N.EL.Y. Kouba, M. Menaa, M. Hasni, B. Boussahoua, and M. Boudour

Transcription

1 Internatonal Conference on Renewable Energes and Power Qualty (ICREPQ 14) Cordoba (Span), 8 th to 1 th Aprl, 214 exçxãtuäx XÇxÜzç tçw céãxü dâtä àç ]ÉâÜÇtÄ (RE&PQJ) ISSN X, No.12, Aprl 214 Automatc Generaton Control n Interconnected Power System wth Integraton of Wnd Power Generaton Usng PID Controller Based on Partcle Swarm Optmzaton N.EL.Y. Kouba, M. Menaa, M. Hasn, B. Boussahoua, and M. Boudour Laboratory of Electrcal Industral Systems, Unversty of Scences and Technology Houar Boumedene, El Ala, 16111, Bab Ezzouar, Algers, Algera, e-mal: koubanour@gmal.com, mmenaa@usthb.dz, hasnmourad21@yahoo.fr, bbouzane72@yahoo.com, mboudour@eee.org Abstract. Today automatc generaton control s becomng more mportant and sgnfcant n nterconnected power system desgn and operaton due to the complexty of power systems, changng structure, and emergng renewable energy sources. One of major functons of automatc generaton control s the load frequency control. Ths paper presents an applcaton of partcle swarm optmzaton (PSO) for determnng the optmal values for the proportonal-ntegral-dervaton (PID) controller for a load frequency control (LFC) of two area nterconnected power system wth ntegraton of wnd power generaton, ths method s compared to the tradtonal Zegler-Nchols method. The transent responses are shown due to mpact of ntegraton of wnd power n area-2. The man prmary objectve s to reduce the fluctuatons of the system frequency and te lne power flow. Key words Load Frequency Control (LFC), PID Controller, Wnd Power Generaton, Zegler-Nchols method, Partcle Swarm Optmzaton (PSO). 1. Introducton In recent years, power system stablty and control s recognzed as an mportant problem. In modern power system, the automatc generaton control (AGC) or load frequency control (LFC) s becomng more sgnfcant. In early 197 Fosha and Elgerd n ther poneerng work appled classcal optmal control methodology to solve LFC problems [1]. The use of renewable energy such as wnd energy was attracted because fossl fuels wll be exhausted n the future, and ts use causes the problem of the envronment (CO2 emsson). The generaton of wnd energy s not stable and can not ensure the constant power generaton because t depends on the wnd as a natural source. However, f a large amount of wnd generaton s nstalled, t becomes dffcult to remove fluctuatons n system frequency. Unbalance between supply and load causes the fluctuaton of the system frequency, whch can degrade the power system performance and makes hs control more dffcult. As major functons of automatc generaton control (AGC), power system frequency regulaton named load frequency control (LFC) has been one of the mportant control problems n electrc power system desgn and operaton. A large frequency devaton degrades the system performance and can damage equpments [2-3]. The prncpal prmary goals of load frequency control s to suppress the fluctuaton of the system frequency and to mantan frequency and power nterchanges wth neghbourng control areas at the scheduled values. To satsfy these objectves, a control error sgnal called the area control error (ACE) s measured. Ths sgnal s a lnear combnaton of net nterchange and frequency devaton, and represents the real power unbalance between supply and load of power [2-4]. A proportonal-ntegral-dervaton (PID) controller s used to regulate the frequency of each area, dstrbuted among the LFC partcpant generator unts n accordance wth ther partcpaton factors to provde approprate control commands for set ponts of specfed plants [5-6]. The nput sgnal of controllers s ACE, whose parameters tunng are selected dependng on the control area characterstcs. PID controllers meet most of the 9% of ndustral needs [7-8]. In 1942, Zegler and Nchols proposed two heurstc approaches based on ther experence and some smulatons to quckly adjust the controller parameters P, PI and PID [9]. For tunng the optmal values for the PID controller parameters, we propose to use the PSO technque, ths technque s a stochastc search developed by Kennedy and Eberhart n 1995 [1], whch has been found to be robust and flexble n solvng optmzaton problem, because t can generate a hgh-qualty soluton wthn shorter calculaton tme and stable convergence characterstc than other stochastc methods [11]. In ths paper, load frequency control (LFC) of two-area nterconnected power system wth ntegraton of wnd power generaton usng optmal PID controller scheme based partcle swarm optmzaton (PSO) algorthm s appled and the proposed approach s compared to the classcal Zegler-Nchols method, and the mpact of ntegraton of wnd power generaton s analyzed.

2 2. Power System Modelng In ths study, IEE Japan East 17-bus-3-machne model system shown n Fg.1 (a) s used, ths system havng dverse sources of power generaton (hydro, thermal, nuclear). It s noted that the LFC sgnals are nputted to the governors of hydro and thermal unts, and the nuclear unt s operated wth a constant output power n Japan. Ths system s dvded nto two control areas, the nterconnected 2-area power system havng dverse sources of power generaton shown n Fg.1 (b) s composed of the governor model, the turbne model, the generator model, the LFC model, and the te-lne model. Each turbne output devaton s nputted to the generator model and the system frequency devaton s outputted. In order to analyze the system frequency, we asume a smulaton of a small control area wth a large penetraton of wnd power generaton nterconnected nto a large control area. For ths the wnd power plant s ntegrated nto area-2. Power system loads are a composte of a varety of electrcal devces. Some loads depend on the change of frequency and others don t. In general the expresson of the electrcal power whch depends on the change on the frequency can be expressed by [12-14]: Pe PL Dw (2) Where P L s Non-frequency-senstve load change, D s load-dampng constant and D w s Frequency-senstve load change. The block dagram form representaton of (1) and (2) s shown n Fg.2. Fg.2. Block dagram representaton of relatonshp between speed and power. In mult-generator system, f all generators are assumed turn wth the same speed of synchronsm, the block dagram representaton of equvalent generator model shown n Fg.3. (a) IEEJ East 17-bus-3 machne system. Fg.3. Block dagram representaton of equvalent generator. The equvalent nerta constant: The equvalent load-dampng constant: Where n s the number of generator. M D eq eq n M 1 (3) n n D 1 (4) n B. Governor Control System Model To brng the frequency back to the nomnal value each generator wth governor adjusts the turbne valve/gate (self regulaton). The schematcs of such governor control system that we used n ths work are shown n Fg.4 and Fg.5 [15-17]. (b) Smulnk model. Fg.1. Two-area nterconnected power system model. A. Generator Model The relatonshp between the mechancal power P m and the electrcal power P e s gven by [12-13]: dw M P m P e (1) dt Where w s rotor speed devaton and M s nerta constant. Fg.4. Block dagram of governor model for thermal unts.

3 Wth: T 2f X T 12 The block dagram representaton for te-lne power model s shown n Fg.8. Fg.5. Block dagram of governor model for hydro unts. C. Load Frequency Control (LFC) Model The man prmary objectve of LFC s adjusted operatng pont reference of governor unts n the control area and to set ther outputs. To evaluate the area requrement (AR) the actual frequency and net nterchange power flow are measured by the ndependent system operator (ISO). The area control error (ACE) s gven by: ACE P w (5) Each generator s generated by PID controller. The PID parameters are tunng wth two methods, frst wth the classcal method of Zegler-Nchols, and second wth the stochastc method (PSO), the dagram of the PID controller used s shown n Fg.6 [18-19]. The PID model s gven by: K I K( S) K P K DS (6) S Tj Fg.8. Block dagram representaton of te-lne power model. E. Wnd Power Plant Model In a modern power system the renewable energy such as wnd power plays an mportant role n the generaton of electrcal power. In ths study we use the load model proposed by T.Mchgam and T. Ish to represent a wnd farm [2]. The wnd power generaton profle s generated by a block dagram shown n Fg.9 [13] [2]. A load conssts of base and frnge components obtaned from the whte nose generators, snce the component wth perod shorter than 5 mnutes or longer than 3 mnutes s out of LFC control, the component s elmnated by applyng flters (HPF and LPF). Fg.6. Block dagram of PID controller model. In order to analyze the system frequency, the LFC model shown n Fg.7 s used n ths paper. Fg.7. Block dagram of LFC model. Fg.9. Block dagram of wnd power plant model. D. Te-Lne Model Usng DC load flow method the te lne power flow P T12 can be expressed by [12] [15]: 1 P T12 ( 1 2 ) (7) X T12 If we assume that the te-lne power flow s gong from area-1 to area-2. The devaton P T12 from the nomnal flow can be expressed: 1 P T12 ( 1 2) (8) X T12 Where 1 and 2 are equvalent to 1 and 2. 1 P T12 ( 1 2 ) (9) X T12 T P T 12 ( 1 2 ) (1) S 3. Zegler-Nchols Technque Over the 194s, many methods have been developed for obtanng the P, I, D controller parameters [5]. In ths work the techncal PSO tunng method s compared wth the tradtonal technque Zegler-Nchols. In 1942, Zegler and Nchols proposed two expermental approaches to quckly adjust the controller parameters wthout knowng the precse dynamc model of the system to adjust. Both methods are emprcal and based on tests. In ths paper we use the second method of Zegler-Nchols, t s a smple technque to tunng P, I, D controller parameters, Ths method requres the system to complete a smple proportonal controller Kp whch ncreases the gan to brng the system to oscllate permanently (Oscllatons mantaned), t s thus at the lmt of stablty. After notng the crtcal gan Kcr and

4 crtcal perod of oscllaton Tcr response, we can calculate the P, I, D controller parameters [9]. 4. Partcle Swarm Optmzaton Technque The partcle swarm optmzaton (PSO) s a heurstc optmzaton method based on swarm ntellgence. It comes from research on the brd and fsh flock movement behavor [1]. PSO s a populaton-based optmzaton method s developed n 1995 by Kennedy and Eberhart [5] [18]. The basc algorthm of PSO s gven by these 7 steps [21]: 1. Create a populaton of agents (called partcles) unformly dstrbuted over X. 2. Evaluate each partcle s poston accordng to the objectve functon. 3. If a partcle s current poston s better than ts prevous best poston, update t. 4. Determne the best partcle (the partcle s prevous best postons). 5. Update partcles veloctes accordng to: t1 t t V V C1rand1 Pbest X (11) t C2rand 2 gbest X 6. Move partcles to ther new postons accordng to: t1 t t1 X X V (12) 7. Go to step 2 untl stoppng crtera are satsfed. The general algorthm of PSO s shown n Fg.1 [21-23]. Usng the analytcal model mentoned n secton (2.E), the mpact of wnd power fluctuaton on the system frequency and Te lne power n two area systems are examned. The fluctuaton of wnd power generaton shown n Fg.11 s nputted nto area-2 (the maxmum devaton s about 5 MW), and the fluctuaton n system frequency and te-lne power flow are shown n Fg.12(a, b) and Fg.13 respectvely. wnd power output(mw) tme (s) Fg.11. Fluctuaton of wnd power generaton (MW). The mpact of LFC control methods on the fluctuatons caused by wnd power generaton s examned; for comparson purpose three cases are carred out, Table I. The parameters of PSO used n ths work are n Appendx B. TABLE I. - SIMULATION CASE Case Area 1 Area 2 1 Wthout LFC Wthout LFC 2 Wth PID (Z-N) Wth PID (Z-N) 3 Wth optmal PID (PSO) Wth optmal PID (PSO) The classcal PID parameters are gven n Table II. TABLE II. - PID PARAMETERS USING ZIEGLER-NICHOLS Fg.1. PSO algorthm. In ths paper, the PSO algorthm s used to attan the optmal PID controller parameters, and the proposed approach s compared to Zegler-Nchols method. 5. Smulaton Results The IEE Japan East 3-machne model system s consdered for the smulaton and the values of the dfferent parameters of the system are gven n Appendx A. crtcal gan Kcr crtcal perod Tcr (s) Kp K Kd PID The optmal PID parameters are gven n Table III. TABLE III. - OPTIMAL PID PARAMETERS (PSO) Kp K Kd

5 The mpact of ntegraton of wnd power generaton and LFC control methods s shown n Fg.12 (a, b) and Fg.13. Frequency devaton n Area-1(HZ) Frequency devaton n Area-2(HZ) Te-lne power flow devaton(mw) wthout LFC wth LFC (PID-Zegler-Nchols) wth optmal LFC (PID-PSO) tme (s) (a) Frequency devaton n Area-1 wthout LFC wth LFC (PID-Zegler-Nchols) wth optmal LFC (PID-PSO) tme (s) (b) Frequency devaton n Area-2 Fg.12. Impact of wnd power fluctuaton and LFC control methods n the system frequency wthout LFC wth LFC (PID-Zegler-Nchols) wth optmal LFC (PID-PSO) tme (s) Fg.13. Impact of wnd power fluctuaton and LFC control methods n the te-lne power flow. We see n Fg.12 (a, b) and Fg.13 when the penetraton of wnd power generaton s large, the fluctuatons of the system frequency and te-lne power are very mportant. However, the system frequency and the te-len power flow are suppressed most effectvely f both areas adopt LFC. Usng the PSO method, global and local solutons could be found smultaneously for a better agreement of the PID controller parameters, and we can see wth optmal PID parameters the fluctuaton of the system frequency and the te-len power flow are better suppressed compared wth the results gven by the Zegler-Nchols method. The results are compared and the proposed approach (PSO) s proven to be better as shown n Table IV. Tunng PID controller technques Max frequency devaton n Area-1 [HZ] Max frequency devaton n Area-2 [HZ] Max te-lne power flow devaton [MW] 6. Concluson TABLE IV. - RESULTS AND COMPARISON Wthout Control Zegler- Nchols PSO In ths paper a load frequency control (LFC) of two area nterconnected power system wth dverse sources of power generaton s studed. The mpact of wnd power generaton on the Te lne power and the system frequency are analyzed. The mpact of LFC control method on the fluctuatons caused by wnd power generaton s also examned. An applcaton of partcle swarm optmzaton (PSO) for determnng the optmal values for the proportonal-ntegral-dervaton (PID) controller for a load frequency control (LFC) s appled here. It s clear that wth optmal PID parameters the fluctuatons of the system frequency and te-lne power flow are suppressed most effectvely. The results are compared to the Zegler-Nchols method and the proposed approach has proven to be very effcent. APPENDIX.A The data of IEEJ 3 machnes. Table A.I. - Thermal Unt Parameters KG 2 Tsr.2 Tsm.2 Tsc.25 Trh / Tco.9 FHP.3 FIP / FLP.7 Table A.II. - Hydro Unt Parameters KG 2 Tgh 1 Tsv.3

6 Table A.III. - Parameters of the Two-Area Power System Area-1 Area-2 Reference Frequency 5 5 [HZ] Inerta constant [s] Load-dampng constant [pu] 2 2 APPENDIX.B Table B.I. - PSO Parameters Populaton sze 3 Intal weghts.9 Fnal weghts.4 Cogntve coeffcent (C 1 ) 1.6 Socal coeffcent (C 2 ).5 Crossover rate (CR).75 Maxmum teraton 3 Mnmum PID gan Maxmum PID gan 3 Mnmum velocty of -5 partcles Maxmum velocty of 5 partcles The Convergence characterstc of PSO algorthm s shown n Fg.B.1. Mnmum Values Convergence Characterstc of PSO Number of PSO Generaton References Fg.B.1. Convergence characterstc of PSO. [1] Mehd Rahman and Nasser Sadat, Two-level optmal load frequency control for mult-area power systems, Internatonal Journal of Electrcal Power and Energy Systems, 213, Vol.53, pp [2] Hassan.Bevran, Robust Power System Frequency Control, Sprnger, 29. [3] B.J.Krby, J.Dyer, C.Martnez, Dr.Rahmat A. Shouresh, R. Guttromson, and J. Dagle, Frequency control concerns n the North Amercan electrc power system, ORNL/TM- 23/41, December 22. [4] Sdhartha Panda and Narendra Kumar Yegreddy, Automatc generaton control of mult-area power system usng mult-objectve non-domnated sortng genetc algorthm-ii, Internatonal Journal of Electrcal Power and Energy Systems, 213, Vol.53, pp [5] Ranuva Nageswara.Rao and P.Rama Krshna Reddy, PSO based tunng of PID controller for a load frequency control n two area power system, IJERA, Vol. 1, Issue 3, pp [6] K. Soundarrajan, Dr.S. Sumath, Partcal Swarm Optmzaton Based LFC and AVR of Autonomous Power Generatng System, IAENG, Internatonal Journal of Computer Scence, 21, Vol.37. [7] H. Rasmussen, Automatc Tunng of PID-regulators, Aalborg Unversty, Dept. of Control Engneerng Fredrk Bajers Vej 7, DK 922 Aalborg, September 6, 22, Denmark. [8] Lalt Chandra Saka and Shash Kant Sahu, Automatc generaton control of a combned cycle gas turbne plant wth classcal controllers usng Frefly Algorthm, Internatonal Journal of Electrcal Power and Energy Systems, 213, Vol.53, pp [9] J.G.Zegler, N. B.Nchols, Optmum settngs for automatc controllers, ASME, Trans, Nov 1942, Vol.64, pp [1] James Kennedy, Partcle Swarm Optmzaton, Proceedng of the IEEE Internatonal Conference on Neural Networks, Perth,Australa, 1995, Vol.IV, pp [11] Qngha Ba, Analyss of Partcle Swarm Optmzaton Algorthm, Computer and Informaton Scence (CCSE), Feb.21, Vol 3, No 1, pp [12] A. J. Wood and B. F. Wollenberg, Power Generaton Operaton and Control, 2 nd ed., John Wley & Sons, New York, 1966, pp [13] Masash Arta, A. Yokoyama, and Y. Tada, evaluaton of battery system for frequency control n Interconnected power system wth a large penetraton of wnd power generaton, IEEE, Internatonal Conference on Power System Technology, 26. [14] P.Kundur, Power System Stablty and Control,McGraw- Hll, 1994, pp [15] P. S. R. Murty,Operaton and Control n Power Systems, 28, ch.6-7. [16] Behrooz Vahd, Mohammad Reza Bank Tavakol, and Wolfgang Gawlk, Determnng Parameters of Turbne's Model Usng Heat Balance Data of Steam Power Unt for Educatonal Purposes, IEEE Tans. Power Syst, Nov 27, Vol. 22, NO. 4, pp [17] 2.html. [18] Wu Dongsheng, Yang Qng, Wang Dazh, A Novel PSO- PID Controller Applcaton to Bar Rollng Process, Proceedngs of the 3th Chnese Control Conference, July 22-24, 211, Yanta, Chna. [19] E. Salm Al and S. M. Abd-Elazm, Optmal PID Tunng for Load Frequency Control Usng Bactera Foragng Optmzaton Algorthm, Proceedngs of the 14th Internatonal Mddle East Power Systems Conference (MEPCON 1), Caro Unversty, Egypt, December 19-21, 21, Paper ID 191,pp [2] T. Mchgam, T. Ish, Constructon of fluctuaton load model and dynamc smulaton wth LFC control of DC power system and frequency converter nterconnecton, n Proc. 22, IEEE Power Engneerng Socety Transmsson and Dstrbuton Conf, vol. 1, pp [21] Marco A. Montes de Oca, Partcle Swarm Optmzaton Introducton, IRIDIA-CoDE, Unversté Lbre de Bruxelles (U.L.B.) May 7, 27. [22] Dr.K.RamaSudha, V.S.Vakula., R.Vjaya Shanth, PSO based desgn of robust controller for two area load frequency control wth nonlneartes, Internatonal Journal of Engneerng Scence and Technology, 21, Vol. 2, pp [23] Yufe Tang, Png Ju, HaboHe, Chuan Qn, and Feng Wu, Optmzed Control of DFIG-Based Wnd Generaton Usng Senstvty Analyss and Partcle Swarm Optmzaton, IEEE transactons on smart grd, March 213, Vol. 4, NO. 1, pp