Protective Relay Coordination of High Voltage Power Systems with Renewable Generation

Transcription

1 Integrated Master in Electrical and Computer Engineering 2010/ Protective Relay Coordination of High Voltage Power Systems with Renewable Generation Gonçalo Belchior DEEC, Sec. Energia, Instituto Superior Técnico Av. Rovisco Pais, 1, Lisboa, Portugal goncalo.belchior@ist.utl.pt Abstract The main goal of this paper is to propose a set of methods and to explain scripts developed in MATLAB and macros developed using the "Computer-Aided Protection Engineering (CAPE), with the purpose of studying, autonomously and automatically, the distance relay behavior when a fault is simulated in the network. All the distance relays in a High Voltage Network are analyzed in this short-circuit study and the most important generation profiles, with greater infeed variation, for each distance relay, are identified. Macros were developed to identify and model the generation in the network that effectively contributes to a fault. Very High Voltage generation was modeled and also the wind generation for each network topology and simulated fault, fulfilling the imposed limitations of the Portuguese legislation. The zones of each distance relay are probabilistically coordinated, using a methodology developed in the nineties for transmission networks and applied to the Portuguese Transmission grid, bearing in mind the inherent measurement errors and the most influent generation profiles obtained after modeling. The main innovation of this work consists in the automated joint of algorithms programmed in MATLAB with macros programmed in CAPE. Index Terms Automation, Computer-Aided Protection Engineering, Distance Relays, High Voltage Network, Optimization, Protective Relay Coordination, Wind Generation. T I. INTRODUCTION HE power system is in constant evolution and has an extraordinary dimension and at the same time vulnerability. Over the years were studied ways to improve its operation, including the optimal protective relay coordination, as presented in the master thesis of Eng. João Afonso [1] and Eng. Reis Rodrigues [2] oriented by Professor Pinto de Sá, for the Portuguese Transmission grid. The High Voltage Network that once had been purely radial, has suffered severe changes in its topology with the arrival of distributed generation, in particular, with the significant introduction of wind farms. The integration of this new distributed generation results in a power system where the power flow is bidirectional in opposition to traditional power systems, where the power flowed in only one way, from large facilities to the consumers [3]. This change in the paradigm of the electricity grid and the new obligations imposed by the Portuguese legislation to wind farms makes necessary rethink the previously defined methods, assumptions and strategies for the protection and control systems. Although the purpose of this type of generation is to increase the capability to meet consumption, taking advantage of the resources locally available, and if possible, reduce losses, there are impacts on power quality and in the protection system that cannot be neglected [4]. This thesis aims the optimization and automation of the protective relay coordination in the High Voltage Portuguese grid. For this study, are identified, when a fault is simulated, the most influential generation profiles for each relay, in other words, are identified the generation profiles who contribute to a greater infeed variation seen by the relay. The CAPE and MATLAB, individually, do not allow the analysis that is desired, limiting it to pre-defined or barely adequate concepts. However, this work is demonstrative that it is possible to achieve an automated joint of algorithms programmed in MATLAB with macros programmed in CAPE, taking advantage of the resources of each program and enabling a perfect marriage of methods and concepts that optimize the protective relay coordination. II. WIND GENERATION MODELING The number of wind farms has increased substantially in recent times. So, it is important to know the impact of the presence of this type of generation in the networks and understand its influence on the coordination of the existing relays.

2 Integrated Master in Electrical and Computer Engineering 2010/ The wind generation can be represented by an equivalent, which simulates a coherent aggregate of all wind generation in area of influence of its substation [5]. For each wind park this production ranges typically from 5% (minimum) to 90% (maximum) of its rated power. However, there is a greater infeed variation if the wind farm operates at the rated power, therefore it was decided, in the present work, simulate only two states for this generation: disconnected from the grid and not contributing with any infeed or connected at rated power. Analyzing the document of Instituto de Engenharia Mecânica e Gestão Industrial about wind farms in Portugal [6], was found that ENERCON is the manufacturer with the largest market share, about 52.5%, and E-82 is its most widely used model, present in about 168 wind farms in Portugal. In the brochure of this model [7], was concluded that the type of generator used is a variable speed synchronous generator (Full Converter Wind Generator), and therefore, in the current work is modeled, through the CAPE, all wind farms in the network with this type of generators. A wind farm may contain dozens of wind generators. These generators are connected by an intricate bus system to the collector of the wind farm substation. While the influence of a single wind generator may be small, the set of all generators in the farm can cause a significant impact when occurs a fault in the network. [8] Therefore, to determine the equivalent of the wind farm, was used the data available on the network provided by EDP. The power of the equivalent generator was calculated by adding the powers of all generators of the farm, the equivalent impedance of the generator was obtained by making the parallel impedance of all the generators and the equivalent impedance of the transformer was determined doing the parallel of all transformers associated and added in series with the transformer of the substation. { For voltage values above 0.9 pu, the wind farm doesn t have to provide any reactive current. To simulate this behavior of wind farms is proposed an algorithm. The algorithm starts with one of the wind farms, whereas all others are set with default values obtained from the data provided by EDP, and are determined the voltage and current on the bus at the network side, imposed by the short-circuit simulation. If the current value is greater than 1.1 pu, due to limits imposed by power electronics, must be limited, it was considered in this work to that value. To determine the equivalent impedance of the generator of the wind farm is necessary to analyze the voltage value obtained. In the case of bus voltage value be less than 0.2 pu, the wind farm is disconnected from the network, as mentioned above. It is allocated for this purpose, the value of zero for resistance and a very high value for the equivalent reactance of the generator of the wind farm. The algorithm simulates in this case the minimum voltage protections of the wind farm. For a voltage between 0.2 and 0.5 pu, in accordance with the legislation, the wind farm has to provide the reactive current of 0.9 pu and not consume any current active during the fault. In the CAPE software, the electromotive force of the synchronous generator is considered equal to 1 pu and the current injected through the wind farm can be represented by the following expression: - (2) To determine the value of the equivalent impedance of the generator (Z G ), is used the equation above with the values of active current and voltage obtained and using the desired reactive value: - - (3) (1) Fig. 1. Equivalent obtained for each wind farm. According to the Portuguese legislation, the wind farm must remain connected to the grid for voltage dips (fault ride through capability - FRTC) with voltage values above 0.2 pu. Bellow this value the wind farm is disconnected. During a voltage dip, the wind generators must supply the network with reactive current, according to the following expression, in order to provide voltage support. The previous value of Z G and the new one are stored in vectors. For values between 0.5 and 0.9 pu, the legislation requires that the wind farm inject a reactive current imposed by the expression (1). So, the equivalent impedance of the generator is given by: (4) The previous value of Z G and the new one are stored in vectors. If the voltage is above 0.9 pu, the wind farm doesn t have to submit any reactive current. If the current has been limited so that the module does not exceed 1.1 pu, the equivalent impedance of the generator is calculated to ensure that current. Otherwise, it uses the value of Z G pre-defined for each farm.

3 Integrated Master in Electrical and Computer Engineering 2010/ For the second wind farm, the values of current and voltage are determined considering that all farms have impedance values pre-defined with the exception of the first, which is simulated with the new value to the impedance of the generator. The new value of Z G is calculated for the second farm and the third farm is simulated with the new values of the two previous farms. This method is repeated successively until be obtained the new value of Z G for all wind farms. In the first cycle the condition of convergence cannot be verified, since it is necessary to calculate the impedances of all wind farms. In the second cycle, starting with the values determined in the first, the new value of impedance is recalculated, so that each wind farm complies with the legislation. In each cycle, the values of impedances of the generators are close to the values obtained in the previous cycle. When the difference between the calculated impedance for the generator and the impedance calculated above is less than for all the farms, we obtain a value for the equivalent generator of each farm, with an error less than 0.1%, which guarantees the compliance with the injection of reactive current provided by the legislation and therefore it is considered that the method converged. This algorithm was programmed in a macro, so it could be applied and simulated on the network using the CAPE software. III. PROBABILISTIC APPROACH A. Study of the Generation Profiles by Macros in CAPE The algorithm explained in the previous section was implemented in a macro. Other macros were created to model the Very High Voltage generation, responsible for the strong current injection on the network, in two profiles of current, minimum and maximum short-circuit currents tabulated by REN [9] and to model the thermal and hydro generations, removing this generation from the network when the voltage is below 0.85pu. The set of all generation in the network corresponds to a generation profile. For each distance relay and fault at the end of their neighbor line, depending on the generation profile at the time of fault, the relay can see more or less infeeds, and consequently, higher or lower apparent impedance. Each generation profile has a different influence for the relay and for the fault in question. To identify the profiles that contribute to a greater variation in the impedance seen by the relay, it is inconceivable to simulate all possible combinations of generation due to the excessive number of combinations. In the network studied there are 40 generations, and since each one can be online or offline, or in case of the Very High Voltage generation, with maximum or minimum short-circuit currents, there are a total of 2 40 possible combinations. So the strategy implemented to identify the most influential profiles was using a weight matrix that combines the weights of the current injected by each generation to their distance to the relay, measured by the impedance value of the lines between generation and relay. Generations with higher injected currents is assigned a greater weight to the currents. Generations with a lower impedance value, is assigned a greater weight to distances. The weights of the currents are twice the weights of the distance due to their significance and because the method of calculating minimum distances is done through a command that returns the path with the least number of buses, which may not necessarily mean the most short path. Multiplying the weights of both vectors are selected generations with higher weights. Starting from the base profile, with the whole generation connected to the network, with the modeling of the Very High Voltage generation with maximum current profile and with the modeling of wind generation for a fault at the end of their neighbor line of the relay, is simulated the generation profiles that have the greatest influence on the relay selecting the six upstream and downstream generations most important for the relay. Fig. 3. Variation of the impedance seen by the relay for different generation profiles. Fig. 2. High Voltage Portuguese grid used in the CAPE. This number is chosen because for each one of the 10 portions of the network there are at most six generations online when a fault is simulated.

4 Cumulative Probabilities Probability(%) Integrated Master in Electrical and Computer Engineering 2010/ Is simulated all combinations for these six generations connect and disconnect from the network and all the impedances seen by the relay are stored. B. Protective Relay Coordination using Scripts in MATLAB The impedance values obtained in the macros are stored in a matrix in MATLAB that records in each column the impedances seen by each relay for a short circuit at the end of a neighbor line for different generation profiles. With the same index but in another matrix are associated the corresponding generation profiles. Each profile has a different probability [9], [10]. The programs developed in MATLAB allow, with data provided by CAPE, the coordination of the four zones of all distance relays in the network. This coordination is based on a methodology developed in the nineties for transmission networks and applied to the Portuguese Transmission grid [1],[11] e [12]. The regulation of the four zones was made as follows: Zone 1: Zone 1 takes into account only the measurement errors. These measurement errors can be represented by a normal distribution with a given mean and standard deviation. The probability density function for zone 1, for an impedance Z of the line is: - - With the advantage of the Central Limit Theorem, considering the different errors that affect the measure of the relay as independent variables, the probability density function converges to a Gaussian function whose standard deviation σ 1 =0.08 is defined by the quadratic sum of standard deviations of the different errors. Zop 1 is defined for all relays with a value of 80% of the impedance of its own line. Zones 1 of all relays in the network are set to the value of Zop 1. Zone 2: Zone 2 is also represented by a probability density function with mean equal to Zop 2 and standard deviation σ 2 = This zone is adjusted in order to coordinate with the zone 1 of the relay in the shortest neighbor line. Each impedance seen by the relay, that is stored in a matrix in MATLAB, has associated a measurement error of zone 2 represented by a normal distribution. Adding all normal distributions associated to each impedance is obtained an equivalent distribution (6). (5) The neighbor line is divided into 100 parts considering for each point that the location of the defect is a random variable with uniform probability density: p(z)=1/zl. So, zone 2 is adjusted taking into account three variables: the apparent impedance seen by the relay, the measurement errors and the fault location. Fig. 4. Equivalent distributions for faults at 10%, at 50% and at 100% of the neighbor line. Zop2 is defined by the mean value of each equivalent distributions and therefore for each fault location is stored the cumulative probability in a matrix in MATLAB. To coordinate these zones, it is considered that the probability of the primary relay doesn t act on zone 1 and the probability of the secondary relay act in zone 2 are independent events (7). The cumulative probability of primary relay doesn t act on zone 1 is given by: Zap(Ω) Is calculated the product of both cumulative probabilities, the cumulative probability of zone 2 for each fault location and the cumulative probability of zone 1, until find a value for Zop 2 that ensures a missing selectivity probability less than 0.05% and that more than 99% of the line of the relay is protected. (7) (8) -( - ) (6) Fig. 5. Cumulative probabilities for zone 2 of secondary relay and for zone 1 of primary relay.

5 Cumulative Probability Cumulative Probabilities Cumulative Probabilities Integrated Master in Electrical and Computer Engineering 2010/ Zone 2 is set to the Zop 2 obtained. Fig. 6. Coordination of the zone 2 of the secondary relay with the zone 1 of the primary relay. (12) The cumulative probability of zone 3 is calculated, making the product of both cumulative probabilities, increasing Zop 3 until find the largest value that ensures a missing selectivity probability less than 0.5%. If the error is greater than 0.5%, the value of Zop 3 is decremented until the product of both cumulative probabilities ensures an error less than 0.5%. In these situations, the protection of the neighbor line is guaranteed by the zone 4 and zone 3 only ensures selectivity with the zone 2 of the primary relay. Zone 3: Zone 3 is adjusted to guarantee selectivity with the zone 2 of the primary relay in the shortest neighbor line taking into account the infeeds. The strategy is similar to that was used for zone 2, but considering the fault only at the end of the line. This zone can be represented by a probability density function with mean equal to Zop 3 and standard deviation σ 2 = Each impedance seen by the relay, that is stored in a matrix in MATLAB, has associated a measurement error of zone 3 represented by a normal distribution. Adding all normal distributions associated to each impedance is obtained an equivalent distribution (9). -( - ) To coordinate these zones, it is considered that the probability of the primary relay doesn t act on zone 2 and the probability of the secondary relay act in zone 3 are independent events (10). (9) (10) Since the functionality of the zone 3 is provide backup to the neighbor lines, it isn t necessary consider an error as low as for the coordination of the zone 2 with the zone 1 of the primary relay. The cumulative probability of secondary relay act in zone 3 is given by: Zone 4: Fig. 7. Coordination of the zone 3 of the secondary relay with the zone 2 of the primary relay. Zone 4 is regulated to protect the longest neighbor line, taking into account the infeeds. Each impedance seen by the relay for a fault in the longest neighbor line has associated a measurement error of zone 4 represented by a normal distribution with standard deviation. Adding all normal distributions associated to each impedance is obtained an equivalent distribution similar to that shown in (9). In order to protect the longest neighbor line with the largest infeed, the value of Zop 4 is regulated to ensure that protects the totality of neighbor line with an error less than 0.01%. This zone may eventually not be necessary in the case of the zone 3 protect the longest neighbor line, when is regulated to ensure the protection of the shortest neighbor line taking account the infeeds. (11) The cumulative probability of primary relay doesn t act on zone 2 is given by: Fig. 8. Regulation of the zone 4 for the secondary relay.

6 Integrated Master in Electrical and Computer Engineering 2010/ C. Organization of the Programs The automated joint of algorithms programmed in MATLAB with macros programmed in CAPE is done according to the phases of the following figure. This linkage between the programs is only possible through the use of txt files. on the command line of the CAPE: input C: \ cape \ tese \ ZescCAPE.txt. The CAPE reading this file identifies the macro and the parameters of each protection: relay number and values for the zones. That macro is executed to all the relays in the network in order to regulate the relays for the values that were obtained from the probabilistic method. However, first should be compared the results with the values initially regulated, using the file comparaprotec.txt that determines the zones of the distance relays in CAPE and prints these zones together with the values obtained in MATLAB in the CAPE screen. IV. CONCLUSION With this work was possible an automated joint of algorithms programmed in MATLAB with macros programmed in CAPE, and was obtained an optimum coordination of distance relays in a portion of the High Voltage Portuguese network. Fig. 9. Representative scheme of the phases of the program. According to the previous scheme, the phase 1 is performed in MATLAB, where is created the file perfiscape.txt. This file serves as a guide for run macros in the CAPE. Phase 2 is performed on CAPE, after the user puts in the command line the command "INPUT" input C: \ cape \tese \ perfiscape.txt. At this stage, the CAPE follows the file perfiscape.txt as a guide, running macros when reading their names (and its parameters, if the macro has input parameters). In the last macro is written a message to the file Pronto.txt and are created, in the same directory, the files: SAIDACAPE.txt with the value of the apparent impedance seen by each relay in the network; PERFISGERACAO.txt with generation profiles and SIR.txt with the values of "System impedance ratio" (SIR) associated with each of the mentioned impedances. The last file is used only to identify the relays that are impossible to coordinate because of the very high value of SIR. In phase 3 when the script in MATLAB receives the message in the file Pronto.txt, initializes the reading of data on SAIDACAPE.txt and PERFISGERACAO.txt, followed by the coordination of the zones of the distance relays. After the regulation of four zones of all distance relays in the network is created the file ZescCAPE.txt with the relay number and value that was obtained for each zone. Similarly is created the file comparaprotec.txt, both in the same directory as the rest. Initializes the phase 4, by placing the command "INPUT" REFERENCES [1] Afonso, J., Coordenação Probabilística de Proteções de Distância da Rede Eléctrica Nacional, IST Master Thesis, Lisbon, April [2] Reis Rodrigues, A. C., Coordenação Sistémica de Proteções Direccionais de Máxima Intensidade em Redes de Transporte de Energia Eléctrica, IST Master Thesis, Lisbon, June [3] Jenkins, N., Allan, R., Crossley, P., Kirschen, D., Strbac, G., Embedded Generation (Power & Energy Ser. 31), The Institution of Engineering and Technology, London, [4] Vu Van T., Driesen J., Belmans R.: " Dispersed generation interconnection and its impact on power loss and protection system," IEEE Young Researchers Symposium in Electrical Power Engineering - Intelligent Energy Conversion, Delft, The Netherlands, March [5] Portaria n.º 596/2010 de 30 de Julho, Diário da República n.º 147 1ª série, Ministério da Economia, da Inovação e do Desenvolvimento. [6] Instituto de Engenharia Mecânica e Gestão Insdústrial, Parques Eólicos em Portugal, Faculdade de Engenharia da Universidade do Porto, December [7] ENERCON, Aerogeradores ENERCON Tecnologia e Assistência Técnica. [8] Muljadi, E., Butterfield, C.P., Ellis, A., Mechenbier, J., Hocheimer, J., Young, R., Miller, N., Delmerico, R., Zavadil, R., Smith, J.C., Equivalencing the collector system of a large wind power plant, Power Engineering Society General Meeting, IEEE, Montreal, October [9] REN- REDE ELÉCTRICA NACIONAL, S.A., Caracterização da Rede Nacional de Transporte para Efeitos de Acesso à Rede em 31 de Dezembro [10] EDP Produção, Produção Números [11] Pinto de Sá, J. L., Afonso, J., Rodrigues, R., A Probabilistic Approach to Setting Distance Relays in IEEE Transactions on Power Delivery, Vol. 12, nº. 2, April 1997, pp [12] Afonso, J., Pinto de Sá, J. L., Rodrigues, R., Probabilistic Coordination of Distance Protection Systems, Proceedings of the 12th Power Systems Computation Conference Vol. I, Dresden, August 1996.