managing frequency within the normal band The studies described in this report indicated that with 1000 MW of PV generation in New Zealand, the effect is similar to the current power system with 30 MW of fictitious load switching 4 times an hour. The magnitude of frequency deviation is small, typically in the region of mhz. A PV penetration level of 1000 MW represents about 25% of the generation mix in midday summer. Considering typical system frequency noise of about 50 mhz (determined through analysis of historical data), the studies indicate that the New Zealand power system can accommodate about 3000 MW of PV generation and maintain acceptable frequency performance within the normal band with the present frequency keeping arrangements. The probability of exceeding the +/- 0.2 Hz normal band is small, about %. This equates to 33 minutes in a year that the frequency will deviate outside the +/- 0.2 Hz band. The study assumed a constant wind speed of 10 m/s. Other variables such as size of the cloud and direction of cloud movement are not considered in the study. In addition, the national PV generation is derived from solar irradiance data from 15 weather stations. The solar irradiance data from this limited number of weather stations made the PV generation profile at each GXP much more correlated than it should be. All the above factors will make the desktop studies more conservative than reality. The analysis used the available data at the time of the study and used some conservative assumptions to simplify the study methodology. The results obtained will be suitable for assessing the performance of the New Zealand power system to maintain frequency within the normal band. It is however prudent to re-check the study assumptions when the national PV penetration level goes beyond 1000 MW. It is expected that solar irradiance data from more weather stations will be available to better estimate the national PV generation. 5.2 Effect of low inertia system An increase of PV generation in the New Zealand power system will change its characteristics and dynamic behaviour, affecting the way frequency risk will be managed. It is expected that many of the largest thermal units will be dispatched off during the high PV generation scenario, leaving the power system with smaller hydro and geothermal units to meet demand. The inter-island HVDC flow will also be reduced, due to PV generation supplying demand locally. Therefore the frequency risk of the power system will be reduced, resulting in less reserve required to maintain the system frequency above the statutory limits. Conversely the power system s lower inertia, resulting from displacement of conventional synchronous generation by PV generation, will require more frequency reserve to mitigate the same quantum of risk. Procuring enough frequency reserve to mitigate the largest risk is a challenge under a high PV generation Effect of Solar PV on Frequency Management in New Zealand Transpower New Zealand Limited. All rights reserved. 52

62 Section 5 Key findings and conclusions scenario. However, a hydro generating unit can be scheduled in Tailed Water Depressed (TWD) mode to provide the required reserve. Procuring more IL is another option. Although the HVDC will be operating at very low levels during the high PV generation scenario, it is beneficial to keep the HVDC operating in round-power mode or monopole configuration to allow inertia sharing between the two islands. The contribution of the HVDC will decrease the RoCoF allowing more time for mitigation measures to act to restore the supply/demand balance following a contingent event. 5.3 Managing an under-frequency event Displacement of conventional generation from the power system removes a critical component for managing a frequency event caused by a supply-demand imbalance. The study showed that at 3250 MW of PV generation, the Rate of Change of Frequency (RoCoF) can be as high as 1 Hz/s with about 150 MW of imbalance in the power system. A higher imbalance will cause the RoCoF to increase. The study showed that the power system can manage frequency for 120 MW of CE risk for PV generation up to 3250 MW. If a larger thermal unit is scheduled, more IR is needed to maintain the system frequency above 48 Hz following the contingent event. TWD and IL are the options to increase the availability of IR for managing the frequency risk. The IR, combined with AUFLS, is adequate to manage the system frequency above 47 Hz in the North Island and 45 Hz in the South Island for an ECE contingency of 240 MW. A 240 MW ECE event will cause AUFLS to shed its first stage for a high PV generation scenario of 3250 MW. The shedding of 10% of the North Island load creates a resultant over-frequency. The design of the AUFLS and IL schemes is crucial for the lower inertia system, to avoid inaccurate shedding of load that could result in a frequency event and impact on security of supply. It is unclear whether the allocation of AUFLS load will be based on gross or net pre-event GXP load for a high PV penetration scenario. It is critical to establish this understanding to design the AUFLS and IL to work effectively when PV generation embedded in the distribution network is high and when there is no PV generation at night. 5.4 Managing an over-frequency event As the level of PV generation increases, the quantity of load that can be disconnected in a single event reduces, thus reducing the risk of over-frequency. Traditionally, the HVDC poses an over-frequency risk in the sending island when transfer is high. But from the generation dispatch report (refer to Effect of PV generation on generation dispatch in New Zealand report), the likelihood of this is low under the high PV generation scenario. In general, the HVDC transfer is expected to be below the 200 MW level. It is expected that the current OFA arrangement is adequate to manage over-frequency in this high PV generation scenario. Additionally, when the frequency reaches a pre-defined Effect of Solar PV on Frequency Management in New Zealand Transpower New Zealand Limited. All rights reserved. 53

63 Section 5 Key findings and conclusions threshold above nominal 50 Hz the solar PV inverter will ramp down its output. This will assist with managing the frequency within the 52 Hz limit in the North Island and 55 Hz in the South Island. 5.5 Effect of low inertia system on under- frequency mitigation measures A low inertia system and resultant high RoCoF prevents mitigation measures from acting in a timely fashion to correct the supply/demand imbalance caused by a contingent event. Governor action, IL, TWD and AUFLS take a finite time before they can react to the system frequency deviation by changing output or shedding load. The higher the RoCoF the lower the frequency before these measures can start restoring the supply/demand imbalance. If the system frequency gets close to the limits before these measures can act, it leaves very little time for it to recover before violating the limits. The effects of a low inertia are of particular concern in the North Island and can potentially cause the AUFLS scheme to lose frequency discrimination between the AUFLS stages, resulting in over-shedding and over-frequency. 5.6 Study limitations The results of this study should be considered in the context of the limitations present in the methodology. These limitations generally relate to the assumptions made in the study, i.e. the information available to create network and component models and the real-world conditions that cannot be adequately factored into the analysis. Some of the limitations are listed below: The generation dispatched is derived from SPD solves assuming the amount of NFR available in the system. This should be reasonably accurate and adequate for this type of study. There is limited solar PV inverter dynamic behaviour information to create an accurate inverter model. The inverter models are created based on inverter testing data from three types of inverter. The three inverters are modelled with no response from an under-frequency event. Two inverters are modelled to ramp down their active power output with droop-like characteristics and one inverter has no response to an over-frequency event. The distribution of inverter types is based on the manufacturer market share data in other countries. The AUFLS model is based on the Extended Reserve four stages scheme, and the quantum of load assigned to each stage is based on gross system load. All PV generation output materialises at the grid level; that is, there is no curtailment due to power quality issues or low-voltage feeder congestion, or loss of output due to local shading of PV panels. Effect of Solar PV on Frequency Management in New Zealand Transpower New Zealand Limited. All rights reserved. 54

64 Section 5 Key findings and conclusions All PV generation is treated as if located at the NIWA measurement site used for each region s PV profile. In reality, the aggregate curve would be spread wider with additional fluctuations due to geographic diversity and clustering of the PV sites. All PV generation is derived from solar irradiance data recorded from a limited number of weather stations. In reality, the PV generation would be less correlated between GXPs, thus making the national PV generation profile smoother. Energy storage and demand side management are not considered. In practice, these may be able to contribute to mitigating the impacts of the high PV penetration. The main consequence of these limitations is that the study results will be more conservative. The other factors that can affect the study results are: Severe weather front or solar eclipse affecting the variability of PV generation Market behaviour affecting frequency risk management Inverter dynamic behaviour Distribution network limit for PV generation export to the grid PV generation performance non-compliance with AS/NZS in riding through frequency and voltage disturbance Effect of Solar PV on Frequency Management in New Zealand Transpower New Zealand Limited. All rights reserved. 55

65 Section 6 Recommendations 6 RECOMMENDATIONS The New Zealand power system can accommodate a significant amount of PV generation before encountering challenges to managing system frequency. It is prudent to continue to improve our capability and tools in readiness for changes in power system characteristics and dynamic behaviour resulting from increasing PV generation. 6.1 Frequency management strategy A new strategy to manage frequency in a low inertia system will be required in future. A faster response reserve may be needed to ensure that mitigation measures work effectively. Future PV generation may have the capability to provide fast response akin to system inertia response to support system frequency during a frequency event. Other inverter based technologies, such as energy storage and Electric Vehicles (EV), are on the verge of becoming main stream. It is possible that these technologies could play a significant part in shaping the future frequency management strategy. The ability of energy storage technology to provide fast reserve will be studied in the next stage of our emerging technology programme of work. The effectiveness of AUFLS and IL will be studied to ensure they can continue to support system frequency during an under-frequency event. It is necessary to understand the AUFLS feeder configurations to accurately determine the quantum of load disconnected when the AUFLS operates. An AUFLS feeder with high PV penetration represents a lower load percentage as compared with a feeder containing little or no PV generation. Improvements in AUFLS computer model used in the power system simulation tools are necessary for more accurate assessments of system frequency performance. AUFLS relay model using probabilistic model for circuit breaker operating time to represent the AUFLS load disconnection action is more accurately. It allows Transpower system operator to study the operation of the AUFLS and IL more accurately and hence can optimise the scheme to work effectively in all system conditions.. With solar PV inverters ability to ramp back its output during an over-frequency event, it is prudent for Transpower to review future requirements and arrangements for overfrequency arming. Traditionally, the system is planned and operated to meet the maximum demand. Managing the minimum demand is a lesser challenge with demand about % less than the peak period. With high PV generation, the minimum demand occurs at midday can be less than 50% of the system evening maximum demand. The frequency management strategy and mitigation measures employed must be effective for both extreme system conditions to ensure adequate frequency reserve is available to manage credible risks in the power system. Effect of Solar PV on Frequency Management in New Zealand Transpower New Zealand Limited. All rights reserved. 56

66 Section 6 Recommendations 6.2 Inverter standard When PV generation penetration increases to a level that can impact the security of supply, it is essential it operates to the same standard as the grid-connected synchronous generators. The under-frequency and over-frequency anti-island set-point values specified in the AS/NZS need to be adhered to. An inverter that does not ride through a frequency event will further deteriorate the power system condition and potentially impact system recovery. It is important for all industry participants to work collaboratively to ensure compliance. 6.3 Long term factors The results and conclusions of this study should be disseminated to interested parties, to provide context for discussions regarding the future of PV generation and its impacts on the power system. Future frequency management strategies for low inertia systems should be formulated and ready for deployment when the growth of PV generation in New Zealand increases to a significant level. It is important to develop the capability to model solar PV technology and other emerging technologies adequately in simulation tools. The characteristics and behaviour of the power system will be impacted by increasing levels of PV generation. This needs to be well understood to ensure we can continue to operate the New Zealand power system securely, stably and economically. Collaboration with distribution network operators is necessary for realising the system benefits achievable from solar PV inverters. Effect of Solar PV on Frequency Management in New Zealand Transpower New Zealand Limited. All rights reserved. 57

67 Appendix A1: Current frequency management practices A1 CURRENT FREQUENCY MANAGEMENT PRACTICES This section of the appendix provides the reader with an overview of the current frequency management practices. A1.1 Contingent event classification The North Island power system has a generation trip risk of around 350 MW while the system has a peak size of 4500 MW and trough of 3500 MW. The HVDC can transfer up to 1200 MW of electrical power from South Island, which poses an under-frequency risk to the North Island and an over-frequency risk to the South Island if disconnected. Transpower assesses system risk routinely to enable secure operation of the power system and to comply with Transpower system operator obligations. The system risk is assessed and categorised based on a least cost criteria and the mitigation measures required to maintain system security during and following a power system event. To avoid power system events from triggering frequency excursions outside the statutory frequency limits, frequency risk is managed based on the risk class and the application of appropriate mitigation measures. Transpower has identified potential credible events that can occur in the New Zealand power system. Categorization of these events is based on the frequency of occurrence, impacts to the power system security and mitigation cost. Two different risk classes are identified namely Contingent Event (CE) and Extended Contingent Event (ECE). These are further described below. A1.1.1 Contingent Event (CE) A CE refers to a power system event that involves the disconnection of a single power system component; such as: A transmission circuit An HVDC link pole A single generating unit A single ancillary service injection point A load block A CE is an event where the impact, probability of occurrence plus estimated cost and benefits of mitigation are considered justification to apply mitigation measures pre-event to avoid post-event frequency exceeding the limits. Sufficient asset capacity and reserve should be dispatched to provide adequate redundancy to maintain the level of quality prescribed in the Code and to avoid post-event unplanned demand shedding. The underfrequency and over-frequency limits are set at 48 Hz and 52 Hz respectively for both islands. Effect of Solar PV on Frequency Management in New Zealand Transpower New Zealand Limited. All rights reserved. 58

68 Appendix A1: Current frequency management practices A1.1.2 Extended Contingent Event (ECE) An ECE refers to a power system event that involves the disconnection of multiple power system components by a single event or simultaneous disconnection of multiple components. Power system events classified under this category are: The loss of the HVDC link bipole The loss of a 220kV or 110 kv bus which can cause a disconnection of multiple generating units ECEs are not considered to justify the cost required to avoid any demand shedding. Hence, post-event demand loss is acceptable A1.2 Managing frequency in the normal band Small frequency deviations within the normal band are generally caused by demand fluctuations between trading periods, before re-dispatch can be carried out to balance supply and demand. These deviations are small and are usually corrected by the combination of free governor response and frequency keeping control. Transpower system operator procures frequency keeping services from one or more generating units to maintain system frequency within the normal band. Frequency keeping is dispatched every trading period and currently, 15 MW blocks of frequency keeping are dispatched in both islands. See the Frequency Management Barometer in subsection 4.3. A1.3 Managing Frequency Stability A1.3.1 Management of under-frequency deviation CE and ECE events involve either disconnection of generating units or loss of the HVDC bipole and will cause the system frequency to drop. Transpower system operator schedules and dispatches frequency reserve to balance the supply and demand during and following the occurrence of a CE or ECE. Sufficient frequency reserve is scheduled to meet the largest possible CE or ECE in either island. Following implementation of a National Instantaneous Reserve Market frequency reserve can be shared between the two islands. For an HVDC link bipole ECE, the receiving island will procure enough frequency reserve to meet the ECE frequency criteria. Transpower s Reserve Management Tool (RMT) is used to determine the amount of frequency reserve needed to cover the CE and ECE risks. Transpower system operator is required to schedule sufficient reserve to meet the specified under-frequency limits and avoid cascade failure. Different types of frequency reserve are available to counter an under-frequency excursion arising from a CE or ECE: Partial loaded Spinning Reserve (PLSR) and Tail-water depressed (TWD) reserve Interruptible Load (IL) Automatic Under-Frequency Load Shedding (AUFLS) Effect of Solar PV on Frequency Management in New Zealand Transpower New Zealand Limited. All rights reserved. 59

69 Appendix A1: Current frequency management practices These are described below. Partial loaded Spinning Reserve and Tail-water depressed reserve This type of frequency reserve supports system frequency by increasing the generators active power output during and following an under-frequency event. Frequency reserves are classified as either fast instantaneous reserve (FIR) or sustained instantaneous reserve (SIR): FIR is required to act in the first six seconds and must maintain its post event output for 60 seconds. SIR is required to act in the first 60 seconds and sustains its post event output for 15 minutes. The intention is to restore supply and demand balance in order to raise system frequency to HZ and above after a frequency event. The generator can operate in either PLSR or in TWD mode while offering frequency reserve: PLSR is partial load spinning reserve. This involves offering spare generation capacity into the reserve market while dispatching energy to supply the demand. The amount of PLSR that can be obtained from a generator depends on the speed of the governor control system setup and turbine control mechanism. TWD is tail water depressed reserve. This involves a generator operating as a motor to offer frequency reserve. In this operating mode the generator shaft spins freely, resulting in no electrical energy being generated. Control logic is configured to activate the changeover to generation mode, which kick starts the electrical energy generating process. The generator can start ramping up its active power output within 2-3 seconds to a predefined MW set point. Interruptible Load (IL) The IL scheme is designed to disconnect a specific amount of load automatically in order to correct any fall in system frequency following a tripping of a generator or the HVDC link. The IL is set to disconnect within 1 second after system frequency falls below 49.2 Hz, if it is offered to the reserve market as Fast Instantaneous Reserve (FIR). If offered for SIR and is set to disconnect within 60 Seconds. Automatic Under-Frequency Load Shedding (AUFLS) AUFLS is a protection scheme that reduces system load in an emergency situation, with the intention of being able then to restore the system to stable operation. AUFLS aims to Effect of Solar PV on Frequency Management in New Zealand Transpower New Zealand Limited. All rights reserved. 60

70 Appendix A1: Current frequency management practices avoid frequency drops below 47 Hz in North Island and 45 Hz in South Island following the discontinuation of a large generation source. The Code mandates for distributors (North Island) and Transpower (South Island) to provide an automatic disconnection of two blocks of demand, each block being a minimum of 16% of the total pre-event demand. The new automated scheme proposed in the Efficient Procurement of Extended Reserve (EPER) wil have four blocks of demand assigned to the scheme in North Island. The total load demand assigned to the new scheme will remain the same at 32% of the total preevent demand. South Island AUFLS configuration remains the same. A1.3.2 Management of over-frequency deviation Managing over-frequency is as challenging as managing under-frequency in a small power system like New Zealand. Disconnection of a large quantum of load or the HVDC bipole link can cause over-frequency in the power system. An over-frequency can activate the over-speed protection of the generators, resulting in subsequent system collapse due to an uncontrolled tripping of generators. Transpower system operator procures over-frequency reserve (OFR) to manage system frequency below 52 Hz power system. OFR is a contracted service fulfilled by a number of generator providers in each island which, when armed, disconnect automatically when the frequency reaches a given threshold. The OFR is armed and disarmed when needed based on the system conditions and the credible events that can pose over-frequency risk to the power system. The OFR trip frequency is configured to trip generators to manage over-frequency below 52 Hz in North Island and 55 Hz in South island. The generator trip frequency differs for each OFR provider in order to provide discrimination between generators. This is to avoid tripping too many generators and causing an under-frequency event. Effect of Solar PV on Frequency Management in New Zealand Transpower New Zealand Limited. All rights reserved. 61

71 Appendix A2: Inverter models A2 INVERTER MODELS As discussed in the main body of this report, the inverters models were produced based on the testing of 3 inverters, looking at 4 key characteristics: Active power control/frequency response Reactive power control/voltage response Frequency/voltage ride through capability Reconnect characteristics This section provides detail of the 4 aspects to the models for the 3 inverters, Inverter A, B and C. A2.1 Active power/frequency control Inverter A, B and C exhibit different active power/frequency control characteristics, with neither displaying under-frequency response. Inverter B showed no over-frequency response. Inverter A and C showed over-frequency response reducing the inverter electrical output linearly to the frequency. Inverter A is configured with a deadband of 0.2 Hz, and Inverter B with a deadband. Inverter C displayed a rapid ramp back to maximum power output once the frequency returned to within 51 Hz. Figure 39 below shows the over-frequency response of the three inverter types. Figure 40 shows the block diagram for the active power component of the Inverter C dynamic model. Inverter A and B are similar, but no ramp back characteristic and different droop, gains and deadband settings. Effect of Solar PV on Frequency Management in New Zealand Transpower New Zealand Limited. All rights reserved. 62

72 Appendix A2: Inverter models 8 Solar PV Inverter Over Frequency Response Active Power (MW) Frequency (Hz) Time (s) Inverter A Inverter B Inverter C Frequency Figure 39 PV generation Over-Frequency Response Effect of Solar PV on Frequency Management in New Zealand Transpower New Zealand Limited. All rights reserved. 63

73 Appendix A2: Inverter models Figure 40 Frequency Control component of Inverter C Dynamic Model Effect of Solar PV on Frequency Management in New Zealand Transpower New Zealand Limited. All rights reserved. 64

74 Appendix A2: Inverter models A2.2 Reactive Power/Voltage control All three tested inverters have selectable voltage control modes. The behaviour of the inverters in each mode varied between inverter types, however to simplify modelling requirements 2 control modes have been applied to all 3 inverters; volt-var mode and constant power factor mode. Both control modes have been modelled based on Inverter A response. The inverters have been modelled with a 33% limit on reactive power, as per the Inverter A tested value. The volt-var characteristic has a deadband and droop, with asymmetric settings for over and under-voltage response. The constant powerfactor characteristic has been modelled controlling the reactive power based on the active power output to maintain constant power factor. Due to the assumption of 0 MVAr output in the power-flow, this mode essentially equates to no voltage response in this study. Figure 41 and Figure 42 show the simulated response of the volt-var and constant power factor models. Figure 43 and Figure 44 show the block diagram for the volt-var and constant power factor component of the dynamic models. 2.0 Solar PV Inverter Voltage Step response - Volt-Var Mode Active/Reactive Power (MW/MVAr) Bus Voltage (pu) -2.0 Inverter Reactive Power (MVAR) Time (s) Bus Voltage (pu) 0.8 Figure 41 PV generation Volt-Var Response to voltage step Effect of Solar PV on Frequency Management in New Zealand Transpower New Zealand Limited. All rights reserved. 65

75 Appendix A2: Inverter models 8 7 Solar PV Inverter Over Frequency Response - Constant Power Factor Active/Reactive Power (MW/MVAr) Frequency (Hz) -2 Time (s) 47 Inverter Active Power (MW) Inverter Reactive Power (MVAr) TSAT Bus Frequency (Hz) Figure 42 PV generation constant power factor response to over-frequency event Effect of Solar PV on Frequency Management in New Zealand Transpower New Zealand Limited. All rights reserved. 66

76 Appendix A2: Inverter models Figure 43 Volt-Var component of Inverter Dynamic Model Effect of Solar PV on Frequency Management in New Zealand Transpower New Zealand Limited. All rights reserved. 67

77 Appendix A2: Inverter models Figure 44 Constant Power Factor component of Inverter Dynamic Model Effect of Solar PV on Frequency Management in New Zealand Transpower New Zealand Limited. All rights reserved. 68

78 Appendix A2: Inverter models A2.3 Voltage/Frequency ride through Inverters A, B and C displayed different characteristics when tested for voltage and frequency ride through. Inverter B is most resilient to under-voltage, with tests showing it remaining connected at 10 V (0.043 pu) for 1 second. Inverter C has an instantaneous under-voltage trip at 160 V (0.696 pu), and Inverter A will trip at 45 V (0.195 pu) for 10ms. All inverters are able to ride through most frequency events which can be expected in the New Zealand power system. Inverter B and C may trip in a South Island over-frequency event assisting in the arrest of over-frequency. Inverter C may trip for a South Island extended contingent event (ECE) as the frequency could potentially fall to 45 Hz. Due to the requirement to also model inverter reconnecting characteristics, the voltage and frequency "trips" have been modelled as limiters on the active and reactive power control blocks, rather than disconnections of the model from the simulation. Figure 45 shows the block diagram for the frequency and voltage trip and reconnect component of Inverter C, and the interface to Pmax and Qmax. Inverter A and B have the same structure, with settings according to Table 4 below. Effect of Solar PV on Frequency Management in New Zealand Transpower New Zealand Limited. All rights reserved. 69

79 Appendix A2: Inverter models Table 4 Inverter voltage and frequency ride trip settings Inverter Type Voltage Frequency Vup1 TVup1 Vup2 TVup2 Vdown1 TVdown1 Vdown2 TVdown2 Fup TFup Fdown TFdown (pu) (s) (pu) (s) (pu) (s) (pu) (s) (Hz) (s) (Hz) (s) Inverter A Inverter B Inverter C Figure 45 Voltage and Frequency trip and reconnect component of Inverter C Dynamic Model Effect of Solar PV on Frequency Management in New Zealand Transpower New Zealand Limited. All rights reserved. 70

80 Appendix A2: Inverter models A2.4 Reconnect characteristics Inverters A, B and C displayed different reconnect characteristics following a frequency or voltage trip. The reconnection was characterised by the reset time and rate at which the inverter returns to maximum power output. Inverter A had a reset time of 90s, and returned to maximum power output with a slow ramp for the first 50% over approximately 50s, then a fast ramp for the remaining 50% over approximately 5s. This was approximated in the model as a ramp to maximum power output over about 40s. Inverter B had a reset time of 50s and reconnected with a step to maximum power output. Inverter C had a reset time of 30s and reconnected with a step to maximum power output. Figure 45 above shows the block diagram for the frequency and voltage trip and reconnect component of Inverter C, and the interface to Pmax and Qmax. Effect of Solar PV on Frequency Management in New Zealand Transpower New Zealand Limited. All rights reserved. 71

81 Appendix A3: Generation Dispatch A3 GENERATION DISPATCH A3.1 Generation dispatch for 630 MW PV generation scenario North island Generator ID Gen. Output (MW) South Island Generator ID Gen. Output (MW) ANI ALD ANI AMS ARA ARG ARI ARG ARI AVI ARI BEN ARI BEN ARI BEN ARI COB ATI COB GLN COB KAG COB KIN COB KMI COB KPI COL KPI CYD KPO HBK KPO KUM MAT MAH MCK MAN MCK MAN MCK MAN MHO MAN MOK MSG MOK OHA MOK OHB MOK A 20 OHC MOK PRU MOK ROX MOK TKA MOK TKB Effect of Solar PV on Frequency Management in New Zealand Transpower New Zealand Limited. All rights reserved. 72

82 Appendix A3: Generation Dispatch North island Generator ID Gen. Output (MW) South Island Generator ID Gen. Output (MW) MOK WHL MOK WTK MTI WTK MTI WTK MTI WTK MTI MTI NAP NGA NTM NTM NTM OHK OKI OKI ONU ONU ONU PPI PTA PTA PTA RKA RKA RPO RPO SFD SFD TAA TAA TAP THI THI Effect of Solar PV on Frequency Management in New Zealand Transpower New Zealand Limited. All rights reserved. 73

83 Appendix A3: Generation Dispatch North island Generator ID Gen. Output (MW) South Island Generator ID Gen. Output (MW) TKU TRC TRH TUI TUI TUI TUK TWC TWF TWF WAA WAA WAA WHE WKM WKM WPA WRK WRK WRK WRK WRK WRK WRK WRK WWD WWD Effect of Solar PV on Frequency Management in New Zealand Transpower New Zealand Limited. All rights reserved. 74

84 Appendix A3: Generation Dispatch A3.2 Generation dispatch for 2500 MW PV generation scenario North island Generator ID Gen. Output (MW) South Island Generator ID Gen. Output (MW) ARI ALD ARI AMS GLN ARG KAG ARG KIN AVI KMI BEN KPI BEN MAT COB MCK COB MCK COB MOK COL MOK CYD MOK HBK MOK ISL MOK MAH MOK A 20 MAN MOK MAN MOK MSG MOK OHA MOK OHB MTI OHC MTI PRU NAP ROX NGA ROX OKI ROX OKI TKA ONU TKB ONU WHL PTA WTK PTA WTK PTA WTK RPO WTK Effect of Solar PV on Frequency Management in New Zealand Transpower New Zealand Limited. All rights reserved. 75

85 Appendix A3: Generation Dispatch North island Generator ID Gen. Output (MW) South Island Generator ID Gen. Output (MW) SFD SFD TAP THI THI TRC TRH TUK TWC TWF TWF WAA WHE WRK WRK WRK WRK WRK WRK WRK WRK WWD WWD Effect of Solar PV on Frequency Management in New Zealand Transpower New Zealand Limited. All rights reserved. 76

86 Appendix A3: Generation Dispatch A3.3 Generation dispatch for 3250 MW PV generation scenario North island Generator ID Gen. Output (MW) South Island Generator ID Gen. Output (MW) ARA ALD ARI AMS ARI ARG GLN ARG KAG AVI KIN BEN KPI BEN MAT COB MOK COB MOK COB MOK COL MOK CYD MOK HBK MOK MAH MOK MAN MTI MAN NGA MSG OKI OHA ONU OHB ONU OHC PTA PRU PTA ROX PTA ROX RKA TKA RPO TKB SFD WHL SFD WTK TAP WTK THI TRC TUK WAA Effect of Solar PV on Frequency Management in New Zealand Transpower New Zealand Limited. All rights reserved. 77

87 Appendix A3: Generation Dispatch North island South Island WHE WRK WRK WRK WRK WRK WRK WRK Effect of Solar PV on Frequency Management in New Zealand Transpower New Zealand Limited. All rights reserved. 78

88 Appendix A4: Summer Sunday generation and grid zones power-flow A4 SUMMER SUNDAY GENERATION AND GRID ZONES POWER-FLOW Figure 46: Generation Mix and HVDC bipole transfer Figure 47: North Island inter-grid Zone electrical power-flow Effect of Solar PV on Frequency Management in New Zealand Transpower New Zealand Limited. All rights reserved. 79

89 Appendix A4: Summer Sunday generation and grid zones power-flow Figure 48: Grid Zone 1 generation, demand and inter-grid Zone power-flow Figure 49: Grid Zone 2 generation, demand and inter-grid Zone power-flow Figure 50: Grid Zone 3 generation, demand and inter-grid Zone power-flow Effect of Solar PV on Frequency Management in New Zealand Transpower New Zealand Limited. All rights reserved. 80

90 Appendix A4: Summer Sunday generation and grid zones power-flow Figure 51: Grid Zone 4 generation, demand and inter-grid Zone power-flow Figure 52: Grid Zone 5 generation, demand and inter-grid Zone power-flow Figure 53: Grid Zone 6 generation, demand and inter-grid Zone power-flow Effect of Solar PV on Frequency Management in New Zealand Transpower New Zealand Limited. All rights reserved. 81

91 Appendix A4: Summer Sunday generation and grid zones power-flow Figure 54: Grid Zone 7 generation, demand and inter-grid Zone power-flow Figure 55: Grid Zone 8 generation, demand and inter-grid Zone power-flow Figure 56: South Island inter-grid Zone electrical power-flow Effect of Solar PV on Frequency Management in New Zealand Transpower New Zealand Limited. All rights reserved. 82

92 Appendix A4: Summer Sunday generation and grid zones power-flow Figure 57: Grid Zone 9 generation, demand and inter-grid Zone power-flow Figure 58: Grid Zone 10 generation, demand and inter-grid Zone power-flow Figure 59: Grid Zone 11 generation, demand and inter-grid Zone power-flow Effect of Solar PV on Frequency Management in New Zealand Transpower New Zealand Limited. All rights reserved. 83

93 Appendix A4: Summer Sunday generation and grid zones power-flow Figure 60: Grid Zone 12 generation, demand and inter-grid Zone power-flow Figure 61: Grid Zone 13 generation, demand and inter-grid Zone power-flow Figure 62: Grid Zone 14 generation, demand and inter-grid Zone power-flow Effect of Solar PV on Frequency Management in New Zealand Transpower New Zealand Limited. All rights reserved. 84

94 Appendix A5: Emerging Energy Programme: plan and outcome strategy A5 EMERGING ENERGY PROGRAMME: PLAN AND OUTCOME STRATEGY Emerging Energy Technologies: Programme Tranche Plan Historic Work 2016/ / /19 Wind Wind Capacity Assessment Solar PV Solar PV Variability Studies Solar PV System Stability Studies Training Battery and Storage Market System, Real Time Operations, Process and People Situational Intelligence Initial Work Battery Storage Trial Invex Stage 1 Situational Intelligence Programme Definition Battery Operations Impact Assessment Situational Intelligence Stage 1 Invex Battery Storage Next Steps Consideration of Economic Options for Investment Situational Intelligence Stage 1 Capex Situational Intelligence Future Phasing Monitoring Progress Against Transmission Tomorrow Future States (ongoing) Work Packages to be Executed with each Emerging Technology Lines Company Data Exchange Review Assessment of Capabilities Load Forecast Review Ancillary Services Review SO Tools Review Policy and Standard Review Effect of Solar PV on Frequency Management in New Zealand Transpower New Zealand Limited. All rights reserved. 85

95 Appendix A5: Emerging Energy Programme: plan and outcome strategy A5.1 Emerging Energy Technologies Outcome Strategy Map Effect of Solar PV on Frequency Management in New Zealand Transpower New Zealand Limited. All rights reserved. 86