Project Report. Project #: TIC706.1-A Islanding Risk of Synchronous Generator Based Distributed Generation Systems. Draft to be approved.

Transcription

1 Project Report Project #: TIC76.1-A Islanding Risk of Synchronous Generator Based Distributed Generation Systems Draft to be approved Submitted To The CANMET Energy Technology Centre Prepared By Wilsun Xu, Ph.D., P.Eng. Dept. of Electrical and Computer Engineering University of Alberta Edmonton, Alberta Canada Walmir Freitas, Ph.D. Dept. of Electrical Energy Systems State University of Campinas Campinas, Sao Paulo Brazil March 31, 28

2 DISCLAMER This report is distributed for informational purposes and does not necessarily reflect the views of the Government of Canada nor constitute and endorsement of any commercial product or person. Neither Canada nor its ministers, officers, employees or agents makes any warranty in respect to this report or assumes any liability arising out of this report. ACKNOWLEDGMENT Financial Support for this collaborative research project was provided in part by Natural Resources Canada through the Technology and Innovation Program as part of the Climate Action Plan for Canada. The authors wish to thank Mr. Sylvain Martel, the project manager, for his support and patience during the course of this project. 2

3 Table of Contents Chapter 1: Introduction...7 Chapter 2: Anti-Islanding Protection for Distributed Generators Electric Island Formed by Distributed Generators Anti-islanding Protection Options for Synchronous DG Survey of Frequency-Based and Voltage-Based Relays...12 Chapter 3: Operating Principles of Frequency and Voltage Based Relays Frequency and Voltage Relays Rate of Change of Frequency Relay Vector Surge Relay...15 Chapter 4: Performance Characteristics of Frequency and Voltage Based Relays The Concept of Anti-islanding Performance Curves Performance Curves of the Frequency-Based Relays Performance Similarities of the Frequency and Vector Surge Relays Key Factors Affecting the Relay Performance Equations to Predict Relay Performance Limitations of Frequency (and Vector Surge) Relay Limitation of the ROCOF Relay Performance Curves of the Voltage Relay Key Factors Affecting the Relay Performance Limitations of Voltage Relay...3 Chapter 5: Non-Detection Zones of Combined Frequency and Voltage relays The Concept of 2D Non-Detection Zone in PQ Plane Non-detection Zone of the Frequency Relay Non-detection Zone of the Voltage Relay Non-Detection Zone of Combined Frequency and Voltage Relays...35 Chapter 6: The Risk of Island Formation The Concept of Islanding Risk Survey of Islanding Risk Research Strategy of the Proposed Work...43 Chapter 7: Assessing the Risk of Islanding The Basic Idea General Procedure for Risk Assessment Modelling of the Non-Detection Zones...53 Chapter 8: Characteristics of Islanding Risks Description of the Study Case Validation Study Results Sensitivity Study Results Types of Relays Relay Settings Generation Level

4 8.3.4 Generator Size Required Detection Time Time Delay Relay Operation Type of Loads The Worst Case Scenario Summary and Recommended Procedures for Risk Assessment...84 Chapter 9: Conclusions...86 Chapter 1: References...89 Appendix A: Methods FOR DeterminING Relay Performance Curves...91 A.1 Simulation System...91 A.2 Relay Models...91 A.3 Simulation and Analytical Studies...93 A.4 Simulation Study of 2D Non-Detection Zone...94 Appendix B. Determination of Non-detection zones...96 Appendix C: Risk assessment Using Dynamic Simulations...1 Appendix D: Risk Assessment for systems with Constant Impedance Loads...12 Appendix E: Example of Time Domain Simulation Results...16 E.1. Simulation Results

5 Executive Summary This report presents the anti-islanding performance characteristics of both frequencybased and voltage-based relays when they are applied to synchronous distributed generators. It also provides methods for determining the risk levels of undetected islanding formation when these relays are used. Anti-islanding protection has become an important requirement for distributed generation applications. A common industry practice to address the anti-islanding requirement is to determine if simple and low cost frequency-based and voltage-based relays are sufficient for a DG installation proposal. If it is not sufficient, more costly and advanced protection schemes are then considered. However, there have been no methods available for utility engineers to conduct the assessment on the applicability of the frequency/voltage based relays for synchronous distributed generators. This project is conducted to fill in this knowledge gap. The goal is to equip Canadian DG industry with techniques and tools for synchronous DG interconnection studies, thereby reducing the technical barrier for DG installation. The main results of this project are summarized here. The anti-islanding performance of individual relays can be characterized using detectiontime versus power-mismatch (or power-imbalance) curves. Detection time is the time needed for a given relay to conclude that an island has formed. Power-mismatch is the power deficit or surplus of the island at the instant of island formation. Research results obtained by this project show that the frequency-based relays are sensitive to active power imbalance and the voltage-based relays are sensitive to reactive power imbalance. If the power imbalance is less than 1% to 2%, the relays may not be able to detect the island formation within typical required time. When both frequency-based relays and voltage-based relays are applied together, a nondetection zone can be established. A non-detection zone (NDZ) is the active and reactive power mismatch levels below which the combined relay schemes cannot detect island formation within an acceptable time delay. The NDZ is highly influenced by the load characteristics in the island. Study results show that the following equation can be used to estimate the typical size of the non-detection zone 8% < ΔP' < 8% 36% < ΔQ' < 16% where ΔP' =.866ΔP. 5ΔQ and ΔQ' =.5ΔP ΔQ, ΔP and ΔQ are the power mismatch levels of the island. In this report, the risk of islanding formation is defined as the probability of a DGcontaining distribution system entering the non-detection zones created by the frequency 5

6 and voltage based anti-islanding devices. Due to the wide variety of synchronous DG interconnection scenarios and distribution feeder configurations, it is not possible to provide typical risk values of island formation. This project therefore develops a practical risk assessment method for use by utility engineers. The method uses relay NDZ characteristics, feeder configuration, load profile and DG generation schedule as input and calculates the probability of the DG system entering the non-detection zones over a specified period such as one day or one week. Case studies using the proposed method revealed a few general characteristics of islanding risks associated with the frequency and voltage based relays. When increasing the required detection time from 5 ms to 1 ms or 15 ms, these relays can be quite effective in reducing the risk level of island formation. Time delay settings of the relays can increase the risk level if short detection time is required. The load characteristics can influence the risk of islanding significantly. The most conservative situations, which lead to higher risk levels, are related to both constant impedance and constant power loads. As the load characteristics of a distribution feeder is hardly known, worst cases that involve both types of load models should be used in the risk assessment. 6

7 CHAPTER 1: INTRODUCTION Distributed generation (DG) has recently gained a lot of momentum due to market deregulation and environmental concerns. An important requirement to interconnect a DG to power distribution systems is the capability of the generator to detect island conditions. Islanding occurs when a portion of the distribution system becomes electrically isolated from the remainder of the power system, yet continues to be energized by distributed generators. Failure to trip islanded generators can lead to a number of problems to the generators and the connected loads. The current industry practice is to disconnect all distributed generators immediately after the occurrence of islands. Typically, a distributed generator should be disconnected within 1 to 2 ms after loss of main supply [1,2,3]. To achieve such a goal, each distributed generator must be equipped with an islanding detection device, which is also called anti-islanding device. The most common devices used for this purpose are the under/over frequency relays, under/over voltage relays and their variations. These relays have very low cost and are widely available. They are the first choice for antiislanding protection. The frequency-based relays operate on the principle that if the generation and load have a large mismatch in an island the frequency of the island will drift. One can therefore detect the islanding condition by checking the amount and rate of frequency change. The voltage relay is based on the understanding that the voltage in an island will also drift because of reactive power imbalance. Unfortunately, these relays are not 1 percent reliable due to their inherent limitations. If the active/reactive power imbalance in an island is small, it will take some time for the islanded system to exhibit detectable frequency or voltage change. As a result, the relays will not be able to provide anti-islanding protection in a timely manner. The corresponding system operating conditions are called non-detection zones of the relays. In view of the significant cost advantages of the frequency and voltage relays, it has becomes imperative for utility companies and DG owners to understand the characteristics of the non-detection zones and the associated risks. The information will greatly facilitate the selection of DG protection schemes and has the potential to achieve significant cost-savings for the DG owners. The objective of this report is to present our research results on the anti-islanding performance characteristics of both frequency-based and voltage-based relays when they are applied to synchronous distributed generators. Over the past 5 to 1 years, non-detection zone research has been concentrated on inverter-based DGs due to the popularity of photovoltaic power supplies. For example, the International Energy Agency (IEA) sponsored a systematic investigation on the subject [4,5,6]. The results have clarified a lot of concerns on the risks associated with inverter-based DGs back feeding an island. In comparison, no similar work has been done for the synchronous machine based DGs. In fact, due to its relatively large size and 7

8 lack of flexibility in output control, the synchronous DGs have become the most challenging type to establish adequate anti-islanding protection [7]. This project is conducted to address the non-detection zone issues associated with the synchronous DGs. The goal is to equip Canadian DG industry with sufficient information so that it can assess the applicability of the frequency/voltage relays and associated risks for various synchronous DG interconnection projects. This report is organized in three parts: The first part, consisting of Chapters 2 and 3, discusses the nature and requirements of antiislanding protection. The frequency and voltage relays are introduced and their operating principles are explained. The second part, consisting of Chapters 4 and 5, presents the anti-islanding protection characteristics of various frequency-based and voltage-based relays. Chapter 4 focuses on the detection-time versus power mismatch curves of the relays. These curves show how long it will take for a relay to operate for a given power mismatch condition. Chapter 5 presents the non-detection zones of the relays. The third part, consisting of Chapters 6, 7 and 8, investigates the risks caused by the nondetection zones. The risk of islanding formulation is essentially the probability of a DGcontaining distribution system entering the non-detection zones created by the anti-islanding devices. A practical risk assessment method is presented to determine if the frequency and voltage relays can be used with confidence. 8

9 CHAPTER 2: ANTI-ISLANDING PROTECTION FOR DISTRIBUTED GENERATORS Anti-islanding capability is an important requirement for distributed generators. It refers to the capability of a distributed generator to detect if it operates in an islanded system and to disconnect itself from the system in a timely fashion. This chapter reviews the background information on the DG anti-islanding protection. Two most common anti-islanding options for synchronous machine based distributed generators, frequency-based and voltage-based relays, are discussed. 2.1 Electric Island Formed by Distributed Generators A typical power distribution system in North America is shown in Figure 2.1. The substation steps down transmission voltage into distribution voltage and is the sending end of several distribution feeders. One of the feeders is shown in detail. There are many customer connection points in the feeder. Large distributed generators are typically connected to the primary feeders (DG1 and DG2). These are typically synchronous and induction generators at present. Small distributed generators such as inverter based PV systems are connected to the low voltage secondary feeders (DG3). Substation 1 13kV 12V F DG3 A 25kV B D C Island DG1 DG2 Figure 2.1: Typical distribution system with distributed generators. An island situation occurs, for example, when recloser C opens. DG1 will feed into the resultant island in this case. The most common cause for a recloser to open is a fault in the 9

10 downstream of the recloser. A recloser is designed to open and re-close two to three times within a few seconds. The intention is to re-connect the downstream system automatically if the fault clears by itself. In this way, temporary faults will not result in the loss of downstream customers. An island situation could also happen when the fuse at point F melts. In this case, the inverter based DG will feed the local loads, forming a small islanded power system. The island is an unregulated power system. Its behaviour is unpredictable due to the power mismatch between the load and generation and the lack of voltage and frequency control. The main concerns associated with such islanded systems are: The voltage and frequency provided to the customers in the islanded system can vary significantly if the distributed generators do not provide regulation of voltage and frequency and do not have protective relaying to limit voltage and frequency excursions. Since the supply utility is no longer controlling the voltage and frequency, the islanding situation could result in damages to customer equipment. Although the supply utility has no control over the situation, it may still be found liable for the consequences. Islanding may create a hazard for utility line-workers or the public by causing a line to remain energized that may be assumed to be disconnected from all energy sources. The distributed generators in the island could be damaged when the island is reconnected to the supply system. This is because the generators are likely not in synchronism with the system at the instant of reconnection. Such out-of-phase reclosing can inject a large current to the generators. It may also result in re-tripping in the supply system. Islanding may interfere with the manual or automatic restoration of normal service for the neighboring customers. The current industry practice is to disconnect all DGs immediately so that the entire feeder becomes de-energized [1,2,3]. It prevents equipment damage and eliminates safety hazards. This calls for a reliable and speedy detection of islanding conditions. The basic requirements for a successful detection are: The scheme should work for any possible formations of islands. Note that there could be multiple switchers, reclosers and fuses between a distributed generator and the supply substation. Opening of any one of the devices will form an island. Since each island formation can have different mixture of loads and distributed generators, the behaviour of each island can be quite different. A reliable anti-islanding scheme must work for all possible islanding scenarios. The scheme should detect islanding conditions within the required time frame. The main constraint here is to prevent out-of-phase reclosing of the distributed generators. A 1

11 recloser is typically programmed to reenergize its downstream system after about.5 to 1 second delay. Ideally, the anti-islanding scheme must trip its DG before the reclosing takes place. The above goal is achieved by equipping each DG with an anti-islanding protection capability each DG must be able to detect if it is islanded and to disconnect itself automatically from the system when islanding occurs. In response to the requirements, many anti-islanding techniques have been proposed and a number of them have been implemented in actual DG projects [4] or incorporated into some of the DG controller. Reference [7] provides a comprehensive review of various anti-islanding techniques. The work of reference [7] further revealed that anti-islanding protection for synchronous distributed generators is the primary concern for Canadian DG industry and supply utilities. 2.2 Anti-islanding Protection Options for Synchronous DG Synchronous distributed generators use synchronous machine as the energy converter. The generators are typically connected to the primary feeder. Their sizes can go as high as 3MW. Synchronous generators are highly capable of sustaining an island. Due to its large power rating, options are limited to control the generators for the purpose of facilitating islanding detection. As a result, anti-islanding protection for synchronous generators has emerged as one of the most challenging tasks facing the DG industry. Methods available for synchronous DG anti-islanding protection can be broadly classified into two types according to their working principles. The first type consists of communicationbased schemes. It uses telecommunication means to alert and trip DGs when islands are formed. The transfer trip scheme well known to utility companies belongs to this type. The telecommbased schemes can be quite expensive and could render a DG project economically unattractive. The second type is to rely on the voltage and current signals available at the DG site. An islanding condition is detected if indices derived from the signals exceed certain thresholds. Frequency-based and voltage-based anti-islanding relays are representative examples of such schemes. The relays trip a DG if the frequency or voltage measured at the DG location drifts outside pre-established safe operation boundaries. Due to their low cost and simplicity, frequency-based and voltage-based relays are the first choice by the DG owners to provide antiislanding protection for synchronous DGs. The frequency-based schemes are the most widely used scheme for anti-islanding protection involving synchronous generators. It is known if the generation and load have a large mismatch in a power system, the frequency of the system will change. In view of the fact that the frequency is constant when the feeder is connected to the transmission system, it is possible to detect the islanding condition by checking the amount and rate of frequency change. Several 11

12 commercial products based on this idea have been developed and are available for use at present. They can be classified into the following three types [8]: Over/under frequency relay, which is called frequency relay in this report; Rate of change of frequency relay, which is called ROCOF relay in this report; and Vector surge (jump or shift) relay, which is called VSR relay in this report. There is only one type of voltage-based relays commercially available for anti-islanding protection. It is the over/under voltage relay and is called voltage relay in this report. The relay operates on the principle of reactive power mismatch in an island. Excessive reactive power will drive up the system voltage and deficit reactive power will result in voltage decline. By determining the level of voltage at the DG terminal, it is possible to detect islanding conditions that cannot be detected by frequency-based relays. Note that a voltage relay is needed for other protection purposes in a DG installation. For example, it is used to prevent over-voltage stress to the DG unit. A voltage relay is, therefore, always available in a DG installation and can be utilised to support islanding detection at no extra cost. As the frequency-based and voltage-based relays are the first choice for anti-islanding protection, it becomes important to understand their performance characteristics and limitations. Such information will help DG owners and supply utilities to determine if the relays can perform the required anti-islanding task reliably. Other anti-islanding options may be considered only after it is determined that the frequency and voltage-based relays are unable to meet the requirements. One of the main objectives of this project is to provide methods and data to help DG owners and utility engineers to assess the applicability of the common frequency-based and voltage-based relays for specific synchronous DG installations. 2.3 Survey of Frequency-Based and Voltage-Based Relays The frequency and voltage relays are common relays found in various power system protection applications. As a result, many relay manufacturers have the products. Because of their commonality, this project only surveyed products that are specifically designed for DG antiislanding protection. Table 2.1 summarises the findings. It is important to note that this is not an exhaustive survey, but rather a sample of what is typically available in the market. It can be seen that anti-islanding products based on the principle of frequency or voltage variation detection are widely available. This is an indication that frequency- and voltage-based relays are the common choice for anti-islanding protection. The survey found that the relay prices vary from $1 to $5. The high priced relays include other DG protection functions. Although the products are widely available, it is not clear how reliable they are in providing anti- 12

13 islanding protection and what are the performance characteristics of the different relays. Chapter 4 will provide answers to these and other related questions. Table 2.1: Sample manufacturers and products of frequency-based anti-islanding relays. Manufacturer Product Name Principle SEL (USA) DG interconnection Relay: SEL-547 Under/over frequency Under/over voltage Basler Electric (USA) BE1-IPS1 Intertie Protection System Under/over frequency Under/over voltage ROCOF Cooper Power Systems (USA) Woodward (USA) UM3SV Vector jump/islanding relay VSR Under/over frequency Under/over voltage MFR-11/G59 Multi Function Mains Protection ROCOF VSR Meidensha Corp. (Japan) Loss of Mains relay ROCOF Sepam (UK) Sepam 1+ ROCOF Crompton Instruments (UK) 256-ROCL Vector Shift and ROCOF relay VSR ROCOF Megacon (UK) KCG592 Loss of Mains Relay ROCOF VSR ABB Oy (Finland) SPAF 14C Frequency Relay Under/over frequency ROCOF DEIF A/S (Denmark) 1) G59 Protection relay package 2) LMR-122D Loss of Mains Relay VSR ROCOF Over/under frequency Over/under voltage 13

14 CHAPTER 3: OPERATING PRINCIPLES OF FREQUENCY AND VOLTAGE BASED RELAYS This chapter presents the principles of the frequency and voltage based relays. Differences and similarities among the three types of frequency-based relays are discussed. The operating principles form the basis to determine the performance characteristics of the relays for anti-islanding applications. 3.1 Frequency and Voltage Relays Measuring frequency and voltage is one of the basic functions performed by modern microprocessor relays. To measure the frequency, a voltage signal supplied from a PT (potential transformer) is first filtered using a band-pass filter. This operation reduces the impact of waveform distortion on the measurement accuracy. The frequency is determined by measuring the time between the zero crossings of the filtered waveform [9]. Each cycle of the waveform yield one frequency value. Typically, the relay works on a moving average of the per-cycle frequencies. The number of cycles used for the moving average calculation varies from 3 to 3 cycles and is selectable by users. When the calculation is based on three cycles, the measurement response time will be short and, consequently, the trip time as well. On the other hand, when thirty cycles are used the response time will be long, but the effect of the noise possibly occurring in the signal will be small. The RMS magnitude of the voltage signal is measured using the following equation [9]: V N 2 rms = v i N i= 1 1 (1) where N is number of samples per cycle and v i is the sample value. As a result, each cycle yields one voltage magnitude value. A moving average is also used to produce a voltage value that is used to compare with a user-specified threshold and thereby activate the relay. Both the frequency and voltage relays have at least one time delay setting in addition to magnitude threshold settings. A delayed activation is often needed to avoid false trips caused by system conditions outside the protective scope of the relays. As an example, typical frequency relay settings to protect a generator from over speeding are 61Hz with an 1. second delay and 63Hz with a.3 second delay. For the DG applications, the DG interconnection guide of Alberta specifies 59.5 Hz as the under frequency threshold and 6.5Hz as the over-frequency threshold. A DG shall be tripped within.5 seconds if either of the thresholds is exceeded. 14

15 3.2 Rate of Change of Frequency Relay The ROCOF relay calculates the rate of change of frequency (df/dt) using two successive moving-average frequency values. The moving average is based on, for example, 3 cycles of voltage waveform. This window size is typically built into the relay and it cannot be easily changed by users. The relay activates when the rate of change of frequency is higher than a userspecified threshold and after a user-specified time delay. Typical ROCOF relay settings for 6 Hz systems are between.5 Hz/s and 2.5 Hz/s. Another important characteristic available in the ROCOF relay is a blocking function according to minimum generator terminal voltage. If the terminal voltage drops below an adjustable level V min, the trip signal from the ROCOF relay is blocked. This is to avoid, for example, the actuation of the ROCOF relay during generator startup or short-circuit faults. 3.3 Vector Surge Relay The principle of the VSR relay can be understood from Figure 3.1 where a synchronous generator equipped with the relay interconnects to a distribution network. There is a voltage drop ΔV between the terminal voltage V T and the generator internal voltage E I due to the generator current I SG passing through the generator reactance X d. Consequently, a displacement angle δ exists between the terminal voltage and the generator internal voltage. The phasor diagram is shown in Figure 3.1(a). If the circuit breaker CB opens, due to a fault for example, the system composed by the generator and the load L becomes islanded. The synchronous machine begins to feed a larger (or smaller) load, which makes it to decelerate (or accelerate). Consequently, the angular difference between V T and E I is suddenly increased (or decreased) and the terminal voltage phasor changes its direction, as shown in Figure 3.1(b). Viewing such a phenomenon in the time domain, we can notice that the instantaneous value of the terminal voltage jumps to another value and the phase position changes, as depicted in Figure 3.1, where the point A indicates the islanding instant. Additionally, the frequency of the terminal voltage also changes. This behaviour of the terminal voltage is called vector surge or vector shift. The vector surge relay is based on such a phenomenon. E I ΔV X d I SG VSR CB I SYS power L V T grid ΔV ΔV E I V T E I V T Δδ δ (a) (b) (a) Network diagram. (b) Voltage phasor ( vector ) diagram. Figure 3.1: The phenomenon of vector surge or vector shift. 15

16 The vector surge relays available in the market measure the duration of an electrical cycle and start a new measurement at each positive-going zero crossing of the terminal voltage. The cycle whose duration is to be measured is compared with the previous cycle (called reference cycle). In an islanding situation, the cycle duration is either shorter or longer, depending on if there is excess or deficit of power in the islanded system. This variation of the cycle duration represents the variation of the terminal voltage angle Δδ. If the variation of the terminal voltage angle exceeds a pre-determined threshold α, a trip signal is immediately sent to the circuit breaker. Usually, vector surge relays allow this threshold to be adjusted in the range from 2 to 2 degrees. The vector surge relay also has a minimum terminal voltage triggered blocking function. If the terminal voltage drops below an adjustable level threshold V min, the trip signal from the vector surge relay is blocked. V(t) measured waveform A reference Δt Δθ new reference Δt Δθ time Figure 3.2: Measurement of the vector surge or shift. It can be shown that the shift, Δθ, is an indirect measurement of the waveform frequency. As a result, this type of relay is expected to have a performance characteristic similar to that of the frequency relay. 16

17 CHAPTER 4: PERFORMANCE CHARACTERISTICS OF FREQUENCY AND VOLTAGE BASED RELAYS The previous chapters have shown that at least four types of simple relays are available for DG anti-islanding protection. It becomes necessary to understand the performance characteristics of these relays. This chapter introduces the concept of performance curves for the relays and discusses the key factors that can affect the relay performance. Furthermore, the concept of a two-dimensional non-detection zone is presented for applications where both frequency and voltage-based relays are used to form a composite anti-islanding protection scheme. 4.1 The Concept of Anti-islanding Performance Curves The frequency-based relays work on the principle of active power imbalance in an island. A large power imbalance will cause fast deviation of frequency in the island and it will take less time to detect the islanding condition. An approach to evaluate the performance of frequencybased anti-islanding relays is, therefore, to understand the relationship between the tripping (or detection) time and power imbalance (or mismatch). This relationship can be represented with a detection-time versus power-mismatch curve as shown in Figure degrees 1 degrees 15 degrees Detection time (ms) Non-detection zone Active power imbalance (pu) Figure 4.1: Typical detection-time versus power-imbalance characteristics of frequency-based relays. The figure uses the vector surge relay as an example. There are three curves each representing a different setting of the VSR. The x-axis is the power mismatch level of the islanded system referred to the rated MVA of the DG. The y-axis is the time needed by the relay 17

18 to operate, since it takes time for the islanded system to exhibit detectable frequency variation. If it is required to trip the distributed generator within 3ms after islanding, one can draw a horizontal line of 3ms. The intersection of this line with the relay curve of 1 degrees gives 33% power mismatch level. If an islanded system has a power imbalance greater than 33%, it would take less than 3ms to detect the islanding condition. So the relay can be used with confidence. One the other hand, the relay will take longer than 3ms to operate if the power imbalance level is less than 33%. Consequently, the relay is not suitable for such cases. The 33% power mismatch level is called the critical power mismatch. The power-mismatch level below the critical power mismatch represents a (one-dimensional) non-detection zone of the relay. Similar performance curves can be developed for the voltage relay. Since voltage is sensitive to reactive power, the performance curves for voltage relays are represented as detection-time versus reactive-power-mismatch curves. Details are shown in Section 4.3. Methods to determine the performance curves for the relays are presented in Appendix A. 4.2 Performance Curves of the Frequency-Based Relays Sample detection-time versus power-mismatch curves for the frequency, ROCOF and vector surge relays are shown in Figure 4.2 for islands consisting of one synchronous generator. It can be seen that the VSR and frequency relays have similar performance. The ROCOF relay has the best performance since its non-detection zone is the smallest. The results also reveal that a non-detection zone of 1% to 3% power mismatch exists for all relay types. Reducing the trip threshold can reduce the non-detection zone. This approach, however, could make the relays too sensitive, resulting in more opportunities of nuisance trips. Because of this reason, the ROCOF relay is more prone to nuisance trips. Detection time (ms) Frequency relay: ± 1. Hz Frequency relay: ± 1.5 Hz Frequency relay: ± 2. Hz Vector surge relay: 6 degrees Vector surge relay: 9 degrees Vector surge relay: 12 degrees ROCOF relay:.5 Hz/s ROCOF relay: 1.5 Hz/s ROCOF relay: 2.5 Hz/s Active power imbalance (pu) Figure 4.2: Characteristics of three types of frequency-based anti-islanding relays. 18

19 4.2.1 Performance Similarities of the Frequency and Vector Surge Relays Figure 4.2 has revealed that the frequency relay and the vector surge relay have almost identical performance characteristics. Extensive research results show that this is not a coincidence. In this section, a comparison between the anti-islanding capability of the frequency and vector surge relays is carried out in detail. In a 6 Hz system, 1 Hz corresponds to 6 electrical degrees. Therefore, a frequency relay setting of.5 Hz can be compared with a VSR relay setting of 3 degrees and so on. The critical power imbalances for typical relay settings and for a required detection time of 3, 5 or 7 ms are presented in Table 4.1. In this table, the power mismatches are presented in percentage of the generator MVA rating; the frequency relay is referred as FR and the vector surge relay as VSR. It can be noted that both relays give very similar critical power imbalances. This further confirms the conclusion drawn earlier. The significance of this finding is the following: The vector surge relay does not offer additional advantages over the frequency relay for antiislanding protection. As a result, there is no need to install a dedicated VSR for anti-islanding application because a frequency relay, which is normally required for any DG installation, is as effective as the vector surge relay for anti-islanding application. The savings can be quite attractive for small distributed generators and the resultant protection system will be much simpler. Because the vector surge relay has almost the same performance characteristic as that of the frequency relay, the VSR relay will not be signalled out for separate analysis and discussion in the subsequent chapters. As a result, we will only focus on two types of relays, the frequency relay and the ROCOF relay in the rest of this report. Table 4.1: Comparison of the critical power mismatch of the frequency and VSR relays. Detection time 3 ms 5 ms 7 ms Settings FR / VSR FR VSR FR VSR FR VSR.5 Hz / 3 o 2.6% 19.2% 15.1% 14.5% 21.% 11.8% 1. Hz / 6 o 31.5% 27.7% 2.9% 19.6% 16.2% 15.6% 1.5 Hz / 9 o 42.2% 36.1% 26.6% 24.3% 2.2% 19.2% 2. Hz / 12 o 53.% 43.3% 32.3% 29.9% 24.4% 23.% 2.5 Hz / 15 o 63.9% 52.9% 37.9% 34.5% 28.3% 26.9% Key Factors Affecting the Relay Performance It is important to note that Figure 4.2 is an illustration of the typical characteristics of frequency-based relays. A number of factors can affect the curves. Research results show that the following factors have significant impact on the relay performance: 19

20 Inertia constant of the distributed generator; Voltage dependency of the feeder loads; and The mode of generator excitation control (if the load is not constant power type) The impact of DG inertia constant, H, can be seen from Figure 4.3(a). A larger H constant will lead to a larger critical power imbalance for the same relay setting. This is understandable since it takes longer time to cause a frequency deviation for DGs with a larger rotor inertia constant. In order to avoid the use of multiple relay curves for different H values, a normalised performance curve is proposed. For this curve, a normalised power mismatch value defined as: ΔP Δ P normalized = H (2) is used as the x-axis value of the curve. As shown in Figure 4.3(b), the normalised curve is the same for different H constants. As a result, a single relay curve can be used to assess DG applications involving different DG sizes H = 2. s H = 1.5 s H = 1. s H = 2. s H = 1.5 s H = 1. s Detection time (ms) Detection time (ms) Active power imbalance (pu) Normalised active power imbalance (pu) (a) Performance curves for different H constant. (b) Normalised performance curves. Figure 4.3: Impact of machine inertia constant on detection curves of a frequency relay. The impact of voltage dependency of the feeder loads is illustrated in Figure 4.4. The constant power load model represents a load characteristic that is independent of voltage. The constant current model represents a load characteristic whose power consumption varies linearly with the supply voltage. The constant impedance model represents a load characteristic whose power consumption varies with the square of the supply voltage. Since the constant impedance load can create a larger power surplus or deficit in an islanded system if the system voltage changes a lot, its performance curves deviate from the constant power curve more significantly. As the load-voltage dependency characteristics of a distribution feeder is hardly known, we have 2

21 to rely on the constant power load curve as a reference to obtain a general understanding of the critical power mismatch of the relays Constant power model Constant current model Constant impedance model Constant power model Constant current model Constant impedance model Detection time (ms) Deficit of active power Excess of active power Detection time (ms) Deficit of active power Excess of active power Active power imbalance (pu) Active power imbalance (pu) (a) Frequency relay (setting = ±1.5 Hz). (b) ROCOF Relay (Setting = 1. Hz/s). (The reactive power is in deficit for both cases) Figure 4.4: Impact of load to voltage dependency on the relay performance characteristics. A distributed generator typically has two modes of controlling its excitation system. One is to maintain constant terminal voltage (voltage control mode) and the other is to maintain constant power factor (power factor control mode) [8]. The tripping time versus power imbalance curves for the two control modes are compared in Figure 4.5 for the frequency relay, assuming there is a shortage of electrical power after islanding and the load is constant impedance type. For the conditions of active and reactive power imbalances simulated in this section, it is found that the critical power imbalance is larger if the excitation system is controlled by power factor than by voltage. This is due to the different response of nodal voltages under different control mode. The voltage control mode will lead to less voltage change in the system. The power shortage in the island is therefore more than that associated with the power factor control mode, which leads to a larger power mismatch for the voltage control mode. So the voltage control mode can result in faster frequency drift and smaller critical power imbalance. For the same reason, the case in which there is excess of electrical power after islanding will result in a smaller critical power mismatch for the power factor control mode than for the voltage control mode. As a result, which mode has smaller non-detection zone is dependent on if the island has deficit or surplus of power. Further study shows that if the load is constant power type, there is no difference between the two control modes. Research results showed that other factors such as feeder length and load power factors have little impact on the relay performance curves. If there are multiple DGs in an island, the frequency-based relays could interact with each other. This is because the tripping of one generator will change the power mismatch level in the island, which in turn affects the variation of system frequencies. The relay behaviours can be difficult to predict under such circumstance 21

22 [1]. Research results also show that the ROCOF relay can cause more nuisance trips than the vector surge relay. The main conclusions from such impact factor study can be summarised as follows: Detection time (ms) V control mode Q control mode ±1. Hz ±1.5 Hz ±2. Hz Active power imbalance (pu) Figure 4.5: Impact of DG excitation control modes. The voltage-dependency of the feeder loads has a significant impact on the relay performance curves. The load type also affects the performance indirectly through DG excitation modes. Since the load-to-voltage dependency is hard to quantify for a given distribution system and for different island formations, we recommend to use the relay curves obtained with the constant power load assumption as a reference. An approximate critical power mismatch level can be determined from the reference curve. A safe margin of.1 to.3 per-unit power mismatch may be added to critical power mismatch level to arrive at a conservative estimate of the non-detection zone. The DG inertia constant also has a significant impact on the performance curves. This factor can be taken into account by using H-constant normalised relay performance curves. The following factors do not have significant impact on the relay anti-islanding performance: feeder length, X/R ratio of the feeder impedance, load power factor, and reactive power imbalance of the island. The reactive power mismatch in an island has some impact on the relay performance. This subject will be discussed in Chapter 5 where a two-dimensional non-detection zone will be introduced. 22

23 4.2.3 Equations to Predict Relay Performance Equations to predict the performance of the frequency-based relays have been developed [1,11,12]. Mathematical models on which the equations are based are explained in Appendix A. For the frequency relay, the equation representing the relay performance under the constant power load condition has the following form: where t d 2Hφ φ = = f ΔP 3( ΔP / H + τ ) t d is the detection time; τ is the time used to compute the frequency value and run the relay algorithm. From manufacturers catalogues this intrinsic delay is around 8 ms. H is the generator H constant; f o is the power system frequency (6 Hz in this report); φ is the relay setting, for example.5hz; and ΔP is the power mismatch between load and generation in absolute per-unit value defined as ΔP= (P gen -P load )/P gen-rated (3) where For the ROCOF relay, the performance equation has the following form: 2H β β + (4) td = Ta ln 1 τ = Ta ln 1 + τ fδp 3( ΔP / H ) β is the relay setting, for example, 1.2Hz/s; τ is the time used to compute the df/dt value and run the relay algorithm. From manufacturers catalogues this intrinsic delay is around 13 ms. Ta is the time constant of a low pass filter that models the averaging algorithm to estimate df/dt. Typical value of Ta is 1 ms (or 6 cycles); For the vector surge relay, the performance equation is: (2ω K( α π )) t d = 2 2K ( α 2π ) where K = ω oδp / 2H = 2π (3ΔP / H ) ; D = (2ω K(α π )) 2 4K 2 (α 2π)(ω 2 α + 2π 2 K); α is the relay setting in the unit of radian. D (5) A set of general purpose curves to predict the relay performance is determined using the above equations and is shown in Figure 4.6. Note that the x-axis is the normalized power mismatch, defined as ΔP n =ΔP/H. Because of the normalization, the curves can be applied to generators of any size. References [1,11,12] further investigated empirical formulas for cases where the load is not a constant power type. Details can be found from the references. 23

24 1 9 8 ± 1. Hz ± 1.5 Hz ± 2. Hz Detection time (ms) Normalised active power imbalance ( ) 9 8 (a) Frequency relay..5 Hz/s ± 1.5 Hz/s ± 2.5 Hz/s Detection time (ms) Normalised active power imbalance ( ) (b) ROCOF relay. 6 degrees 9 degrees 12 degrees Detection time (ms) Normalised active power imbalance ( ) (c) Vector surge relay. Figure 4.6: Generalised characteristic curves of frequency based relays Limitations of Frequency (and Vector Surge) Relay The frequency (and vector surge) relay relies on frequency deviation to detect islanding conditions. Ideally, the relay should respond as fast as possible to frequency deviation so that 24

25 critical power mismatch is minimized. However, not all frequency derivations are caused by islanding conditions. As a result, one cannot set a frequency relay too sensitive. In fact, technical guides for DG interconnection recommend that the generators should not be disconnected due to small frequency variation [2]. A properly designed DG protection scheme must satisfy both the anti-islanding and frequency-variation immunity requirements simultaneously. This section analyzes if a frequency or a vector surge relay can satisfy both requirements. A common frequency-band employed by utility companies to avoid nuisance frequencydeviation-caused trips is 59.5Hz to 6.5Hz [2]. A frequency relay should not trip a generator if the frequency deviation is within that band. Accordingly, the most sensitive relay setting is.5hz for the under-frequency situation. On the other hand, utility and DG protection requirements also specify that a frequency relay must operate if the frequency goes below 57Hz. This leads to a relay setting of 3Hz. As a result, the available frequency deviation settings for a frequency relay to perform anti-islanding function are from.5hz to 3.Hz. This situation can be shown graphically in the trip-time versus power-mismatch plane as a feasible region called the application region of the frequency relay (Figure 4.7(a)). In the figure, the required time for islanding detection is assumed as 5ms. This requirement further restricts the options available to a frequency relay and establishes a horizontal boundary for the application region. A similar analysis can be performed for the case where there is surplus power in an island. The upper limit for over frequency is 6.5Hz, which leads to a minimal relay setting of.5hz. The 6.5Hz is the also the limit above which a generator must be tripped. As a result, the application region of the relay for anti-islanding purpose has the form shown in Figure 4.7(b) Upper limit curve: 57 Hz Detection time (ms) APPLICATION REGION OF FREQUENCY - BASED RELAYS D etection time line Freq. variation immunity curve: 59.5 Hz Active power imbalance (pu) (a) power deficit (under frequency) case Detection time (ms) APPLICATION REGION OF FREQUENCY - BASED RELAYS Detection time line Freq. variation immunity curve: 6.5 Hz Active power imbalance (pu) (b) power surplus (over frequency) case Figure 4.7: Application regions of a frequency relay. The following conclusions can be drawn from this analysis: There is an application region for the frequency relay. Anti-islanding settings are acceptable only if they result in a relay performance curve that falls into the application 25

26 region. Operating close to the boundary of the region for the purpose of improving antiislanding sensitivity is likely to increase the chances for nuisance generator trips. Because of the restriction of the application region, a non-detection zone in the range of at least 1% to 2% power mismatch always exists when a frequency relay is applied for anti-islanding protection. This region is more significant for power deficit (under frequency) case than for the power surplus (over frequency) case. In order to improve the anti-islanding performance, one may choice to use one setting for under-frequency case and another for the over-frequency case. Such a relay is expected to have a smaller overall non-detection zone for both power deficit and surplus case Limitation of the ROCOF Relay The ROCOF relay is based on the rate of frequency change. The rate of frequency change is essentially in proportion to the power imbalance in the islanded system. As a result, if the power mismatch is smaller than certain value, the rate of frequency change may never exceed the ROCOF relay setting, even if the frequency has deviated from its nominal value significantly. It implies that the ROCOF relay has an inherent non-detection zone. The following analysis will clarify this subject further. The swing equation of an islanded synchronous generator has the following form: 2H dω = P ω dt dδ = ω ω dt M P L = P SYS = ΔP where H is the generator inertia constant, ω = 2πf is the synchronous speed, f is the system nominal frequency and ΔP is the power mismatch in the island. The rate of change of frequency can be calculated as: df dt 1 dω f ΔP = = ΔP = 3 2π dt 2H H The above equation shows that the rate of change of frequency is proportional to the power imbalance. If we omit the averaging process needed to determine df/dt and the associated time delays, the relay activation criterion becomes: ΔP df ΔP (8) 3 = > β or > β / 3 H dt H (6) (7) 26