2004 IEEE Intemational ConferenceonElectric Utility Deregulation, Restructuring and Power Technologies (DRPT2004) April 2004 Hong Kong
Generate new relay settings
I Intelligent Method for Protection Coordination C.W. So, Member, IEEE, K.K. Li, Senior Member, IEEE Abstract: This paper presents the application of Artificial Intelligence in protection coordination. It can substantially improve the coordination of protection relay operations. The relay settings and coordination requirements are formulated into a set of constraint equations and an objective function is developed to manage the relay settings by the Time Coordination Method (TCM). Modified Evolutionary Programming (MEP) is employed to search for the optimum relay settings with maximum satisfaction of coordination constraints. The results show that the intelligent method for protection coordination can optimize the protection relay settings, reduce relay mis-coordinated operations, and increase supply reliability. The efficiency of TCM taking the fault current changes into consideration is also discussed in this paper. Index Terms - Protection Relay Coordination, Time Coordination Method, Power System Reliability. I. INTRODUCTION A modem protection system consists of various types of protective relays, which functions to detect and isolate system abnormalities swiftly. For various voltage level, the various combinations of main and backup relays are installed to provide complete protection to various power apparatus. The main protection relay works in unit protection principle and removes in-zone fault instantly. The backup protection relay is designed to backup the main protection relay in case it fails. The overcurrent relay is a major type of backup protection relay. As it is non- unit protection, it is discriminated by their operating time, which forms a sequence of backup relay operations for the unclear system fault. Any incorrect operation of the backup relays will result in a large area of supply interruption [ 11as well as decreasing the supply reliability. The backup protection relay coordination is thus necessary. It ensures that the fault clearance actions are in correct sequence and minimizes the supply interruption. As the power system is changing from time to time, the coordination work should be carried out upon any significant change in power system configurations [2]. Based on the communication ability equipped in modem digital relay, the coordinated relay settings can be downloaded through the communication network. The Time Coordination Method (TCM), which formulates the coordination of relay settings into a set of constraint equations and an objective function, is proposed in [3] to The authors would like to thank The Hong Kong Polytechnic University for supporting the research and publishing this work. C.W. So (e-mail: paulsot~~comnuter.or~) is currently working in a PhD project with the Hong Kong Polytechnic University. Dr. K.K. Li (e-mail: cckkli@polwi.cdu,hk) is with the Dcpartment of Electrical Engineering, Hong Kong Polytechnic University, Hong Kong. manage the relay settings. In distribution network, the major backup protections are Inverse Definite Minimum Time Lag (LDMTL) overcurrent and earth fault relays. Their operations depend on the fault current magnitude. Thus, the relay operation is affected by the change of fault current magnitude [7]. In fact, any unclear fault will be cleared by the backup relay operations. Any circuit breaker operation will result in changes in fault current distribution and magnitude experienced by backup relays. In order to boost the performance, the TCM must be able to handle a reasonable number of fault current changes. This paper shows how the TCM can significantly improve the supply reliability and reduce relay mal-operations. 11. PRINCIPLES OF TIME COORDINATION METHOD (TCM) Initialize relay settings I Objective value calculation I E No Fig. 1 Time Coordination Method The process flow chart of the TCM is shown in Fig 1. It formulates the power network and protection system into an objective function and a set of constraint equations. The objective function in the TCM is shown in equation (1). where Ri is each relay operation time and regulated by a scale factor a. C y is the operation time difference for each pair of relays and regulated by a scale factor p. CV, is the number of coordination constraint violations and regulated by scale factor x and 6. Note: Variable i, j and k are iterated for all possible system configurations. 0-7803-8237-4/04/$17.0002004IEEE 378 The initialization is to generate several sets of relay settings randomly. Each set of relay settings will be checked against all system constraints and configurations. The effectiveness of the relay settings is calculated by (1). Any violation of system constraints will be penalized and this will be reflected in the increased objective value. The optimization is used the Modified Evolutionary Programming (MEP) [ 3] to search for the optimum relay settings. The relay settings improvement process to produce the next generation is carried out by a sub-process of MEP called mutation. It is based on Gaussian normal distribution noise as shown in Eqn (2). x, =x, +or NI (0,l) (2) 0, =Jm where xi is the i element of relay settings. pi is the scale factor for EA mutation. yi is the offset for EA mutation. @(x) is the objective value of the relay settings. N, ( 0 , ~ ) The scale factor pi and the offset yi are to control the performance of EA and typically set to 1 and 0 respectively. is the Gaussian normal distribution noise. These newly generated relay settings should be passed to constraints checking and objective value calculation for the next EA generation. The process will stop after a fixed number of generations. The number of generations required to carry out the optimum relay setting depends on the pattern and the number of initial relay settings. III. MODIFIED EVOLUTIONARY PROGRAMMING The relay setting coordination is modeled according to its constraint optimization [ 3] . As the relay setting is discrete and constrained by the adjacent relay, it is difficult for the conventional gradient search method [5] to find the solution. The MEP is a stochastic parallel search method for multi-variables. It can increase the chances to search for the optimum relay settings in a fixed number of generations with maximum satisfaction of coordination constraints. Initialization process is used to generate several sets of protection relay settings in random basis to set up the initial population pool. A larger initial population pool will broaden the solution search area [6]. Each initialized protection relay setting is passed to the Protection Performance Evaluation module for coordination constraints checking [3]. Unfortunately, the pure random generated initial protection relay settings may not pass the coordination constraints checking. The setting pusher technique [7] is thus developed to improve the success rate of coordination constraint checking in 2004 IEEE International Conferenceon Electric Utility Deregulation, Restructuring and Power Technologies (DRPT2004) April 2004 Hong Kong 379 order to minimize the initialization time. The performance of EA is determined by population size, number of generations and mutation rate [SI and will be discussed in the result section. IV. FORMULATION +us BO Bus B1 Line L5 Bus B7 Circulating current protection relay Overcurrent protection relay Fig. 2 Sample distribution system The sample system as shown in Fig. 2 is considered in this study. Each circuit is protected by circulating current (CC) protection, Inverse Definite Minimum Time Lag (LDMTL) Overcurrent (OC) and Earth Fault (EF) protection. . The CC protection is a unit protection with a fixed setting and do not need to be changed adaptively. The IDMTL OC & EF protection is a non-unit protection. It operates on excessive current flowing through the circuit in which the TCM is applied to coordinate the settings. The TCM is required to calculate the relay setting for all possible system configurations. The relay with optimized settings should be capable to handle the change of fault current due to the tripping of adjacent circuits [7]. In fact, most of the faults can be cleared by two to three protection relay operations. For instance, for a busbar fault occurred on Bus B5, if the relay located at Line L25 operates first, the relays located in other lines will experience a fault current change. If the relay located in Line L36 operates next, the relays located in Line L14, L46, L67, L57 will experience the second change of fault currents. Finally the relay located in Line L57 operates and the fault is completely cleared. The relay operation in Line L36 is definitely a mal-operation. In this case, the TCM should at least handle all the relay operations up to two step changes of fault current. It should also be able to handle other kind of mis-coordinated relay operations 2004EEE International ConferenceonElectric Utility Deregulation, Restructuring and Power Technologies (DRPT2004) April 2004 Hong Kong which will caused fault current changes. In the result section, the efficiency of the TCM with different number of fault current changes will be examined. The more number of fault current changes to be handled, the larger number of system configurations should be considered. The possible system configurations are considered as follows: For a fault in a particular busbar, the number of system configurations C to be studied due to the number of lines N in the system is: C=2N The number of system configurations C' if r out of N circuits will trip to isolate the fault is shown in equation (4). (3) i=O In the sample system as shown in Fig 2, although the number of configurations is C=2' =256, but some rare system configurations may be ignored. If the maximum allowable circuit outages =3, the number of system configurations to process C' =C: +C: +C: +C: =93 . The processing time reduction due to the minimized system configurations is equal to 6- 93 -36,3% . The effect of the reduced number of system configuration in constraint checking will be discussed in the result section. C 256 During the constraint checking, the constraint violations are checked against whether the operation time difference between the upstream and downstream relays is smaller than the coordination margin [3,4]. The objective value is also calculated during constraint checking. One of the major objective of the TCM is the minimization of the number of constraint violations. V. RELIABILITY EVALUATION The benefits of the TCM may be measured by the supply reliability indices. The basic principle of the supply reliability is presented by Billinton and Allan [9,10]. The effectiveness of protection relay operations contributed to the supply reliability [I 11may be computed by the three classic calculations. The first class of reliability calculation is the stuck breaker and is shown in equations (4) and ( 5) . us ( b) =' brb (4) ' 1 (b) =' b (5) Where A,, is the failure rate of stuck breaker. rb is the switching time to restore system due to stuck &(b) is the supply b failure rate due to stuck breaker. Uv(b) is the supply b restoring time. The second class of reliability calculation is the breaker. busbar fault and is shown in equations (6) and (7). Where Afis the failure rate of busbar. rf is the repair time of busbar. &(b) is the supply b failure rate due to busbar fault. Ub(b) is the supply b repair time. The third class of reliability calculation is the protection fault and is shown in equations (8) and (9). U, ( b) =Pf ',r, (8) (9) Where A, is the failure rate of the cable. r, is the switching time of the cable. Pf is the probability of protection failure. AJb) is the supply b failure rate due to protection U,(b) is the supply b restoring time. fault. The resultant reliability of the supply b due to protection operation is shown in equations (lo), (1 1) and (12). A (b) =As@) + +&,(b) (10) [E] U@) =Us@) +Ub(b>4- Up@) r(b) =U(b)/A (b) Where A (b) is the failure rate of supply b. U(b) is the repair time of supply b. r(b) is the mean duration of supply b interruption. The supply reliability for every supply busbar may be evaluated by simulating the busbar fault, stuck breaker and protection failure in each busbar. For instance, a single-phase-to-earth fault occurs at busbar bi, the protection relay operates according to the fault current distribution and resulting in busbar bj loss of supply (bifbj). The supply reliability of busbar bj should be updated according to equations (6) and (7). The protection failure and stuck breaker cases are also simulated by applying busbar fault on the circuit connected to the busbar and update the reliability indices by equations (4), (9, (8) and (91). The resultant reliability indices will be the sum of all individual simulated reliability indices according to equations (lo), (1 1) and (12). VI. SIMULATION RESULT AND DISCUSSION The relay information and system parameters of the sample distribution system are shown in Table 1 and 2 respectively. 380 2004 IEEE Intemational Conferenceon Electric Utility Deregulation, Restructuring and Power Technologies (DRPT2004) April 2004 Hong Kong Note: All per-unit (pu) values are based on 100MVA. Table 3: Si1 The TCM can be set of different population size, number of generations and number of fault current changes. The results shown in Table 3 are used to determine the case for the best TCM performance. The contribution of constraint violations in objective value is proportional to the number of system configurations. Moreover, from the results, the number of constraint violations depends on the number of fault current changes, the comparison of the objective value is only valid for those cases with the same number of fault current changes, i.e. case 1 to 3,4 to 6, 7 to 9, 10 to 12, 13 to 15 and 16 to 18. For cases 1 to 3, they have been checked based on no fault current change. The TCM then only checks the relay operations with single and three phase busbar faults at E31 to B7, i.e. 12 fault cases are being studied. In cases 16 to 18, the number of system configurations based on equation (2) for a fault in either busbar is c =2 c p = 219 . As there are six busbars in the system and two types of fault are simulated, a total of 219 x 6 x 2 =2628 fault cases are to be studied. Thus, the number of constraint violations for cases 16 to 18 are considerable larger than cases 1 to 13. When a busbar fault occurred, two or more tripping should be carried out by OC relay to isolate the fault completely. For cases 1 to 6, since the number of fault current changes is less than 2, they are impractical to implement. The number of constraint violations and the processing time per generation of cases 7 to 9 are less than cases 10 to 18. Case 8 has the smallest number of constraint violations. Case 14 has better reliability indices with 0.6160 f/yr and 1.2016 hr/yr. It implies that case 13 to 15 has less chance in loss of supply due to relay mis- coordination. As the roll of distribution is to provide a reliable power network to customers, relay settings of case 14 should be the best for retaining the supply reliability. The performance of MEP depends on the population size, the number of generations and the number of fault current changes. For cases 13 to 15, a better objective value occurs in case 14 for population size of 50 rather than case 13 or 15, similar behaviors occur in every three consecutive cases. It implies that population size of 100 may be over-crowed, in which better protection settings are not easily to be selected to survive during the selection process in MEP. VII. CONCLUSION r = O The TCM can make relays more intelligent and adapt to the system configuration. The supply reliability can be improved by reducing the number of mis-coordinated relay operations. The setting of the TCM with different number of fault current changes is examined. The result shows that better improvement of supply reliability do not occur in larger number of fault current changes. It implies that the unrealistic number of fault current changes will cause the TCM to over-test the system and the TCM force the protection relays to handle those unrealistic system configurations. VIII. REFERENCE [l ] R.P. Graziano, V.J . Kruse, G.L. Rankin, "Systems Analysis of Protection System Coordination: A Strategic Problem for Transmission and Distribution 38 1 2004 IEEE I ntemational Conference on Electric Utility Deregulation, Restructuring and Power Technologies (DRPT2004) April 2004 Hong Kong Reliability, IEEE Transactions on Power Delivery, Vol. 7, NO. 2, pp 720-724, April 1992. [2] William J . Ackerman, Substation Automation and the EMS, I999 IEEE Transmission and Distribution Conference, April 1999, Los Angeles, USA, Vol. 1, pp. [3] C.W. So, K.K. Li, Time Coordination Method for Power System Protection by Evolutionary Algorithm, IEEE Transactions in Industry Application, Vol. 36, No. 5, [4] C.W. So, K.K. Li, K.T. Lai, K.Y. Fung, Application of Genetic Algorithm for Overcurrent Relay Coordination, IEE 6Ih International Conference on Developments in Power System Protection, Nottingham, UK, March 1997, pp. 66-69. [5] R. Salomon, Evolutionary Algorithms and Gradient Search : Similarities and Differences, IEEE Transactions on Evolutionary Computation, Volume 2, pp [6] C.W. So, K.K. Li, K.T. Lai, K.Y. Fung, Overcurrent Relay Grading Coordination Using Genetic Algorithm, IEE APSCOM-9 7 International Conference, Hong Kong, Vol. 1, pp. 283-287, November 11-14, 1997. [7] C W So, K K Li, Overcurrent Relay Coordination by Evolutionary Programming, Journal of Electric Power System Research, Volume 53, pp 83-90, 2000. [8] A. E. Eiben, R. Hinterding, and Z. michalewicz, Parameter Control in Evolutionary Algorithms, IEEE Transactions on Evolutionary Computation, Volume 3, pp [9] R. Billinton, and R.N. Allan, Reliability Evaluation of Engineering Systems: Concepts and Techniques, Plenum Press, New York, USA, 2d Edition, 1992. [ 101R. Billinton, and R.N. Allan, Reliability Evaluation of Power Systems, Plenum Press, New York, USA, 2d Edition, 1996. 274-279. Sqt/Oct 2000, pp. 1235-1240. 45-55, J uly 1998. 124-141, J uly 1999. C11lJ.J. Meeuwsen, W.L Kling, S.P.J . Rombouts The influence of protective relay schemes on the reliability indices of load points in meshed operated mv networks, I 4rH International Conference and Exhibition on Electricity Distribution, Part 1 : Contributions, Congres International des RCseaux Electriques de Distribution (CIRED), J une 1997, IEE Publication No. 438, Vol;. 4, pp. 14/1-14/5. IX. BIOGRAPHIES C.W. So rcceived the BEng and PhD degree from the Hong Kong Polytechnic University in 1996 and 2001 respectively. He is working as an Engineer in CLP Power Ltd. in Hong Kong, responsible for Protection and Substation Automation. His rescarch interest is power system protection, the applicahon of artificial intelligent, substation automation and power system computer programming K. K. Li received the M.Sc. and the Ph.D degree from the University of Manchester Institute of Science and Technology, Manchester, U.K., and City University, London, U.K. Heis currently an Associate Professor in the Department of Electrical Engineering, Hong Kong Polytechnic University, Hong Kong. His research interests are power systcmprotection and A I applications in power systems. 382