Coordinated PID Secondary Voltage Control of a Power System Based on Genetic Algorithm

Transcription

1 Helwan University From the SelectedWorks of Omar H. Abdalla Winter December 27, 26 Coordinated PID Secondary Voltage Control of a Power System Based on Genetic Algorithm Prof. Omar H. Abdalla, Helwan University Prof. A. M. Abdel Ghany, Helwan University Hady H. Fayek, Heliopolis University Available at:

2 Proc. 26 Eighteenth International Middle East Power Systems Conference (MEPCON), Paper ID: 58, pp , Helwan University, Cairo, Egypt, Dec. 26. Coordinated PID Secondary Voltage Control of a Power System Based on Genetic Algorithm Omar H. Abdalla, Life SMIEEE and A.M. Abdel Ghany, MIEEE Electrical Power and Machines Department Helwan University Cairo, Egypt ohabdalla@ieee.org, ghanymohamed@ieee.org Hady H. Fayek Electromechanics Department Heliopolis University Cairo, Egypt hadyhabib@hotmail.com Abstract This paper presents the design of secondary voltage control by tuning the parameters of coordinated PID controllers to eliminate voltage violations in power system contingencies. The coordinated PID control parameters are tuned by using genetic algorithm. The study is presented on modified IEEE 4 bus system. The coordinated secondary voltage control is assigned to the generators and STATCOMs in order to eliminate voltage violations in system contingencies. Here, a contingency means a disturbance leading to tripping a generator at a certain bus. Measurements of pilot buses voltage magnitudes are done by using minimum number of phasor measurement units. Simulation and optimization studies are performed by using MATLAB / Simulink software. Key words: Coordinated secondary voltage control, Coordinated PID controller, Genetic algorithm (GA), Pilot buses. I. INTRODUCTION One of the main reasons of blackouts or brownouts of a power grid is the voltage instability. Voltage stability phenomenon is defined in [] as the ability of a power system to maintain steady voltages at all buses in the system after being subjected to a disturbance from a given initial operating condition. Voltage stability also plays the most important role in the reliability and security evaluation of the grid [], [2]. Voltage control in power system consists of three hierarchical control levels: Automatic Voltage Regulation (AVR), Secondary Voltage Control (SecVC), and Tertiary Voltage Control (TerVC) [3]. AVR aims to control the voltage magnitude of the buses where the reactive power sources (such as synchronous generators, synchronous condensers, static var compensators, Static Compensators (STATCOM), etc.) are installed. The measurements used in this control level are taken locally [4]. SecVC aims to control the voltage magnitude of a pilot bus which is a certain load bus required to be controlled. This control level is normally performed by using hardware that leads to adjust the reference point of the AVR(s) which is/are located at the same region of the pilot bus at a speed slower than that of the AVR control level. SecVC is also responsible for the determination of the voltage control regions and indicates its association with each load bus. As power system operating conditions continuously changes so SecVC should be flexible enough to change the control regions and the set point which cope with the entire grid conditions. TerVC is responsible for the determination of optimal reference value of each load bus voltage magnitude in the grid. The optimal values are calculated from optimal load flow computations which are done to achieve a certain objective function such as minimum power losses or minimum load shedding or optimizing reactive power reserve. TerVC has a time range from half an hour to a complete hour to be periodically performed [4]. In modern power systems (smart grid) more facilities will be available for monitoring, communication, controlling and protection among all the grid levels. The three voltage control levels will be widely required especially with the high penetration level of renewables. Smart grid is characterized by wide area measurement using Phasor Measurement Units (PMUs) based on Global Positioning System (GPS). The coherent and time-synchronized measured quantities are essential for accurate and real-time wide-area power system voltage control [5], [6]. This paper is focused on SecVC by using a coordinated PID controller which is designed using genetic algorithm. The controller used to eliminate voltage violation under different contingencies such as generator outage. The proposed approach employs minimum numbers of PMUs to get real time measurements as control feedback signals, so that a precise and fast secondary voltage control can be achieved. The subsequent sections are organized as follows: Section II describes the SecVC and the design of its proposed PID controller. Section III shows the system description. Section IV presents a proposed method for determining the minimum number of PMUs. Section V briefly describes the GA. Section VI presents the application of the coordinated SecVC to the IEEE 4-Bus system. Section VII presents the simulation results and Section VIII summarizes the main conclusions. II. SECONDARY VOLTAGE CONTROL As mentioned in details in [4], for SecVC application the power grid is divided into many voltage control regions; for example the IEEE 4 bus has 8 regions, assuming that each Published in IEEE Xplore, DOI:.9/MEPCON

3 load bus represents a region. The voltage magnitude of each load bus is controlled by a selected set of reactive power sources in the same region. The main objective of SecVC is to control the pilot bus voltage magnitude by changing the reference points of the regional generators AVRs and STATCOM S firing circuits in a coordinated way. Fig.. SecVC overall structure [3]. control level function. This is done by setting the dominant time constant of the SecVC more than that of all the AVRs in the power grid and the TerVC repetition time is higher than that of all SecVCs [3]. The voltage control sequence is done with the following procedure: The AVR controls the bus voltage of the generating unit or the station, which is done in most cases by a using local controller. The time constant of the AVR is set to be near.5 sec. Unit Cluster Control (CC) is shown in Fig. 2. The AVR voltage reference VG S could be varied between V min and V max which obtains the unit reactive power production Q G corresponding to its reference value Q ref as illustrated in (). V GS K t G ( Q ref Q G dt ) V V max min () Where K G is the regulator integral gain and it is equal to (X TG + X eq ) / (T G ) and tuned in such a way that the closed loop has a dominant time constant (T G ) of about 5s, X TG represents the transformer generator reactance while X eq represents the equivalent reactance of the line connecting the supporter generator with the pilot bus. The AVR dynamic response should be fast enough to deal with the local network disturbances to avoid being affected with the reactive power loop. The reference value Q ref is obtained from the product of the reactive power level q by the unit capability limit Q GL as shown in (2). Q GL is computed on-line based on the operation conditions of the reactive power source. Q ref q * QGL (2) The reactive power level q may be provided by the pilot Central Area Control (CAC). In closed-loop and realtime, the reactive level q in the interval between its minimum q min = -% and its maximum q max = % which achieves load bus voltage magnitude V p corresponding to its reference value V pref. [4]. Fig. 2. Generator SecVC scheme. Fig. shows the overall structure of the three voltage control levels in a power system with a zoom in SecVC configuration. The three voltage control levels are constructed in such a way that each level takes a time less than that of the higher level. In other words the inner loop is faster than the outer ones to avoid the overlap between each Vp Vpref Vp (3) t dvp q KpVp KI VpdtK (4) D dt. where K P, K I, K D are the parameters of the coordinated proportional integral derivative controller. The V pref is given originally by optimal load flow calculations to achieve a certain objective function (such as minimum load shedding or minimum power losses or even optimal reactive power reserve) at a certain condition with restriction to some constraints as active and reactive power limits and also bus voltage limits. In this paper V pref is assumed to be per unit.

4 III. SYSTEM DESCRIPTION This research is applied to the IEEE 4 bus system after performing some modifications. The test system used in Fig. 3 consists of eight PQ buses and five PV buses. The idea of the SecVC is that the reactive power sources at the PV buses should support the loads in PQ buses in case of any disturbance. The reactive power sources installed in the IEEE 4 bus network are generators at buses and 2, STATCOM at bus 6 and synchronous condensers at buses 3 and 8. Each generator has its own AVR based on its type and size; typical examples are shown in Fig. 4 and Fig. 5. Fig. 3. Modified 4-Bus System. Fig. 4. AVR for some steam units. Fig. 5. AVR for some hydraulic and nuclear units []. IV. MINIMUM NUMBER OF PMU S FOR VOLTAGE MEASUREMENTS At the end of the past century, supervisory control and data acquisition (SCADA) system were widely used to transfer power information (such as power flow computations, bus voltage measurements, etc.) from point to point [7]. The problem of the SCADA is that it needs to be connected point to point which means transferring information over large distances is very difficult, also now days with the aim to convert the power system to smart grid a real time synchronized monitoring, measurements are needed to perform real time control and protection. PMUs were first introduced in last 98s are now one of the main components of any smart grid as it allows having wide area of measurements at same time and capturing the power system data at real time. Measurements which is transferred by using global positioning system (GPS) with accuracy better than micro second, enables WAMPAC system to provide real-time measurements of bus voltage and branch current phasors [8]. Installing PMU in each bus will give accurate real time measurements. The main problem of the PMU is that still expensive so the idea is to keep the entire grid observable by locating the PMUs in proper buses. Accessing the optima number of PMUs to have a complete observable power network requires a suitable optimization technique which has been widely handled by several researchers in recent years. In this research sequential linear programming optimization technique is used to get the minimum number of PMU s to be used as a feedback control on the load buses [9]. The optimization problem is defined as follows: - Objective function: minimizing the number of PMUs n (Min. i w ix i ) Where n is the number of buses. - Variables: Buses include PMU s - Constraints: Each bus should be measured by at least one PMU ( f ( x) ) (PMU can measure the voltage of the installed bus and the buses which are directly connected through transmission line or transformer to the installed bus). In this paper the optimization was made on the 4 bus network shown in Fig. 3. After performing the optimization on MATLAB, we set the PMU s at buses 2, 6, 8, and 9. The PMU plays an important role in the proposed SecVC not only to be used as feedback signal with the actual value of the pilot bus voltage magnitude but also will be used as an indicator for the contingency element in the network. Each contingency will

5 have its own SecVC controller design which will be discussed in section V. To facilitate the optimization the network is converted to a spiral shape as shown in Fig. 6. Fig Bus system spiral shape. V. GENETIC ALGORITHM During the last decades there has been a growing interest in problem solving system based on principles of evolution and hereditary. GA is a type of evolution-based computer program to search for the fit solution of a particular problem []. Genetic algorithms are attractive techniques that employing natural evolution to solve problems in a wide field of complex applications. The GA is an iterative optimization technique, working with a number of candidate solutions (known as a population). If knowledge of the problem domain, is not available a priory, the GA begins its search from a random population of solutions []. A suitable coding (or representation) for the problem to be solved, must be defined first. In addition, a fitness function should be defined to assign a figure of merit to each coding solution. During the run, if the termination condition is not satisfied, parents must be selected for reproduction. Then, these are recombined to generate offspring using the reproduction, crossover and mutation operators to update the population of candidate solutions. Usually, a simple genetic algorithm consists of three operators: selection, crossover, and mutation. The application of these operations produces new strings (offspring), new population, and new generation, respectively. The overall process is then repeated with the new generation until the appropriate criterion is clearly satisfied. The GA applied in this work is performed by using the double vector population type with population size of 2. The Elite count reproduction is 2 and the crossover fraction is.8. VI. APPLICATION OF THE COORDINATED SECVC The SecVC control scheme shown in Fig. 2 has constant parameters all of them can be calculated easily from the network except the coordinated PID controller parameters which should be designed to achieve the objective of the SecVC, i.e. to reach the reference value of the load bus at a time range longer than that of the AVR at any disturbance. In this paper a steady state disturbances is chosen to test the SecVC control scheme. In each steady state disturbance which will be a contingency of an effective element in the network as the generator or the transmission line, the load bus coordinated PID should be designed to reach the reference under that condition. This means that under each contingency the design of PID controller parameters will be changed to achieve the desired reference at a certain time. Moreover the design will be varied also if the reactive power source which will support the load bus is changed or even increased. The design of the PID controller parameters will be made for each condition separately using genetic the algorithm tool box in MATLAB. The optimization problem is described as follows: - Objective Function: Minimizing integration of square error (ΔV p ) - Variables: PID parameters - Constraints: V GS and q limits. After finishing the PID controller design for each contingency and simulate this design in each case to show its effectiveness on the network, a what-if analysis should be created for all the possible contingencies and the coordinated PID controller design for each case. The scenarios should be stored in an artificial intelligence element (as neural network) which should get the optimal values of the PID controller to reach the desired voltage magnitude at certain contingency with certain reactive power source(s). The PMU reading will be the input of the neural network based on the stored condition inside it; the optimal values of the PID designed for the entire contingency should be set as shown in Fig. 7. Fig. 7. Neural Network used in SecVC. The neural network used in this work consists of two inputs, three outputs and a hidden layer having neurons. The network was trained by the well-known Lavenberg- Marquardt backpropagation method. VII. SIMULATION RESULTS In this paper load bus 5 is chosen to be the controlled bus at different tests. The aim is to let voltage magnitude of bus 5

6 Voltage magnitude in PU Voltage magnitude in PU voltage in pu Voltage in PU reaches. pu (as assumed in this research) under different disturbances by the reactive power supporters surrounding bus 5. Test : A contingency is to trip the generator installed at bus. The generator which is installed at bus 2 will provide the support to load bus 5. The PID controller parameters designed by GA at this condition are set to be K P =.74, K I =.36 and K D =.24. The GA takes 22 iterations to converge. Fig. 8 shows that the SecVC has returned the bus 5 voltage to its preset steady-state value (. pu), while using AVR only will result in a significant steady state error in the voltage. Test 2: Now, we assume a contingency of the generator at bus 2 leading to tripping this generator. The generator at bus will function to support voltage at load bus 5..4 Voltage at bus 5 support to load bus 5. The PID controller parameters which were designed in Test are used here. Fig. shows the voltage response at load bus 5. The voltage response is unsatisfactory. It transiently exceeds.4 pu and still away from the desired value more than s. Test 4: A contingency as Test takes place and leading to tripping the generator at bus. Now both the generator which is installed at bus 2 and the STATCOM installed at bus 6 will work together to provide support to load bus 5. The PID controller parameters designed by GA at this condition are: K P =.56, K I =.52and K D =.62. The GA requires 9 iterations to converge..5 Control without ANN.2 SecVC AVR time in sec Fig. 8. Load bus-5 voltage magnitude when the network subjected to test. The PID controller parameters designed by GA at this condition are set to be K P =2.74, K I =.56and K D =.874. The number of iterations taken by the GA is 7..4 Bus 5 Voltage magnitude Time in sec Fig.. SecVC to bus 5 without neural network Time in sec Fig.. Load bus 5 voltage response when the network subjected to Test Time in sec Fig. 9. Load bus 5 voltage response when the network subjected to Test 2. Test 3: A contingency 2 is to trip the generator installed at bus 2. In this case the generator installed at bus will do the The response in Fig. shows that using more than one reactive power source improves the performance of the SecVC. Compared to the response in Fig. 8, the response in Fig. has a lower undershoot value and the overshoot value is almost eliminated. Moreover it gives the chance to have more reactive power reserve.

7 VIII. CONCLUSIONS This paper investigates the coordinated secondary voltage control by reactive power sources to eliminate the steadystate voltage violations in power system contingencies. The parameters of the coordinated PID controller differ from disturbance to another. Therefore, a neural network has been used to store the controller parameters of each event. The PMUs are essential in the proposed work to provide real time voltage measurements to be used as inputs to the neural network, which provides appropriate values of the PID controller parameters. The simulation studies have shown that the proposed secondary voltage control scheme provides improved voltage regulation and allows more effective utilization of available reactive power sources. REFERENCES [] Prabha Kundur, John Paserba, Venkat Ajjarapu, Göran Andersson, Anjan Bose, Claudio Canizares, Nikos Hatziargyriou, David Hill, Alex Stankovic, Carson Taylor, Thierry Van Cutsem, and Vijay Vittal, Definition and classification of power system stability, IEEE Trans. Power Systems, July 24. [2] Bo Hu, Secondary and tertiary voltage regulation based on optimal power flows, 2 IREP Symposium - Bulk Power System Dynamics and Control VIII (IREP), Buzios, RJ, Brazil, Aug. -6, 2. [3] Qinglai Guo, Hongbin Sun, Mingye Zhang, Jianzhong Tong, Boming Zhang, Bin Wang Optimal voltage control of PJM smart transmission grid: study, implementation, and evaluation, IEEE Trans. Smart Grid, vol. 4, no. 3, Sept. 23. [4] Sandro Corsi, Massimo Pozzi, Carlo Sabelli, Antonio Serrani, The coordinated automatic voltage control of the Italian transmission grid Part I: Reasons of the choice and overview of the consolidated hierarchical system, IEEE Trans. Power Systems, vol. 9, no. 4, Nov. 24. [5] Yi Su, Chih-Wen Liu, An Adaptive PMU-Based Secondary Voltage Control Scheme IEEE Trans. Smart Grid, vol. 4, no. 3, Sept. 23. [6] Coordinated voltage control in transmission network, CIGRE Task Force, C4.62, Tech. Rep., Feb. 27. [7] A. Bose, Smart transmission grid applications and their supporting infrastructure, IEEE Trans. Smart Grid, vol., no., pp. 9, June 2. [8] Mahari, A., Seyedi, H. Optimal PMU placement for power system observability using BICA, considering measurement redundancy Electrical Power System Res. 3, 78 85, 23. [9] Chakrabarti, S., Kyriakides, E., Eliades, D.G Placement of synchronized measurements for power system observability, IEEE Trans. Power Delivery, vol. 24, no., pp. 2 9, 29. [] Omar H. Abdalla, Wedad M. Refaey, Mohamed K. Saad,Gamal Sarhan, Coordinated design of power system stabilizers and static VAR compensators in a multimachine power system using genetic algorithms, Proc. of the 6th International Conference on Electrical Engineering (ICEENG), Military Technical College, Cairo, Egypt, May, 28. [] D. Dracopoulos, Genetic algorithms and genetic programming for control, in Dasgupta and Michalewicz (eds.), Evolutionary Algorithms in Engineering Applications, Spriger-Verlag, pp