Localization of FACTS Devices for Optimal Power Flow Using Genetic Algorithm

Transcription

1 13 Internatonal Conference on Electrcal Informaton and Communcaton Technology (EICT) Localzaton of ACTS Devces for Optmal Power low Usng Genetc Algorthm A.K.M. Rezwanur Rahman, Md. Shahabul Alam, Md. Zakr Hossan * and Md. Shahjahan Dept. of Electrcal and Electronc Engneerng, Khulna Unversty of Engneerng & Technology, Bangladesh * E-mal: zakreee6@yahoo.com Abstract Ths paper presents about the effectve localzaton of the ACTS (lexble AC Transmsson Systems) devces n power system by a global search GA (Genetc Algorthm) technque. Ultmate goal s to mprove the stablty of power system as well as to reduce the generaton cost, transmsson losses by ncreasng loadablty and mprovng voltage profle wth ntroduce ACTS devce at the most effectve regon. Ths method s employed by consderng the cost of ACTS and ther optmal utlzaton n the system. Optmal locaton of ACTS, ther types and rated values are optmzed smultaneously. Three ACTS devces such as TCSC, TCPAR and SVC are smulated n ths study. Smulaton s carred out on IEEE3 bus and IEEE 118 bus power system wth dfferent ncreased load-ablty. We search the effcency of ths method on the bass of power generaton cost, ACTS nvestment cost and transmsson loss reducton. The employed algorthm s emerged as an effectve and practcal method for the choce and allocaton of ACTS n large power systems. Keywords ACTS devces, Genetc Algorthm (GA), Optmal Power low (OP), Improved loadablty, Load low. I. INTRODUCTION In recent years, deregulaton of electrcty has emerged for ts huge demand. Due to the deregulaton of the electrcty market, study regardng ths matter has become mperatve. Varous ntatves are taken to overcome that, but utlzaton of ACTS devce attracts everyone s attenton. ACTS devces are beng played vtal role for better utlzaton of the exstng power system wth the ncreased demaned [1], []. On the other hand, transmson and dstrbuton orentaton has become more severe due to the lack of proper arrangement. The major power loss occurs for system loss whch s ncreasng day by day around the world and has emerged as a challange for the developng countres to run wth lmted resources. To mnmse ths transmsson power losses and ensure optmal power flow, ACTS s ntroduced n power system. Dfferent parameters and varables of the transmsson lne such as lne mpedance, termnal voltages and voltage angle can be controlled by ACTS devces n a fast and effectve way [3]. Varous types of ACTS devces such as Thyrstor Controlled Seres Compensatons (TCSC), Thyrstor controlled phase angle Regulators (TCPR), Unfed Power low Controllers (UPC) and Statc Var Compensator (SVC) etc, are used to control the power flow n the network. These ncrease the flow n heavly loaded lnes, there by resultng n ncrease load ablty, lowerng system losses, mproved stablty of network and reduced cost of producton [4]. Although ACTS devce has an great mpact on power system for optmal power flow, but t requres optmal alocaton for proper stablzton and localzaton. or that reason, many researches were made on the optmal locaton of ACTS devces wth many dfferent ways. We use GA technque to search the optmal localzaton of ACTS devces. Genetc algorthms (GA) s a parallel and global search technque [5], [6], whch generate solutons to optmzaton problems usng natural evoluton, such as nhertance, mutaton, selecton, and crossover. It s more lkely to use for convergng toward the global soluton because t evaluates many ponts n the parameter space smultaneously. GA dffers from other optmzaton and search procedures n four ways [7], such as () t can easly handle the nteger or dscrete varables because of codng of the parameter set, not the parameters themselves, () t searches wthn a populaton of ponts, not a sngle pont whch may provde a globally optmal soluton, () t can deal wth the non-smooth, non-contnuous and no dfferentable functons whch are actually exst n a practcal optmzaton problem, because t utlzes only objectve functon nformaton, not dervatves or other auxlary knowledge, (4) t uses probablstc transton rules, not determnstc rules. Although GA seems to be a good method to solve optmzaton problem, but sometmes the soluton obtaned from GAs s only a near global optmum soluton. In ths paper general GA s appled to mprove stablty, voltage profle of power system as well as to reduce transmsson and generaton losses usng three types of ACTS devces. In [1], locaton of ACTS devces are found on the bass of two parameters - overload and over voltage whle n [11], only nstalaton and generaton costs are consdered. Our optmzaton ncludes all of above parameters. Moreover, reducton of generaton cost, transmsson losses, mproved loadablty and system stablty were consdered n the objectve functon of the GA for better mprovement of the power system whch are not consdered altogether by prevous work. IEEE 3 and IEEE 118 bus of power system are used for smulatons. The rest of the papers are organzed as follows. Secton II descrbes about optmal power flow where subsectons A and B analyzes about optmzaton and analyss respectvely. ACTS devces are descrbed n secton III where generaltes & choce and modelng are explored n subsectons A & B accordngly /13/$ IEEE

2 We descrbed about GA n secton IV. Results are optmzed n secton V. nally we conclude the work n secton VI. II. OPTIMAL POWER LOW A. Power low Optmzaton Optmal power flow (OP) s a nonlnear programmng problem whch s also called load flow analyss, s very gure 1. Confguraton of 4-bus power system. mportant for analyss of an electrcal system. By analyss ths the characterstcs of the power flow drecton n dfferent lne and the voltage at dfferent buses can be easly obtan [8]. g. 1 shows an example of confguraton of 4-bus power system where two bus has alternator to generate power and all bus share ths power wth dfferent demand.where red and blue ndcate peak and off peak demand respectvely. Due to ncrease n sudden load some buses are overloaded and some are wth lag of load whch causes dfference n bus voltages and power flow n transmsson lne. So we need to calculate overall generaton of power, cost, transmsson loss, voltage stablty n that tme n an economc way. or that reasson power flow analyss s necessary. B. Optmal Power low Analyss The lne resstance s small compared to the reactance and transverse capactance s close to zero for ntereconected power system network that obyes the Krchoffs law. we consder only lne reactances for nterconecton between and j bus whch s shown n g.. P j s the real power flow and Q j s the reactve power flow between two buses by a lne s related by the followng equatons: Q VV j j = snθj, (1) X j P ( V VV cosθ ) 1 j = j j, () X j Where, V and V j voltages at buses and j respectvely, X j reactance of the lne, j angle between V and V j. Here X j & j controls real and reactve power. We analyze these by Matlab power smulaton package Matpower 4.1 [9]. X j gure. Bus and bus j s conected by lne wth reactance Xj. j III. ACTS DEVICES A. Introduce wth ACTS Devces The ACTS s power electroncs based devce whch s used for AC power transmsson and dstrbuton. Due to rapd response, ablty for frequent varatons and smooth adjustablty wth output the applcatons of ACTS devces are ncreased day by day [1]. Dfferent types of ACTS devces are used for power transmsson and dstrbuton such as SVC (Statc Var Compensators), xed Seres Capactors (SC), Thyrstor- Controlled Seres Compensator (TCSC), Thyrstor Controlled Phase Angle Regulator (TCPAR), Unfed Power low Controller (UPC), STATCOM etc. Power qualty, avalablty, system stablty, transmsson capablty can be mproved usng ACTS as well as t mnmzes transmsson and dstrbuton losses. Here just TCSC,TCPAR and SVC are used for the optmal power flow analyss. B. Choce and Modelng Three dfferent types of ACTS devces have been chosen for the controllng of power flow. These are TCSC, TCPAR, and SVC. TCSC s used to modfy the reactance of the transmsson lne X j. or controllng the phase angle j the TCPAR s used. The SVC s used to absorb or nject reactve power whch s connected n shunt wth the lne. Both the TCSC and TCPAR are connected n seres wth the lne. Each ACTS devce s represented wth fxed dscrete values for mathematcal analyss, where t has two possble characterstcs, capactve or nductve accordngly n order to decrease or ncrease the lne reactance, phase angle, reactve power n lne usng TCSC, TCPAR and SVC respectvely. Maxmum and mnmum value of each ACTS devce s fxed and type of each devce s also specfed. TABLE I represents the specfcaton of ACTS devces. g. 3 shows the connecton model of ACTS devces. TABLE I. SPECIICATION O ACTS DEVICES Name/Specfcaton TCSC TCPAR SVC Devce type 1 3 Mnmum value -.8 X L - 5 deg. -1 MVar (Capactve) Maxmum value. XL (Inductve) +5 deg. +1 MVar (a) (b) (c) gure 3. Connecton models of ACTS devces. (a) TCSC (b) TCPAR (c) SVC The real value of the ACTS devce [1] V real s calculated wth there loccaton accordng to the model of the ACTS by V ) real = Vmn + ( Vmax Vmn V (3)

3 Where, the normalzed value s V, the maxmum and mnmum settng value are V max and V mn respectvely. IV. GENETIC ALGORITHM A. GA overvew GA was proposed on the bass of the evolutonary deas of natural selecton and genetcs [5], [6]. It represents an ntellgent explotaton of a random search used to solve optmzaton problems whch s a rapdly growng area of artfcal ntellgence (AI). Because of ndependent of the choce of the ntal confguratons GA fnds hgh qualty solutons. Moreover, they are computatonally smple and easy to mplement. Genetc dversty or varaton s confgured usng dfferent operators such as reproducton (selecton), crossover (recombnaton) and mutaton. B. Model of GA The man objectve of the optmzaton s to fnd the best locaton for a gven number of ACTS devces n the power system based on defned crteron. Three parameters are utlzed for encodng ndvdual these are the locaton, type of devce and rated value [1], [1]. Each ndvdual s represented by n number of three strngs, where n s the number of ACTS devces nstalled n the power system. TABLE II shows the ndvdual format. Indvdual s made n three stages, frst a set of branches are randomly selected and s put n the frst strng. In the second strng type s also randomly selected. In the thrd strng devce settng value s randomly selected. Ths approach s repeated for obtanng desred populaton. Then the entre populaton s computed wth respect to objectve functon whch s the measure of obtanng best locaton for the ACTS devce. chromosomes to be mxed at the gene level. Consder the two parents selected for crossover. If the mxng rato s about.5, then half of the genes n the offsprng wll come from parent 1 and rest from parent. Then boundary mutaton s appled. g.4. shows the unform crossover and boundary mutaton technque clearly. Crossover and mutaton s done after selecton of parents. Blue and black colors represent parent 1 and parent respectvely. Then we obtan offsprng accordng to crossover of parents and mutaton. The red color shows the mutaton results where other for crossover. gure 4. Crossover and Mutaton approach. The entre methodology of GA s explaned n g. 5. rstly ndvduals are selected randomly. Then ftness value s calculated for each ndvdual based on the ftness functon. Best ndvdual s found accordng to selecton, crossover and mutaton when fnal crteron s reached. Start TABLE II. INDIVIDUAL ORMAT OR GA Locaton ACTS Type Normalzed value N ndvdual generatons tness to each Indvdual Selecton, Crossover Mutaton New ndvdual s generated based on the results obtaned from the old generaton. or ths 1 st GA operator selecton s used. In ths case Proportonal Roulette Wheel Selecton technque s used. In proportonal roulette wheel, ndvduals are selected wth a probablty that s drectly proportonal to ther ftness values.e. an ndvdual selecton corresponds to a porton of a roulette wheel. Let be the ftness value and P be the selecton probablty, then P = N = 1 Based on the selecton probablty P, ndvdual s randomly selected by roulette wheel. After that nd GA unform crossover s appled [5]. Unform crossover wth some probablty knows as the mxng rato. The crossover operator allows the parent (4) Yes Stoppng crteron? Best Indvdual gure 5. low chart of the optmzaton strategy. V. OPTIMIZATION AND RESULT A. Objectve functon The am of the optmzaton s to fnd the best locatons for the gven number of ACTS devces wthn the defned No

4 constrans for the best utlzaton of the exstng system. We want to mnmze the power generaton costs and reduce the transmsson and dstrbuton losses. So the objectve functon s based on the mnmzaton of cost whch can be expressed as C Total C1( f ) + C( PG) + C3( PL) =, (5) Where, C Total, C 1 (f), C (PG), C 3 (PL) are the total cost of objectve functon, average nstallaton costs of ACTS devces at each observaton per hour, total generaton costs and cost of power transmsson losses respectvely. The cost functons for SVC, TCSC and TCPAR are developed on the bass of the Semens AG Database [8], [15] The cost functon for SVC and TCSC are: C svc.3s.351s ( US$/ K var) C t.1s.713s ( US$/ K var) = (6) csc = (7) Where S s the operatng range of the ACTS controllers n kvar. Dependng on the nstallment cost, the cost functon of TCPAR can be expresed as C tcpar = 14.5( US$/ K var) (8) The cost functon for SVC, TCSC and UPC from Semens AG Database s shown n g. 6. C The cost functon for power loss s represented as N 3 ( PL) = = PL * Eloss * dt (11) 1 Where, N, PL, Eloss and dt are denotes the number of used ACTS devces, transmssn losses, cost of the losses n per hour and ACTS devces utlzaton tme respectvely. Now the ftness functon for the genetc algorthm s found as tness=1/c Total (1) B. Results Accordng to varaton of ftness functon ndvduals are generated usng GA to optmze the power flow. The smulaton s carred out by free Matlab power smulaton package Matpower 4.1 [9]. Based on the GA, best fttest ndvdual s found for the optmal power flow n IEEE 3 and IEEE 118 bus power system wth ncreased amount of demand. Reducton of the power loss and mprovement of the voltage profle durng transmsson are ntroduced here those are shown n gs. 9 and 1 accordngly. gs.7 and 8 show the ftness value of the ftness functon wth respect to generaton for IEEE 3 and IEEE 118 bus power system respectvely. 8.5 x 1-4 tness value Generaton Number gure 7. tness functon curve wth generaton for IEEE 3 bus system gure 6. Installaton cost curve Now f C(f) s the summeton of the used ACTS devce nstalaton cost. The generaton cost s calculated n per unt that s US$/Hour and the nstallaton costs of ACTS devces are n US$. or that reason lfe tme of the ACTS s consdered. In ths paper, three years s appled to evaluate the cost functon [1], [11]. We calculate the average values of the nstallaton costs usng the followng equaton, where 876 s the total hour n a year. C( f ) C1( f ) ( US$/ Hour) = (9) The generaton cost functon s represented by a quadratc polynomal as follows: C ( PG) = α + α1pg + α ( PG) (1) Where PG s the output of the generator (MW), and, 1 and are cost coeffcents. tness Value 5.5 x Generaton Number gure 8. tness functon curve wth generaton for IEEE 118 bus system. g. 9 shows both for the IEEE 3 bus and IEEE 118 bus that before usng ACTS devce power loss through lne s more whch s showed wth red mark. After usng ACTS devce power loss through lne reduced and t s showed by blue mark. Although n some case lttle ncreasement of power loss after usng ACTS devce but t s neglgble. So overall

5 performance s much better after usng ACTS devce. It has found that obtaned locaton s showng satsfactory output. Power loss n MW Power loss after usng ACTS devce Power loss before usng ACTS devce Branch Number Power loss n M W (a) Before usng ACTS devce After usng ACTS devce Branch Number (b) gure 9. Comparson of power loss before and after usng ACTS devce n (a) IEEE 3 bus (b) IEE118 bus. g. 1 shows both for IEEE 3 and IEEE 118 bus t s found that voltage magntude (VM) profle s better whle usng ACTS devce than wthout ACTS devce. As voltage magntue sholud stay n the lmt 1.5 and.95 per unt whch s marked by color green and black, wthout ACTS devce VM s marked by red color whch s very poor. On the other hand after ACTS devce utlzaton voltage profle s marked by blue and t s more stable. Optmal locatons n the power system are detected by usng genetc algorthm for the ACTS devce. At the same tme specfed ACTS devce wth specfed value whch s hghly effectve for optmal power flow s too determned. After applyng these obtanted outcome n the power system, fnal optmal power flow s observed n the IEEE 3 and IEEE118 bus system. In the IEEE 3 and IEEE118 bus system t was observed that transmsson loss reduced n the system after nstallng ACTS devce durng certan ncrease n the load n the system. It was also too observed that voltage profle at each bus of the system mproved after nstallng ACTS devce. Voltage Magntude Voltage Magntude Voltage Magntude before usng ACTS devce Voltage Magntude after usng ACTS devce Upper voltage lmt 1.5 p.u Lower vlotage lmt.95 p.u Bus Number (a) Voltage magntue before usng ACTS devce Voltage magntue after usng ACTS devce Upper voltage lmt 1.5 P.U Lower voltage lmt.95 P.U Bus Number (b) gure 1. Voltage Magntude profle comparson before and after usng ACTS devce n (a) IEEE 3 bus (b) IEEE118 bus VI. CONCLUSION A genetc algorthm has been presented wth larger parameters than prevous methods to optmally locate ACTS devces n the power system. Here only three types of ACTS devces are used and smulaton s carred out on IEEE 3 bus and IEEE 118 bus. After smulaton ACTS devces are used n the obtaned locaton and the power flow of the system s observed. It s found that power transmsson loss has reduced for usng ACTS devces n case of IEEE 3 bus for 9% cases and n case of IEEE 118 t s about 75%. So transmsson losses are reduced. In case of bus voltage profle, about 95% bus voltage reman wthn the lmt due to use of ACTS devce n IEEE 3 bus and n case of IEEE 118 bus t s about 8%. So t can be sad that overall ACTS devce has a great mpact n power system for optmal power flow and n that case Genetc algorthm exhbts a great mpact for selectng the perfect locaton. REERENCES [1] S. Gerbex, R. Cherkaou and A. J. Germond, "Optmal locaton of mult-type ACTS devces n a power system by means of genetc algorthms," IEEE Trans. Power Systems, vol. 16, pp , August. 1. [] K. Ghoshand and V. C.Ramesh, An opton model for electrc power markets, Electrcal Power and Energy Systems, vol.19, no., [3]. D. Galana, K. Almeda, M. Toussant, J. Grffn and D. Atanackovc, Assessment and control of the mpact of ACTS devces on power system performance" IEEE Trans. Power Systems, vol. 11, no. 4, Nov

6 [4] D. J. Gotham and G. T. Heydt, Power flow control and power flow Studes for systems wth ACTS devces, IEEE Trans. Power Systems, vol.13, no.1, eb [5] R. C. Chakrabarty. undamental of Genetc Algorthms AI course, Lecture39-[Onlne],Avalale: http// [6] X. P. Wang and L. P. Cao, Genetc Algorthms Theory, Applcaton and Software Realzaton, X an Jaotong Unversty, X an, Chna, [7] T. S. Chung and Y. Z. L, A hybrd GA approach for OP wth consderaton of ACTS devces, IEEE power engneerng Revew, pp , ebruary, 1. [8] I. O. Elgrd, Electrc Energy System Theory- An Introducton, McGraw Hll Inc., New York, [9] R. D. Zmmerman, C. E Murllo-Sanchez and R. J. Thomas, MATPOWER: Steady State Operatons, Plannng & Analyss. Tools for Power Systems Research & Educaton, IEEE Transactons on Power Systems, vol. 6, no.1, pp.1-19, eb. 11. [1] N. G. Hngoran and L. Gyugy,. Understandng, ACTS - Concepts and Technology of lexble AC Transmsson Systems. IEEE, 1999 [11] L. Ca, I. Erlch, G. Stamtss and Y. Luo, Optmal Choce and Allocaton of ACTS Devces n Deregulated Electrcty Market usng Genetc Algorthms, Conf. on Bulk Power System, Daynamcs & Control-VI, August -7, 4.