Automatic Under-Frequency Load Shedding (AUFLS)

Transcription

1 System Operator Report: Automatic Under-Frequency Load Shedding (AUFLS) Economic and Provision Page 1 of 43 Automatic Under-Frequency Load Shedding (AUFLS) Rate of Change of Frequency Testing & Recommendation System Operator 20/07/2012

2 System Operator Report: Automatic Under-Frequency Load Shedding (AUFLS) RoCoF Testing Page 2 of 43 COPYRIGHT 2012 TRANSPOWER New Zealand LIMITED ALL RIGHTS RESERVED The information contained in the report is protected by copyright vested in Transpower New Zealand Limited ( Transpower ). The report is supplied in confidence to you solely for your information. No part of the report may be reproduced or transmitted in any form by any means including, without limitation, electronic, photocopying, recording, or otherwise, without the prior written permission of Transpower. No information embodied in the report which is not already in the public domain shall be communicated in any manner whatsoever to any third party without the prior written consent of Transpower. Any breach of the above obligations may be restrained by legal proceedings seeking remedies including injunctions, damages and costs. LIMITATION OF LIABILITY/DISCLAIMER OF WARRANTY Transpower make no representation or warranties with respect to the accuracy or completeness of the information contained in the report. Unless it is not lawfully permitted to do so, Transpower specifically disclaims any implied warranties of merchantability or fitness for any particular purpose and shall in no event be liable for, any loss of profit or any other commercial damage, including but not limited to special, incidental, consequential or other damages. Date Prepared by: System Operator July 2012 This report and the appendices are available to download from the System Operator website at

3 System Operator Report: Automatic Under-Frequency Load Shedding (AUFLS) RoCoF Testing Page 3 of 43 TABLE OF CONTENTS 1 EXECUTIVE SUMMARY INTRODUCTION Background and Purpose Providing comments to the System Operator Survey of North Island Distributors The AUFLS testing process (current and future) Cost recovery of scheme changes The reliable use and implementation of the RoCoF scheme Objective of the AUFLS Scheme The challenges of achieving block discrimination The benefit of faster AUFLS operation Proposed benefit of RoCoF INTERNATIONAL USE OF ROCOF FOR AUFLS Use of RoCoF in Tasmania CALCULATING THE RATE OF CHANGE OF FREQUENCY Frequency Calculation Rate-of-Change-of-Frequency Calculation ROCOF BENCH TESTING - METHODOLOGY Initial RoCoF relay performance requirements Test equipment arrangement and process Test Cases Settings Used The role of the frequency guard BENCH TESTING RESULTS Bench Testing Results Summary Logic Stability Uniformity and Response Time Accuracy INACCURACY FROM POWER SYSTEM OSCILLATIONS Limitation in calculating frequency and RoCoF Impact of system oscillations on RoCoF Calculations Oscillations during Under Frequency Events Mitigating the effects of power system oscillations The Frequency Guard Increasing the RoCoF calculation time Low-Pass Filtering ROCOF RECOMMENDATION SUMMARY SOUTH ISLAND BLOCK 2 INCREASE Background Cost Benefit Analysis Methodology Cost Benefit Analysis Results NEXT STEPS SUMMARY OF QUESTIONS FOR CONSIDERATION ACKNOWLEDGEMENTS REFERENCES... 43

4 System Operator Report: Automatic Under-Frequency Load Shedding (AUFLS) RoCoF Testing Page 4 of 43 1 Executive Summary Automatic Under-Frequency Load Shedding (AUFLS) is vital to managing power system security and acts as a safety net to prevent power system collapse and blackout following large, rare system events. New Zealand s AUFLS scheme is made up of a minimum of two 16% blocks in each island. This means that 32% of customer demand can be automatically disconnected to restore stability to the power system. In 2009, the System Operator determined that a technical review of these arrangements was required. The results of the technical review concluded that the Reserve Management Tool (RMT) and the under-frequency products available for use by the System Operator should prevent system collapse from large defined risks (such as the sudden disconnection of HVDC bi-pole) at all times 1. However, the results of the technical review demonstrated that the operation of the current AUFLS scheme could result in over-frequency and potentially system collapse from some of the defined risks. Further, the current AUFLS scheme does not provide the System Operator with sufficient confidence that it will prevent the system from collapsing following undefined larger risks. The System Operator identified technically feasible options for improving the performance of the AUFLS system and under took a cost-benefit analysis of those options. One of the options proposed included the use of rate-of-changeof-frequency (RoCoF) relays. The analysis revealed the scheme, using RoCoF elements, resulted in the largest benefit for North Island. Following the cost-benefit analysis the System Operator conducted a literature review which revealed the use of RoCoF relays for under-frequency load shedding schemes is rare. In addition, some North Island distributors voiced concern whether the RoCoF relays would increase risk of trip due to maloperation. The System Operator tested several RoCoF relays to address the issue of relay reliability and to identify the Code requirements to ensure the implementation of a reliable AUFLS scheme. This report presents the findings and conclusions of the RoCoF relay bench testing. The testing process aimed to verify the logic, stability, uniformity, response time, and accuracy of the relays to identify the requirements for reliable use of RoCoF relays on the New Zealand power system. Using RoCoF technology for the proposed AUFLS scheme requires a level of logic to implement a frequency guard and backup under-frequency settings. All of the relays tested had some form of logic programming that allowed users to implement the required combinations of elements. Each manufacturer has a different way of arranging and setting their logic elements and care is required to ensure the desired result is achieved when programming and setting the various relays. The bench testing aimed primarily at ensuring the RoCoF relays remain stable during a range of system disturbances that would not normally result in AUFLS operation. The tests results indicate that the sample RoCoF relays were stable under all of the disturbances to which they were subjected. The guard frequency used in these tests likely plays a major role in preventing the undesired operation of RoCoF. 1 This could require the transfer on the HVDC link to be limited to below its maximum capability under certain system conditions to ensure power system security.

5 System Operator Report: Automatic Under-Frequency Load Shedding (AUFLS) RoCoF Testing Page 5 of 43 A level of uniformity is required to ensure the installed AUFLS system achieves the desired levels of operation should a range of different relays be used. The tests demonstrated that the sampled RoCoF relays do not behave uniformly when subjected to identical inputs. This is not necessarily a problem provided that relays comply within a specified maximum allowed response time. The RoCoF relays must be able to accurately calculate the rate of system frequency decay when subjected to a variety of grid conditions to be used for load shedding. The consequences of inaccuracy could result in the relay maloperating by tripping inappropriately. The relays performed within the specified accuracy margins for clean frequency decays. However, the testing results did verify that RoCoF relays are susceptible to calculation inaccuracy due to system oscillations. RoCoF relays use low pass filters to manage the impact of power oscillations on the RoCoF calculation. Ideally, the low pass filters would attenuate the noise and oscillations while leaving the slower-changing underlying RoCoF relatively unaffected. It is recommended that a filtering standard be required for all relays, regardless of the RoCoF settings. The System Operator has concluded the use of RoCoF relays is feasible on the New Zealand system provided the following requirements are specified in the code: 1. The required RoCoF calculation filtering to mitigate the inaccuracy caused by an identified level of oscillations during under-frequency events. 2. The maximum response time required for the RoCoF settings proposed so that the various relays provide a degree of uniformity. 3. A test regime to verify the logic elements of the installed relays meet the performance requirements. The System Operator requests that Industry reviews the tests results and provides feedback on the level of confidence in the stability of the relays. For the South Island, the System Operator believes it is prudent to hold off proposing new AUFLS schemes until there is further clarity of the future of the frequency band and AUFLS provision at the Tiwai grid exit point. In the interim, the System Operator recommends the trip setting of the second AUFLS block is increased to 46.5 Hz to improve the capability of the current South Island AUFLS scheme. The implementation cost for the change is assessed as low so a high level costbenefit analysis was undertaken to avoid costly studies. The cost-benefit analysis suggests the setting increase will be economically beneficially. The testing results and South Island benefit analysis were presented and discussed with industry at System Operator workshops commencing on 9th July Following the workshops, the System Operator will consider industry feedback before making recommendations to the Electricity Authority regarding the future AUFLS arrangements in both islands.

6 System Operator Report: Automatic Under-Frequency Load Shedding (AUFLS) RoCoF Testing Page 6 of 43 2 Introduction 2.1 Background and Purpose AUFLS is the acronym for Automatic Under-Frequency Load Shedding and describes the set of relays in New Zealand which automatically trip blocks of load following a severe under-frequency event to seek to restore the system frequency. These relays are relied upon by the System Operator to prevent the collapse of the system from under-frequency following events which have the potential to cause a system blackout. New Zealand s current AUFLS scheme is made up of a minimum of two 16% blocks in each island. This means that 32% of customer demand can be automatically disconnected to restore stability to the power system. The AUFLS obligations are set out in Part 8 Schedule 8.3 Technical Code B of the Electricity Industry Participation Code ( the Code ). Distributors 2 in the North Island and Grid Owners in the South Island are required to provide a minimum of 2 x 16% blocks of AUFLS as described in Table 1 below: Table 1 Required AUFLS settings North Island South Island Block 1 Block 2 Block 1 Block 2 Trip Frequency (Hz) Time Delay (sec) nd Trip Frequency (Hz) Time Delay (sec) The current AUFLS arrangements are largely based on historical practice. In 2009, the System Operator determined that a technical review of these arrangements was required. This technical review was completed in The results of the technical review concluded that the Reserve Management Tool (RMT) and the under-frequency products available for dispatch should prevent system collapse from large defined risks, such as the sudden disconnection of HVDC bi-pole, at all times 4. However, the results of the technical review demonstrated that the operation of the current AUFLS scheme could result in over-frequency and potentially system collapse from some of the defined risks. There is also concern that the current AUFLS scheme does not provide the System Operator with sufficient confidence that it will be effective to prevent the system from collapsing following a range of undefined larger risks. The technical review found the lack of discrimination between AUFLS blocks can result in more AUFLS tripping than is required which leads to dangerous overfrequency conditions and potential combined cycle generators tripping, which then can cause another under-frequency event and, without any AUFLS remaining, system collapse. To address the issues identified in the technical review, the System Operator worked through a process of identifying technical feasible options aimed at improving the performance of the AUFLS system and undertaking cost-benefit 2 The obligations includes consumers connected directly to the grid, also known as direct connects 3 The technical report can be found at 4 This could require the transfer on the HVDC link to be limited to below its maximum capability under certain system conditions to ensure power system security.

7 System Operator Report: Automatic Under-Frequency Load Shedding (AUFLS) RoCoF Testing Page 7 of 43 analyses of those options. One of the options proposed included the use of rateof-change-of-frequency (RoCoF) relays. The economic analysis revealed the scheme using RoCoF elements resulted in the largest benefit range for North Island 5. The results of the economic analysis were presented to industry in August 2011 and industry feedback was received on the proposed changes. Any changes to the AUFLS scheme will require changes to the Electricity Industry Participation 2010 Code (Code). In November 2011, the Electricity Authority agreed that the System Operator would complete the RoCoF testing work required to compile code change recommendations. The System Operator finalised the scope of the project after collecting feedback from the Electricity Authority and industry participants. The purpose of RoCoF testing work is to determine the requirements that will provide a reliable, secure, and efficient RoCoF AUFLS system to deliver greater certainty on system integrity during major under-frequency events. To make the code recommendations the RoCoF testing work aimed at the following: evaluating the technical requirements for the reliable use of RoCoF relays; outlining the required installation design outcomes to provide flexibility in meeting the requirements; and developing a proposed implementation plan to maintain system security during rollout. The work also included completing a high level cost-benefit analysis of proposed South Island setting changes. The purpose of this report is to set out the System Operator s findings on the above 6 and to seek industry and stakeholder views on such findings. A high level overview will be presented here while a detailed technical report is attached as Appendix B. Once the System Operator has considered any comments received in respect of this report s findings, it expects to make any code recommendations to the Electricity Authority. 5 The economic report can be found at 6 This excludes the code recommendations themselves and the implementation plan which will be published at a later date.

8 System Operator Report: Automatic Under-Frequency Load Shedding (AUFLS) RoCoF Testing Page 8 of Providing comments to the System Operator The System Operator requests comments on this report s findings by Friday 17 August 2012 so that it can continue the next phase of work. The System Operator s preference is to receive submissions in electronic form. The electronic version should be ed with the phrase Submissions on AUFLS scheme options in the subject header to justin.blass@transpower.co.nz. Hard copies can be posted to the following address: Mr Justin Blass Transpower New Zealand Limited Level 7, Transpower House 96 The Terrace PO Box 1021 Wellington New Zealand The System Operator will acknowledge receipt of all submissions electronically. The System Operator values openness and transparency and therefore comments will be published on the System Operator s website. Those providing comments should discuss with the System Operator any intended provision of confidential information, prior to sending the information.

9 System Operator Report: Automatic Under-Frequency Load Shedding (AUFLS) RoCoF Testing Page 9 of Survey of North Island Distributors Following the economic review the System Operator engaged in discussions with North Island distributors 7 to ensure the concerns of the obligated parties were captured prior to completing the next scope. The discussions were grouped into the following categories: Load allocation (specifically the management of critical load) Relay ownership (specifically management of relays at the Grid Exit Point Level) RoCoF Feasibility and Implementation Market Arrangements and load optimisation The discussions helped to shape the work being carried out in the AUFLS programme. The views presented during these discussions were varied. However, the commonly re-occurring themes were 8 : 1. The AUFLS testing process (current and future) 2. Cost recovery of required scheme changes 3. The reliable use and implementation of the RoCoF scheme In light of the industry discussion and the present scope of work the System Operator invited industry participants to participate in an advisory group. The purpose of the advisory group was to provide additional input into the testing of the RoCoF relays. Participation in the advisory group included providing comment on the testing methodology and discussing these comments via teleconferences 9. The advisory group had nine members from different distribution companies that provided input into the initial testing process. The System Operator would like to thank the participants for their input and cooperation with the development of the RoCoF relay bench testing process The AUFLS testing process (current and future) The Code currently requires the testing of AUFLS operation every 4 years regardless of relay type 10. However, various parts of the Code stipulate the testing requirements for analogue, non-self-monitoring digital, and self-monitoring digital protection systems 11. The self-monitoring digital protection systems are required to be tested every 10 years while the others are every 4. Several of the distributors suggested bringing consistency to the code requirements by allowing AUFLS provided on self-monitoring digital protection systems to be tested every 10 years. In addition, some expressed concern over the clarity of roles when AUFLS operation is provided at the Grid Exit Point (GXP) level. Distributors had a range of experiences managing the testing and status of AUFLS relays installed inside Transpower controlled sites. Although the obligation for testing and operation of the relays remains with the distributor; there appears to be no clear process to enable the coordination required. It was suggested that a process be developed to ensure consistency 7 Due to availability and timing constraints the following parties were not surveyed; Centralines, Scanpower, and Electra. 8 Other concerns included proper incentives for scheme development,.4s IL trip time proposal, and visibility of load aggregation and AUFLS provision. 9 Minutes from the advisory group teleconferences can be found here 10 Schedule 8.3, Technical Code A, Appendix B, clause 6 and 7 11 Schedule 8.3, Technical Code A, Appendix B, clause 13(a),(b), and (c)

10 System Operator Report: Automatic Under-Frequency Load Shedding (AUFLS) RoCoF Testing Page 10 of 43 and clarity in managing GXP level AUFLS relays. The Grid Owner has been notified and is establishing a person responsible for this coordination. Further concern was expressed with the development of the testing process for the proposed use of RoCoF elements. Many distributors suggested further work is done to develop the testing process and to ensure the practicality of testing is considered. The System Operator is considering the different technical requirements and the testing process, and expects to publish guidelines with code recommendations Cost recovery of scheme changes Another common concern raised by the distributors was their ability to budget for and recover the costs associated with any code changes. A two year lead time is considered necessary to roll out changes to the AUFLS scheme. It was suggested that any implementation plan would need to consider the cost recovery process. Further consideration will be included as part of the code recommendation and implementation plan The reliable use and implementation of the RoCoF scheme The third common concern raised by distributors revolved around the use of RoCoF elements to trigger load. Several distributors queried whether the use of RoCoF elements would increase the risk of mal-operation of AUFLS. In addition to mal-operation, other expressed concerns about their ability to implement the additional RoCoF blocks. This was of particular concern where recloses are currently used to provide AUFLS. The System Operator completed testing work, which is outlined in this report, to specifically address the concern of mal-operation of the RoCoF relays. The testing methodology and results are highlighted in a further section. Concern over implementation will be considered as part of the code design process.

11 System Operator Report: Automatic Under-Frequency Load Shedding (AUFLS) RoCoF Testing Page 11 of Objective of the AUFLS Scheme Before determining the performance requirements of the RoCoF elements for under-frequency load shedding it is important to consider the primary objective of the AUFLS scheme; to perform load reduction to match significant generation loss to arrest the frequency decay and return the frequency to the normal operating while maintaining the frequency within the safe operating range. The performance criteria of the AUFLS scheme are: Arrest the system frequency within the safe operating range Return the system frequency to the normal range while remaining within the safe operating range. Achieving the stated objected of the AUFLS scheme relies upon two key factors: block discrimination and speed of operation The challenges of achieving block discrimination Ideally the AUFLS scheme will shed only the amount of load required to arrest the system frequency and return it to normal. Excessive load shedding can cause the frequency response to overshoot. This is of special concern for the North Island, as there are significant amounts of thermal generation which will trip on over-frequency protection when operating above 52 Hz. If an over-frequency event occurred following AUFLS operation, then it is possible some thermal units may trip and there will be a second dip in system frequency which may not be recoverable as there may be insufficient AUFLS remaining to balance this subsequent loss of generation. Having a number of small blocks enables better matching of load shedding to generation loss which reduces the potential for over-frequency and overvoltage 12. An effective AUFLS scheme made up of a number of small load shedding blocks requires the blocks to be able to each operate before the next block triggers; this is known as discrimination. In general, discrimination ensures the scheme sheds only the load required to arrest and return the frequency to the normal range 13. As the rate of frequency decay increases following major system disturbances, so does the difficulty in maintaining discrimination between AUFLS blocks. The New Zealand system currently has only 1 Hz range (48 Hz 47 Hz) within which AUFLS blocks must operate within. The limitation is imposed to ensure that the AUFLS scheme does not operate for a contingent event. This limits the ability to obtain block discrimination. The current scheme will not achieve discrimination for most AUFLS events. 12 System Operator, Appendix C - Calculating the df/dt settings, found here: 13 Alternate scheme designs are feasible that may not require discrimination between blocks provided there are numerous blocks that are small enough to achieve some level of discrimination (i.e. between every other block). See Ireland s and Tasmania s systems.

12 System Operator Report: Automatic Under-Frequency Load Shedding (AUFLS) RoCoF Testing Page 12 of 43 Figure 1 Overview of requirements for block discrimination Figure 1 demonstrates the challenge for the current AUFLS scheme to achieve discrimination. Relays equipped for AUFLS utilise a time delay to ensure the calculated frequency stays at or below the trip point for a set amount of time; this is represented in Figure 1 by td1. The time delay is often adjustable and is chosen by the obligated parties when tested (test reports show it can range from ms). Once the time delay expires, and the conditions are met, the relay sends a trip signal. Time taken for the signal to be processed and the circuit breaker to open is represented by td2. The time from the relay signal sent to circuit breaker operation is not tested and so assumptions are made by obligated parties to ensure the code requirements are met (anywhere from 80ms -120ms has been assumed). For an event that causes the frequency to fall at a rate of 1 Hz/s the second block of AUFLS has already initiated its time delay before the first block of AUFLS has tripped. If the frequency does not increase above the block 2 trip point (47.5 Hz) before the time delay has expired block 2 will trip. The system has very little time to recover and prevent the second block of AUFLS from triggering, which will likely result in frequency over-shoot The benefit of faster AUFLS operation In addition to block discrimination, the speed of operation was shown in the AUFLS technical review to be critical to arrest the system frequency following major system disturbances. Increasing the speed of operation enables additional time for generators to respond which will assist in arresting the frequency. Tripping load at higher frequencies increases the minimal frequency reached following disturbances and enables the system to respond to larger disturbances.

13 System Operator Report: Automatic Under-Frequency Load Shedding (AUFLS) RoCoF Testing Page 13 of Proposed benefit of RoCoF The existing AUFLS scheme uses under-frequency elements to trigger the AUFLS blocks. This means the AUFLS will trip once the frequency has dropped below a set frequency for a set period of time. Another way to trigger AUFLS is to use rate of change of frequency (RoCoF) elements, often referred to as df/dt. These elements will trip the AUFLS blocks once the frequency fall has reached a certain speed. One of the benefits of utilising the RoCoF elements is that it enables AUFLS to be triggered at frequencies higher than the current contingent event target frequency; 48 Hz 14. Triggering above 48 Hz increases the speed of operation and will raise the minimum frequency following most large events. This enables the AUFLS scheme to cover for larger events while maintaining the system frequency above 47 Hz limit. Triggering AUFLS blocks above 48 Hz also enables additional blocks to be added and maintain block discrimination. A critical factor in block discrimination is the rate of frequency fall. RoCoF elements enable the AUFLS blocks to be arranged to better match of load shed and generation loss 15. This should reduce the amount of unnecessary load shed and also reduce the amount of frequency over-shoot that can follow AUFLS operation. The cost benefit considered the reduction in load shed as the key benefit for using RoCoF relays and resulted in a net benefit range of $16 million to $89 million over 15 years. In addition increasing the minimum frequency following AUFLS operation enables greater levels of DC transfer, without the procurement of additional reserves. 14 This ability requires discrimination from CE events. Previous recommendations suggest increasing the speed of interruptible load to ensure greater difference between the speed of frequency fall for CE and ECE/Other events. 15 Articles in the literature review outline that the initial rate of frequency fall after an event is an indicator of the power system imbalance. If factors such as the system inertia can be calculated there is potential future of AUFLS scheme to have an adaptable method to load shedding [1][2].

14 System Operator Report: Automatic Under-Frequency Load Shedding (AUFLS) RoCoF Testing Page 14 of 43 3 International Use of RoCoF for AUFLS An investigation was carried out to identify systems utilising RoCoF elements in their under-frequency load shedding schemes and to learn from their experience. The systems investigated revealed that the use of RoCoF elements for underfrequency load shedding schemes is rare. The majority of systems rely on underfrequency triggering elements and do not report the same drivers (i.e. low system inertia and narrow AUFLS frequency band) that would necessitate the use of RoCoF elements. A more common use of RoCoF elements is for generator unit islanding protection. However, there are systems that have employed RoCoF for underfrequency load shedding and others that are investigating RoCoF as a potential option 16. Most notable, for the current investigation, is the Tasmanian power system. The Tasmanian system provides an ideal case study as their grid is similar to the New Zealand grid. They have been using RoCoF in their AUFLS system since the late 1980 s 17. The literature review did reveal that Transpower considered the use of RoCoF elements in the 1990 s. However, the reasons the relays were not utilised could not be found [3]. 16 Other systems investigating or using RoCoF for under-frequency load shedding include Sri Lanka, Qatar, and regions of India. 17 The information summarised comes from D. E. Clarke, Tasmanian experience with the use of df/dt triggering of UFLSS. See reference [4] for further information.

15 System Operator Report: Automatic Under-Frequency Load Shedding (AUFLS) RoCoF Testing Page 15 of Use of RoCoF in Tasmania Tasmania began developing an under-frequency load shedding scheme following a system wide black out in 1979 caused by the mal-operation of protection equipment. They sought to develop a scheme that would protect their system from a collapse following what they classified as reasonably foreseeable contingencies while not operating for events covered by reserves. In 1986, they formally defined reasonable foreseeable contingencies, which included system disturbances up to 60% of total system load and events that result in system splits creating small sub-systems with 50% island load to generation deficit. The initial AUFLS scheme designed to provide this level of protection included the use of RoCoF elements. The initial RoCoF settings were identified by: investigating the worst case RoCoF following events covered by reserves to ensure discrimination; and identifying a RoCoF measurement initiation frequency (similar to a frequency guard) to reduce the possibility of mal-operation. identifying back-up under-frequency settings to cater for slow decaying disturbances. The resulting design of the AUFLS scheme included six load shedding steps 18. RoCoF elements were utilised on the first two steps and were connected to major industrial load, as their load profiles are relatively constant. In 2003, the AUFLS scheme was reviewed for changes to the Tasmanian frequency standards to incorporate modern gas turbines and wind generation. The operating range of AUFLS was reduced from 2.2 Hz to 1.5 Hz. Tasmania noted that the reduction was only achieved with acceptable discrimination and acceptable over-shoot by using rate of change of frequency settings to operate early and slow down the frequency decline for large disturbances. Further re-design of the AUFLS system took place with the installation of the Basslink (HVDC connection to Australia). The review found that the Tasmanian power system is prone to oscillations during large disturbances and the use of average RoCoF was required. This concept will be discussed in more detail later in this report. Tasmania reported the benefit of using of RoCoF elements together with higher under-frequency settings improved the minimum system frequency reached for AUFLS events and resulted in at least one less block of load shed. The review of the RoCoF scheme in 2008 reported that there have been no recorded instances of the RoCoF elements of relays not operating correctly [4]. 18 The load shedding steps are not a set percentage of load as in New Zealand. The size of load shed in each step is different and appears to be organised around load type (major industrial and retail load) with the majority of load provided by major industrials.

16 System Operator Report: Automatic Under-Frequency Load Shedding (AUFLS) RoCoF Testing Page 16 of 43 4 Calculating the rate of change of frequency The current AUFLS scheme uses relays to calculate the system frequency in real time and then shed load when a set frequency level has been met. The system frequency is not a measurable value; it must be calculated from another waveform (generally voltage). The rate of change of frequency is also a calculated value. The RoCoF calculation methods are important as they can impact the reliability of the relays. To better understand the calculation limitations of RoCoF relays the methods were investigated and necessary tests were constructed to ensure their reliability Frequency Calculation There are many methods to calculate the system frequency. A more detailed analysis of calculation methods exists in Appendix C to illustrate the concept. The following section will provide a very brief summary of a common method. Most relays calculate the system frequency from the voltage waveform. The method each relay uses to calculate frequency is often not specified. A simple method is called the zero crossing method. The relays measure the time between the points where the voltage sine wave transitions through the zero value, as displayed in Figure 2. Figure 2 Example of zero crossing calculation method No matter what method a relays uses to calculate the frequency it is important to note that frequency is a calculated value. Every calculation method will have a degree of inaccuracy associated with it.

17 System Operator Report: Automatic Under-Frequency Load Shedding (AUFLS) RoCoF Testing Page 17 of Rate-of-Change-of-Frequency Calculation The rate of change of frequency is generally calculated from the frequency value rather than directly from the original voltage waveform. Similar to the frequency measurement, relay manufacturers do not make it transparent how they calculate the rate of change of frequency values. The simplest form of RoCoF calculation is based on two successive frequency calculations with a set time difference between the frequency calculations. The difference between frequency calculations is divided by the time difference to calculate the RoCoF. However, relay manufacturers do not make it clear how they calculate RoCoF and it is unlikely any would rely on the crude method describe above. Frequency is calculated quantity so the RoCoF is derived from a calculated quantity. Further, it is well known that differentiation is generally an unstable operation, especially in the presence of noise. This can make the calculation of RoCoF unstable and filtering is required for it to work well. The limitation of the calculation method is outlined in greater detail in Section 7.1.

18 System Operator Report: Automatic Under-Frequency Load Shedding (AUFLS) RoCoF Testing Page 18 of 43 5 RoCoF Bench Testing - Methodology The objective of the RoCoF relay bench testing is to evaluate the technical requirements for the reliable use of RoCoF relays. The testing methodology was developed to verify the following performance criteria: the relays are capable of operating the required logic; the RoCoF calculation is stable during a range of expected system disturbances; the variety of RoCoF calculation methods used deliver a predictable response within a range of uniformity when subjected to the same inputs; the RoCoF calculation responds within a sufficient timeframe to frequency changes; and the RoCoF calculation is accurate when subjected to a wide variety of input waveforms. The testing work is not seeking to approve the use of any particular relay but to better understand what requirements need to be specified to ensure reliable operation of RoCoF relays. This section outlines the steps taken and assumptions made to complete the RoCoF relay testing. The bench testing included the following steps: 1. Establish initial RoCoF performance requirements to create a sample relay selection 2. Develop test equipment arrangement and process 3. Design test cases objectives and generate required files 4. Carry out bench testing 5. Review results and identify next steps 6. Complete additional testing iterations as required 5.1 Initial RoCoF relay performance requirements The first step was to propose a set of initial RoCoF relay performance requirements that would enable selection of sample RoCoF relays to be used in the bench testing process. The investigation included an analysis of additional user requirements (i.e. Transpower and lines companies). The initial performance requirements are set out in Appendix A. A selection of the requirements for the sample relays is provided below in Table 2.

19 System Operator Report: Automatic Under-Frequency Load Shedding (AUFLS) RoCoF Testing Page 19 of 43 Table 2 Selection of the RoCoF Relay Performance Requirements Feature Minimum requirement Purpose Fixed f + t (81U) element df/dt (81R) element Setting range 45.0 to 50.0 Hz At least 4 stages or levels or elements Setting range 0.0 to 2.0 Hz/s At least 4 stages or levels or elements Traditional RoCoF Trip logic Boolean internal, programmable To enable frequency guard To combine 81x with 27 elements Operating range 45 to 55 Hz Accuracy (81U) ±0.05Hz of reference frequency* Precision of trip settings Accuracy (81R) ±0.1mHz/s of reference frequency* Precision of trip settings Repeatability Trip within ±20.0 ms over three tests using the same input waveforms and trip settings Predictability of trip From the list of relay requirements a survey was done of major manufacturers. The investigation resulted in the final sample set of relays provided in Table 3. Table 3 Relays used in RoCoF Bench Testing Manufacturer Relay model Description Target markets* SEL 751 Multifunctional feeder protection Transmission / Distribution SEL PS Multifunctional feeder protection Transmission / Distribution GE SR760 Multifunctional feeder protection Transmission / Distribution Siemens 7SJ64 Multifunctional feeder protection Transmission Siemens 7SJ80 Feeder protection Distribution ABB RED670 Line differential protection Transmission ABB REU615 Voltage protection Distribution These are all multifunctional numerical relays that can be programmed to provide a range of different protection functions in addition to frequency protection. Two additional relays were also provided for testing, namely a RMS 2H34-S frequency relay and an Alstom P145 feeder protection relay. The RMS relay was found to have a bug that prevented settings from being changed reliably, and the Alstom relay arrived half way through the second round of testing. Neither of these relays were tested as a result. However, the Alstom relay will likely be tested at a later date since, unlike all the other relays tested, its averaging window can be explicitly set.

20 System Operator Report: Automatic Under-Frequency Load Shedding (AUFLS) RoCoF Testing Page 20 of Test equipment arrangement and process To identify the requirements for reliable use of RoCoF relays on the New Zealand power system, a testing process was developed to verify each element of the performance criteria. The testing simulated specific system events and conditions through the relays and observed the effect on the relay s RoCoF calculation and behaviour. Most of the relays tested do not provide a readily accessible output of their calculated frequency or rate of change of frequency. Therefore, determining these calculated values requires some reverse engineering. The relay s trip signal was used as it is a readily accessible output and directly relates to the function of the relay. The point at which the relay signals to trip can be referenced against the voltage waveform to better understand the calculated RoCoF value. The testing work involved injecting voltage waveforms into the sample relays to allow the relay to calculate the RoCoF and then observing if the relay responds with a trip signal. Figure 3 provides an overview of the arrangement used to test the relays. Figure 3- Relay bench testing configuration The Omicron test set is used to convert data files into actual voltage waveforms injected into the relay.

21 System Operator Report: Automatic Under-Frequency Load Shedding (AUFLS) RoCoF Testing Page 21 of 43 A typical test will involve: Creating a voltage waveform to inject into the test relay Loading the selected voltage waveform into the Omicron Programming the test relay with the desired settings Playing the selected voltage waveform into the test relay Recording when the test relay operates Running the same voltage waveform through Matlab to calculate frequency and RoCoF using a known algorithm Comparing the operating point with the voltage waveform and Matlab calculated frequency and RoCoF on the same time axis. 5.3 Test Cases Several test cases were created specifically to understand each aspect of the performance criteria (logic, stability, uniformity, response time, and accuracy). Each test case required a voltage waveform and, depending on the objective of the test, a waveform was either generated or taken from real system events. A sample waveform used to test the logic of the relays can be seen in Figure 4. Figure 4 - Example waveform used to test the relay logic elements One of the areas of greatest concerned outlined by the Distribution Companies was around the stability of the relay. Several test cases were created to verify the stability of the relay during system disturbances which included; a noisy bus waveform (Glenbrook); HVDC Commutation failures; loss of phase; no volts; abnormal frequencies; and islanding. The literature review gave particular focus to the effect of power system oscillations effect on the accuracy of the RoCoF calculation. Multiple test cases were developed to further investigate this potential issue.

22 System Operator Report: Automatic Under-Frequency Load Shedding (AUFLS) RoCoF Testing Page 22 of Settings Used The testing work utilised the AUFLS settings proposed in the last stage of the AUFLS review 20 for the majority of test conducted 21. The settings are in Table 4. Table 4 - Proposed NI AUFLS operation criterion Accelerated element Under frequency element 1 Under frequency element 2 Block Size Df/dt set f guard T d3 f set1 T d1 f set2 T d2 Block 1 N/A N/A N/A 47.8 Hz 0.3s N/A N/A 8% Block 2 N/A N/A N/A 47.5 Hz 0.3s 47.8 Hz 15s 8% Block Hz/s 48.5 Hz 0.4s 47.3 Hz 0.3s 47.5 Hz 15s 8% Block Hz/s 48.8 Hz 0.4s 47.3 Hz 0.3s 47.5 Hz 15s 8% The role of the frequency guard The proposed AUFLS scheme utilises different elements to achieve the designed logic output. One of the key elements is the frequency guard. The frequency guard prevents the relay from tripping on the calculated RoCoF value alone. The relay will only operate if the RoCoF is greater than or equal to the set level AND the frequency measured is less than or equal a set level for specified duration. A high level logic diagram is displayed in Figure 5. RoCoF Calculation AND Guard frequency and time delay OR Trip Backup frequency and time delay Figure 5 High Level Logic Overview The frequency guard helps prevent spurious operation from frequency fluctuations not resulting from large system disturbances. Therefore, the frequency guard helps to minimize the risk of mal-operation. 20 See Appendix C of the AUFLS Stage II review. 21 It is noted that some test utilised other settings that were being investigated for possible scheme improvements prior to Pole 3 commissioning.

23 System Operator Report: Automatic Under-Frequency Load Shedding (AUFLS) RoCoF Testing Page 23 of 43 6 Bench Testing Results The purpose of the testing work was to provide greater visibility of the strengths and weakness of RoCoF relays for the purpose of triggering AUFLS load. Seven numerical relays fitted with RoCoF elements were tested to help determine their suitability. The tests devised focussed on understanding the logic capability, stability, uniformity, response time, and accuracy of the RoCoF relays. This section highlights the observations made from the test results and makes recommendations to ensure the installed RoCoF relays deliver the desired outcomes. The detailed tests and results are in Section 6 of Appendix B; the main points are summarised below. 6.1 Bench Testing Results Summary Logic The proposed AUFLS scheme using RoCoF requires a level of logic including a frequency guard and backup under-frequency settings. The testing aimed to ensure the relays are capable of operating the required logic. All of the relays tested had some form of logic programming that allowed users to implement various combinations of RoCoF, frequency, and time delay elements to achieve a desired tripping response. Each manufacturer has a different way of arranging and setting their logic elements and care is required to ensure the desired result is achieved when programming and setting the various relays. This was illustrated in test 3 where the two of the relays exhibited an unexpected response to a simple logic test. It is suspected that this could be addressed by adjusting the logic arrangements used in these relays. It is recommended that the Code specifies a test outcome to verify that the logic elements of the installed relays meet the performance requirements Stability A primary aim of the bench testing was to ensure the RoCoF relays remain stable during a range of system disturbances that would not normally result in AUFLS operation. The tests results indicate that the sample RoCoF relays were stable under all of the disturbances to which they were subjected. This indicates that the RoCoF algorithms are designed to reject noise and fast transient events which will help to avoid undesired operation of the RoCoF. The guard frequency used in these tests is likely to play a major role in preventing the undesired operation of RoCoF. The above stability tests were carried out using the RoCoF settings that were available at the time of testing. The results of these tests may change if different settings are used (i.e. if the frequency guard is set to a higher frequency). It is recommended that RoCoF relays are stable for use. It is suggested that Industry reviews the tests results and provides feedback on the level of confidence in the stability of the relays.

24 System Operator Report: Automatic Under-Frequency Load Shedding (AUFLS) RoCoF Testing Page 24 of Uniformity and Response Time A level of uniformity is required to ensure the installed AUFLS system achieves the desired levels of operation with a range of different relays. The tests demonstrated that the sampled RoCoF relays do not behave uniformly when subjected to identical inputs. This lack of uniformity was reflected in all of the tests, and it indicates that different relay manufacturers use different algorithms for calculating RoCoF. The nature of the algorithms used in these relays means that it will always take a finite amount of time to detect a change in frequency because it involves calculating over a set number of cycles. The response times of the relays are dependent on the RoCoF settings. The tests demonstrate that the closer the rate of change is to the set point, the longer the response time, the greater the difference, and the faster the relay operates. Depending on the RoCoF settings used, the relays use different lengths of time (numbers of cycles) for the RoCoF calculations, as shown in Figure 6. Figure 6 Example of response time The lack of uniformity means relays operate over a range of different response times during the frequency excursions. This results in AUFLS load tripping at different times even though all relays have identical settings. This is not necessarily a problem provided that relays comply within a maximum allowed response time. It is recommended that the Code specifies the maximum response time allowed for RoCoF relay operation.

25 System Operator Report: Automatic Under-Frequency Load Shedding (AUFLS) RoCoF Testing Page 25 of Accuracy The ability to rely on RoCoF relays for load shedding requires the relay to be able to accurately calculate the rate of system frequency decay when subjected to a variety of grid conditions. The consequences of inaccuracy could result in the relay mal-operating by tripping inappropriately or not tripping when needed. As stated previously, frequency is a calculated value from which the RoCoF is derived. The RoCoF is derived (differentiated) from a calculated value. It was assumed that most relays would simply calculate the RoCoF from two successive frequency calculations with a set time difference between the frequency calculations. However, the testing revealed the calculation method from most RoCoF relays are more complicated and it is difficult deduce the exact method used. The relays performed within the specified accuracy margins (±0.05 Hz/s) for clean frequency decays. However, the test results did verify that RoCoF relays are susceptible to calculation inaccuracy due to system oscillations. The impact of power system oscillations is dependent on the relays settings used.