Kenya, P.O Box GPO

Transcription

1 Internatonal Journal of Emergng Technology and Advanced Engneerng Webte: (ISS , ISO 900:2008 Certfed Journal, Volume 3, Iue 7, July 203) Solvng The Actve Dtrbuton etwork Reconfguraton (ADR) Problem Takng Into Conderaton A Stochatc Wnd Scenaro and Load Uncertanty By Ung HBFDE Method Mr.Peter Muau Moe, Dr. codemu Abungu Odero 2 Pot Graduate Student, 2 Senor Lecturer, Department of Electrcal and Informaton Engneerng, The Unverty of arob- Kenya, P.O Box GPO Abtract--Pat lterature ha attempted to olve the problem of network reconfguraton wth Dtrbuted Generator (DG) wthout takng nto conderaton the ntermttent renewable at a cloe proxmty. Dtrbuton etwork Reconfguraton (ADR) mut account uncertan behavor of load and wnd when the commercal wnd baed DG, Doubly Fed Inducton Generator (DFIG) upport a gnfcant part of network. In th paper, a new Hybrd Bacteral Foragng and Dfferental Evoluton (HBFDE) algorthm condered for the ADR problem wth mnmum lo and an mproved voltage profle. In the HBFDE algorthm the Dfferental Evoluton (DE) algorthm combned wth the Bacteral Foragng (BF) algorthm to overcome low and premature convergence of BF. Indeed, the propoed algorthm baed on the evolutonary nature of BF and DE, to take ther advantage of the compenatory property, and avod ther correpondng drawback. In addton, to cope wth the uncertanty behavor of load and wnd, a tochatc model preented to olve the ADR problem when the uncertanty related to wnd and load forecat modeled n a tochatc framework on cenaro approach ba. The propoed algorthm teted on the IEEE 33-Bu Radal Dtrbuton Tet Sytem. The reult of the mulaton how the effectvene of propoed algorthm real tme and real world optmzaton problem facng the mart grd. Keyword-- Actve Dtrbuton etwork Reconfguraton (ADR), Bacteral Foragng (BF), Dfferental Evoluton (DE), Doubly Fed Inducton Generator (DFIG), Hybrd Bacteral Foragng and Dfferental Evoluton (HBFDE). I. ITRODUCTIO The confguraton of electrc dtrbuton network motly a radal for proper protecton coordnaton. Due to ome obectve uch a: upply all of the load, reduce power lo, ncreae ytem ecurty and enhance power qualty; the confguraton of thee network may be changed wth automatc or manual wtchng operaton. Dtrbuton network reconfguraton alo releve the overloadng of network component. 26 Th change performed by openng ectonalzng (normally cloed) and clong te (normally open) wtche of the network. Thee wtchng operaton are performed n uch a way that all of the load are energzed and the radalty of the network preerved. From the mpact of DG on power dtrbuton network, the dtrbuton ytem reconfguraton found to have more lne loe and reducton of termnal voltage a compared to tranmon network. For optmzng power lo, the new reconfguraton can be ued a the feeder reconfguraton a a ytematc method to operate the dtrbuton ytem at mnmum cot and wth mproved ytem relablty and ecurty. By openng or clong the feeder wtche, load current can be tranferred from feeder to feeder thu helpng to tudy the effect of DG on the dtrbuton network wth reference to network reconfguraton problem. The algorthm dealng wth feeder reconfguraton lke heurtc baed and modern optmzaton method have been propoed to olve the problem. The GA baed method and Ant Colony Optmzaton (ACO) were propoed for optmal reconfguraton n preence of DG for dtrbuton ytem power lo reducton repectvely. The other conventonal method lke Honey Bee Colony Optmzaton (HBCO) and Dg Slent oftware were alo uggeted. The feeder reconfguraton and DG placement proce not only reduce the power lo but alo mprove the voltage profle. DG are not only employed to provde real and /or reactve power compenaton n dtrbuton ytem but alo to reduce the power loe and to mantan the voltage profle wthn acceptable lmt. Th paper wll focu on the wnd pecalzed DFIG. The numerou advantage of the DFIG n term of ther low or zero emon and ther much maller ze than conventonal central power plant have created more ncentve than before to ue thee knd of generator wth free energy ource.

2 Internatonal Journal of Emergng Technology and Advanced Engneerng Webte: (ISS , ISO 900:2008 Certfed Journal, Volume 3, Iue 7, July 203) Therefore, t neceary to tudy the mpact of DFIG unt on the dtrbuton network epecally on the dtrbuton network reconfguraton problem []-[2]. A: General Revew Many tude have undertaken the challenge of lo mnmzaton n the area of the network reconfguraton. One of the frt paper on th area had been ntroduced by Merln and Back [3]. They olved the problem by the branch and bound method. But the ntroduced method ha two drawback. Frtly, the convergence of the oluton wa not guaranteed, and econdly t requred a huge amount of calculaton for a real network. In [4], a mple nnovatve method for calculatng lo through the network reconfguraton, whch baed on ome mplfcaton n order to calculate the change of lo n load tranfer from one feeder to another, wa propoed. Aaa and Galana mproved the method n [4], wth a few correcton. The man negatve pont of the method are that there no guarantee to reach the global oluton and the fnal oluton depend on the ntal tatu of the wtche [5]. In [6, 0], a problem for lo reducton and load balancng a an artfcal bee colony (ABC) problem wa modeled. Qwang et al ntroduced a power-flow-mnmum heurtc algorthm for ADR problem [8]. In [9], a model baed automated trategy wa condered the DR problem takng the mart grd nto conderaton. Da propoed a GA and Spannng tree mult-obectve approach to olve the ADR problem []. Many paper have been ued evolutonary algorthm to olve ADR problem. In [7], an algorthm wa condered for effcent and automatc network reconfguraton (EAR) baed on the ervce retoraton and load balancng n a real-tme operaton. The method wa baed on dfferental evoluton. Chou et al preented a varable calng hybrd dfferental evoluton (VSHDE) to olve the ADR. In h work, mnmzng the power lo wa condered a the obectve functon [2].uno and Suana propoed a method for network reconfguraton to mnmze power lo by ung the mproved mxed-nteger hybrd dfferental evoluton [3]. In [4], a PSO baed algorthm for mult-obectve ADR wth fault retoraton propoed. Zhang et al propoed a hybrd algorthm baed on GA and PSO for ADR problem [5]. Yu et al [6] propoed an mproved genetc algorthm wth nfeable oluton dpong for DR problem. In all mentoned paper, the ADR problem ha been olved wthout conderng the DFIG unt, and n th paper, condered for the frt tme 27 B: ADR wth DFIG Recent tude on the ADR problem wth wnd ha been done by conderng general wnd turbne. To acheve the reconfguraton of dtrbuton network wth wnd turbne, Shuang et al [7] raed a probablty-baed cenaro analy to form typcal cenaro of wnd turbne wth conderaton of tattcal charactertc of wnd energy output. By generatng the wtch tate graph of dtrbuton ytem, network reconfguraton wa mplfed to the optmzaton problem of network wtch tate combnaton. For model oluton, a genetc algorthm (GA) and the way of "breakng crcle" wa employed to eek optmal network reconfguraton cheme wth wnd turbne under mono-cenaro and the multcenaro. In [8] a chance contraned programmng formulaton for dtrbuton reconfguraton wth wnd power generator (WPG) wa propoed that am at mnmum power lo and ncrement voltage qualty. A cenaro analy method wa appled to decrbe the random output of WPG through the cenaro probablty and cenaro output. A multple obectve partcle warm algorthm (MOPSO) wa employed to olve the mult-obectve dcrete nonlnear optmzaton problem. The dedcated partcle encodng wth the meh nformaton of the dtrbuton network can effectvely avod producng a large amount of nvald oluton. Wth MOSPO, t wa poble to obtan the optmum oluton et of each obectve whle helpng the operator to chooe the mot approprate plan for reconfguraton. In [9], a novel cenaro dtrbuton network reconfguraton model wa preented n whch the cenaro analy method wa appled to decrbe the random output of the wnd power generator (WPG) and t nfluence through the cenaro electon and cenaro voltage. Multple WPG and wnd farm connected wth a network wa alo condered n th model. And then, an effcent genetc algorthm (GA) wa preented for the cenaro dtrbuton network reconfguraton model. Through the no unfeable codng rule n the ntal populaton trategy, cro trategy and eugenc trategy, ndvdual n the evoluton alway form the feable oluton whch can meet the requrement of the actual dtrbuton network. Phycal optmzaton baed on cenaro voltage n the proce of evoluton reduce the optmzaton tme and the dependence of the ntal populaton. All thee tude are baed on general wnd turbne generator and are therefore not utable for the commercal DFIG.In Addton the hybrd optmzaton method are the tate of the art mean for a real tme and real world power ytem.

3 Internatonal Journal of Emergng Technology and Advanced Engneerng Webte: (ISS , ISO 900:2008 Certfed Journal, Volume 3, Iue 7, July 203) In [20] a ont optmzaton algorthm of combnng reactve power control of wnd farm wth DFIG and network reconfguraton wa propoed. In the propoed ont optmzaton algorthm, an mproved hybrd partcle warm optmzaton wth wavelet mutaton algorthm (HPSOWM) wa developed for voltage profle mprovement whch utlzed reactve power output of wnd farm a the control varable. In each partcle updatng ntance at each teraton of reactve power output optmzaton algorthm, a bnary partcle warm optmzaton algorthm (BPSO) utlzed to fnd the optmal network tructure. Th problem wll be revted n th paper by ung an HBFDE algorthm In the dtrbuton ytem, due to many canddate wtchng combnaton and alo preence of DFIG unt, the ADR problem modeled a a mxed-nteger nonlnear programmng problem. Varou method uch a lnear programmng, mxed-nteger programmng, quadratc programmng, can be ued to olve th problem. However, n many cae, the clacal method fal to provde the global optmum oluton and only reach local oluton. In recent year, The BF and DE algorthm have been propoed a two powerful optmzaton evolutonary algorthm n the feld of optmzaton [2]-[22]. Although n many cae, BF eventually determne the dered oluton; t convergence rate low and t local earch weak. In other de, DE algorthm ha a good performance n the local earch area; but t global earch weak and t performance depend on the proper electon of the ntal populaton. In th paper, hybrd method employed to combne the BF and DE algorthm (HBFDE) baed on the evolutonary nature of them, to take the advantage of the compenatory property of them, and avod ther negatve pont. Mot mportant advantage of propoed HBFDE algorthm hgh accuracy n fndng the optmal wtchng combnaton of the dtrbuton ytem to mnmze the network lo. In addton, the decon maker mut be able to balance the conumpton and generaton of energy at both hort and long tme nterval. Due to ntermttent nature of wnd and load, th dffcult for the network operator to balance the conumpton and producton of energy when the wnd turbne are feedng the electrcal network [23]-[25]. Th the man reaon why a good predcton ha a gnfcant role n handlng of the DFIG and load n electrcal network. Hence, n th paper, the ytem uncertanty ncludng the DFIG and load are modeled n a tochatc approach baed on cenaro. The cenaro are generated ung the roulette wheel mechanm and Monte Carlo Smulaton (MCS) whch model the tochatc behavor of the wnd peed and load uncertanty. Moreover, an aggregaton method ued to decreae the computaton burden and extract one cenaro among all generated cenaro o that the uncertanty of all cenaro are condered n aggregated cenaro baed on ther probablty. Fnally, the HBFDE algorthm ued to optmze the problem for determntc and tochatc (aggregated) cenaro. II. ADR PROBLEM WITH WID The ADR problem conderng the DFIG a mxed nteger nonlnear optmzaton problem. In th problem, there are many dfferent obectve functon ncludng lo mnmzaton, balancng the load on tranformer, balancng the load on feeder, maxmum loadng of feeder and mnmzng the devaton of voltage from nomnal value. In th paper, lo mnmzaton condered a the obectve of ADR problem wth regard to DFIG, whle the other obectve are condered a contrant. The mnmzaton of total real power lo calculated a follow [23]: br f (X) R I The vector X ha three ecton: Te, that repreent the tuaton of te wtch n th loop and get 0 and value correpondng to open and cloe tate, repectvely. Indeed, f the Te 0, t mean that the te wtch of th loop mut reman unchanged (open), alo f t, t mean that te wtch of th loop mut be cloed and the Sw (obtaned n econd part of vector X ) that one of the ectonalzng wtche n th loop, replace the te wtch of that loop. The thrd part of vector X the power factor of th DFIG. The contrant of the propoed problem are lted a follow [23, 24]: Dtrbuton lne lmt, P Lne P Lne, max (2) 2 () 28

4 Internatonal Journal of Emergng Technology and Advanced Engneerng Webte: (ISS , ISO 900:2008 Certfed Journal, Volume 3, Iue 7, July 203) Dtrbuton power flow equaton, P Q bu bu V V Y V V Y co( ) n( ) Radal tructure of the network In th paper, the man cloed loop of the ytem are ued to check the radal tructure of the network. The number of man loop calculated a follow: FL br bu (4) Lmt on current of tranformer, I,2,..., max t, It, Lmt on the current of feeder, I,2,..., max f, I f, Actve power contrant of the DFIG, p p p mn, w, w, max, w, (7) Bu voltage ampltude are lmted a: V V V mn max (8) t f (3) (5) (6) The contnuou dtrbuton functon of the wnd forecat error along wth t dcretzaton hown n Fg. accordng to whch even nterval are centered on the zero mean and each of nterval are one wnd peed forecat error tandard devaton, a preented n [26]. Fnally, a roulette wheel mechanm [27] mplemented to generate cenaro for the propoed tme. In the roulette wheel mechanm, cenaro are elected on the ba of dfferent wnd and load forecat level and ther probablte obtaned from the PDF. Probablty denty 7 Wnd level 7 5 Wnd level 5 3 Wnd level 3 Wnd level Wnd error Fg : Dcretzaton of PDF of wnd forecat error For th purpoe, at frt, the probablte of dfferent wnd and load forecat level are normalzed uch that ther ummaton become equal to unty. Then, by roulette wheel mechanm a cenaro obtaned. In th regard, a hown n Fg. 2, the range of 0- occuped by the normalzed probablte. After that, random number are generated between 0 and. Each random number fall n the normalzed probablty range n the roulette wheel.the elected nterval aocated wth a bnary dgt equal to, and other nterval become zero. Wnd level 2 2 Wnd level 4 4 Wnd level 6 6 Power factor contrant of the DFIG, pf mn, pf pf max, (9) Wnd level 6 Wnd level 5 Wnd level 7 Wnd level III. THE STOCHASTIC MODEL OF DFIG AD LOADS If decon maker want to balance the conumpton and producton of energy n a ytem conderng DFIG, he/he mut make a good predcton of wnd peed and load. In th paper to make a good predcton of uncertanty parameter, the wnd and load uncertanty are condered baed on the wnd and load forecat error [24, 25]. So, the probablty dtrbuton functon (PDF) of the wnd and load forecat error hould be contructed (In th paper a typcal PDF ha been condered.). 29 Wnd level 4 Roulette wheel 0- Wnd level 3 Wnd level 2 Fg 2: Roulette Wheel Selecton

5 Internatonal Journal of Emergng Technology and Advanced Engneerng Webte: (ISS , ISO 900:2008 Certfed Journal, Volume 3, Iue 7, July 203) After obtanng the wnd and load forecat level for all DFIG and load, a cenaro correpondng to all DFIG and load produced. A cenaro a vector of bnary parameter that dentfy the wnd and load nterval: Th procedure repeated untl the dered number of cenaro generated. The probablty of all cenaro calculated by equaton [24]: p WTG 7 load 7 WTG load wn,,. n,,. wn, r, m. n, r, m m r Senaro n en WTG 7 load 7 WTG load wn,, n,, wn, r, m n, r, m k m r n,2,... en (... ) (0) Fnally, obtaned wnd peed converted to the actve power of DFIG by followng equaton [30]: 0 vw v, v v Pw ( vw) v vw v2 p v v v w 3 n 2 w 3 () Where, v, v 2 and v 3 are, cut n, rated and cut-out wnd peed repectvely X. BF _ Populaton.. X for, 2 te 2 te 2 ( BF ) (2 te WTG ) X [ Te, Te,..., Te, Sw, Sw,..., Sw BF, pf, pf,..., pf ],,2,..., WTG (2 te WTG ) BF (2) The chemotax tep compoed of a et of conequence wmmng followng by tumble. At frt, baed on the rotaton of the flagella, the bacterum elect the drecton that t hould move (tumblng). Then, f the new poton of bacterum better than prevou, the bactera wll begn to move n the ame prevou drecton (wmmng). A unt length n a random drecton repreent a tumble; ( ), th random drecton pecfe the drecton of the movement after a tumble. The contant runlength unt, C(, ), determne the ze of the tep taken n the random drecton. In the propoed algorthm, the locaton of the th bacterum at the th chemotactc tep, kth reproducton tep and lth elmnaton/dperal event repreented by (, k, l), and the obectve functon hown by J(,, k, l ). After a tumble the locaton of th bactera hown by the followng equaton [2]: IV. THE HBFDE ALGORITHM A: Bacteral Foragng (BF)Algorthm The Bacteral Foragng algorthm (BF) ha been mplemented n the global optmzaton feld durng the recent year. BF algorthm baed on the foragng behavor of the E. col bactera. Each bacterum by t electon behavor tre to elmnate the poor foragng trategy and modfe the ucceful foragng trategy. A the other evolutonary algorthm, th algorthm degned to olve non-gradent optmzaton problem wth non-dfferentable and complex obectve functon uch a DR problem. In BF algorthm, each poble oluton condered a bactera. There are three man tep n the orgnal BF algorthm, namely; chemotax, reproducton and elmnaton or dperal [2]. Before gong to the man tep, the ntal populaton of bactera mut be produced n ther earch pace a follow: (, k, l) (, k, l) C(, ) ( ) (3) If the obtaned obectve functon at (, k, l) better (lower) than J(,, k, l), another tep of ze C(, ) n the ame drecton taken. Th wmmng operaton repeated a long a a lower obectve functon obtaned untl a maxmum preet number of tep are reached. The obectve functon of each bacterum n the populaton affected by warmng, that performed by the cell-to-cell gnalng releaed by the bactera group to form warm pattern. Th procedure expreed a follow [22]: cc J (, P(, k, l)) J (, (, k, l)) cc p p 2 dattract wattract m m hrepellant wrepellant m m m m [ exp( ( ))] [ exp( ( ) )] (4) 30

6 Internatonal Journal of Emergng Technology and Advanced Engneerng Webte: (ISS , ISO 900:2008 Certfed Journal, Volume 3, Iue 7, July 203) Where, d attract, w attract, h and repelllant w are repellant coeffcent repreentng the charactertc of the attractant and repellant gnal releaed by the cell and the mth component of th bactera poton. Alo, the poton of each member of the populaton of the S bactera how by p(, k, l) and defned a: p(, k, l) (, k, l),2,... m (5) cc Where, S the ze of the bactera populaton. The functon J (, P(, k, l)), whch repreent the cellto-cell gnalng effect, added to the J(,, k, l ). After c tep of chemotactc, the reproducton proce performed. In th tep, the populaton halved o that the leat healthy half de and each bacterum n the other healthet one plt nto two bactera that take the ame poton. S Sr 2 An elmnaton/dperal proce performed after takng a maxmum number of chemotactc tep, re. In th operaton each bacterum could be moved to the new poton of the earch pace B: Dfferental Evolutonary (DE)Algorthm Dfferental Evolutonary (DE) algorthm wa propoed by D.Pal et al [7].Th algorthm a very mple and traghtforward optmzaton trategy.the man concept of DE conducted by mean of four man operaton: - ntalzaton, 2-mutaton, 3- croover, 4- electon. In the ntalzaton tep, DE vertex pont are randomly generated and the ftne of the obectve functon evaluated. The generated populaton orted baed on the obectve functon value a followng matrx: X. DE _ Populaton.. X DE ( DE ) (2 te WTG ) X [ Te, Te,..., Te, Sw, Sw,..., Sw, pf, pf,..., pf ],,2,..., 2 te 2 te 2 WTG (2 te WTG ) DE (6) In the mutaton tep, a donor vector X m created by addng the weghted dfference between the two vector to the thrd vector a followng equaton: Xm X r3, (Xr, X r2, ) (7) Where, ndexe r, r2, r3[, ] are elected randomly. It noted that random ndexe have to be dfferent from each other and from the runnng ndex (). The a contant value n the range of 0 to 2 and control the amplfcaton of the dfference vector of randomly choen ndvdual and number of teraton. The factor determned baed on performance of the optmzaton routne n prevou experment. In the croover, the new ndvdual are generated ung the followng cheme n croover procedure: u X X f r CR f r CR (8) Where, r a random number generator r [0, ] and u repreent tral ndvdual. CR[0, ] a contant value. In the electon tep, the equaton (9), ued for the electon wth the modfcaton, X Where, u f f (u ) (X ) f X otherwe (9) f functon value of (u ) u C: The HBFDE Algorthm and f (X ) are the obectve and X, repectvely. The combnaton of BF and DE algorthm wth each other and other evolutonary algorthm ha been condered by other author n recent year. In [2] and [32], BF and DE ha been combned to perform economc load dpatch wth non-convex load and mprove the performance of algorthm by ncreang the effect of warmng. 3

7 Internatonal Journal of Emergng Technology and Advanced Engneerng Webte: (ISS , ISO 900:2008 Certfed Journal, Volume 3, Iue 7, July 203) In [2], a varable calng hybrd DE algorthm ha been propoed to domnate the problem of the electon of a mutaton operator. But, th paper am at the ntegraton of BF and DE algorthm, to combne ther advantage and avod ther negatve pont. For example, DE algorthm a very effcent local earch procedure, but t not alway avalable nce t very entve to the choce of the ntal condton and there no guarantee to obtan the global optmzaton. Alo, the accuracy of the BF algorthm not very hgh and t local earch weak. But, by combnng thee algorthm, the new algorthm namely, HBFDE algorthm ha a better convergence rate n comparon to orgnal BF and DE algorthm. The algorthm for the DFIG cenaro generaton ummarzed a follow. Step : Set I a the ntal number of tral; =+ Step 2: Import the bac nformaton; that the bu and load data. Step 3: Generate 000 cenaro for the DFIG peed. Step 4: Determne the aggregate of the 000 generated cenaro ung equaton (0). Step 5: Ue equaton () to determne the DFIG output. The HBFDE Algorthm Mechanm Step 6: Generate the ntal populaton. Step7: Let J be the number of HBFDE teraton, J=J+. Step 8: Evaluate the obectve functon for ntal populaton; that the power lo. Step 9: Sort the ntal populaton baed on augmented obectve functon value. Step 0: Devde the orted populaton to BF and DE algorthm. The populaton ze et at 3, when olvng an -dmenonal problem. The ntal 3 oluton are randomly generated and orted by ther ftne, and the top = DE oluton fed nto the DE method to mprove ther poton. The other 2= BF oluton are updated by the BF method. In the BF algorthm neghborhood of bacterum propoed to expand the local earch area. The neghborhood bet bacterum are elected by frt evenly dvdng the 2 bacterum nto neghborhood and degnatng the bacterum wth the better ftne value n each neghborhood a the neghborhood repreentatve. After frt teraton of BF and DE, the frt combnaton completed. In the econd combnaton, the 3 updated oluton are orted agan for repeatng the entre run. Th proce performed after takng a maxmum number of combnaton. Step : Combne the updated populaton from tep 0 to generate the HBFDE populaton Step 2: Compare the HBFDE teraton J and the ntal number of tral.e. J J max then update the HBFDE teraton to J=J+ and go back to Step 7.Otherwe go to Step 3. Step 3. If max go back to Step. then output reult, otherwe The HBFDE algorthm employed to olve the ADR problem, conderng tochatc model of wnd and load. A een above tep, the propoed approach a two tage optmzaton method. In the frt tage, the tochatc cenaro produced for DFIG and load, and n the econd tep, ung of the HBFDE algorthm the optmal confguraton of network obtaned for determntc and tochatc cenaro when power loe obectve functon. V. SIMULATIO RESULTS A: IEEE 33-bu Radal Dtrbuton etwork In th paper to evaluate the feablty of the propoed approach, we ue IEEE 33-bu Radal Dtrbuton etwork [5,33], whch a practcal dtrbuton network. The chematc of th network hown n Fg 3a). The dtrbuton network for reconfguraton cont of 33bue and 5 te lne; the total load are kw and kvar. The normally open wtche are 33, 34, 35, 36, and 37 repreented by the dotted lne and normally open wtche to 32 are repreented by the old lne. For th bae cae, the ntal loe are 22.7 kw. 32

8 Internatonal Journal of Emergng Technology and Advanced Engneerng Webte: (ISS , ISO 900:2008 Certfed Journal, Volume 3, Iue 7, July 203) Algorthm Table. Comparon of average oluton of 0 tral wthout conderng DFIG. Intal populat on Iteraton Bet oluton Wort CPU Average oluton Tme() BF ~29 DE ~24 HBFDE 00 BF=60 DE=40 External loop=0 BF loop=30 DE loop= ~26 Fgure 3a) Fgure 3b) Fg Bu IEEE, Radal dtrbuton network a) n the ntal reconfguraton b) after reconfguraton wth DFIG B: A Comparon of BF DE and HBFDE wthout DFIG In Table a comparon between, HBFDE, BF and DE algorthm for 0 tral gven for DR problem wthout DFIG unt. Accordng to th table, the hybrd HBFDE algorthm more effectve than orgnal BF and DE algorthm n that the tandard devaton of propoed HBFDE algorthm le than of BF and DE algorthm. Indeed, the HBFDE algorthm could obtan optmum oluton n 9 tral of 0 tral and only n one tral the obtaned reult are dfferent. C: ADR problem wth Wnd The ADR problem conderng DFIG ha not been tuded o far. Hence, n the Table 2 to demontrate the performance of propoed HBFDE algorthm, the total actve power lo of the propoed dtrbuton ytem optmzed wthout conderng DFIG unt hown.. Accordng to th table, the obtaned reult by HBFDE algorthm are better than other n term of lo reducton, CPU tme and load flow teraton. To obtan the mnmum power loe correpondng to tochatc tuaton, 000 realzaton of DFIG and load have been developed.in th cae, t can be poble to olve the problem for each cenaro eparately, but t quetonable whch cenaro mut be followed by decon maker. For th mean, all 000 generated cenaro aggregated ung equaton (20) to extract a cenaro whch nclude the uncertanty n all cenaro baed on ther occurrence probablty. Aggregated Scenaro { Pcenaro ( )} (20) After mplementaton of the above equaton, a decon maker can clam that the output control varable can optmum all cenaro whle the contrant n all cenaro are atfed. In Table 2, the obtaned aggregated cenaro ncludng load and DFIG actve power hown to llutrate the mportance of the uncertanty parameter n DR problem, by comparng the determntc and tochatc tuaton. 33

9 Internatonal Journal of Emergng Technology and Advanced Engneerng Webte: (ISS , ISO 900:2008 Certfed Journal, Volume 3, Iue 7, July 203) In th table, to how the mpact of DFIG n DR problem, the determntc tuaton nclude two cae; DR conderng the DFIG and wthout conderng the mpact of DFIG n network. In lat column of Table 3 the optmum te wtche correpondng to each cenaro hown. Table.2. The Smulaton Reult for DFIG n Stochatc Scenaro Lo n the bae confguraton 222.7kw Lo n the optmal confguraton Optmal confguraton Lo reducton 33,4,8,32, kw Lo reducton (%) 39.% CPU tme 0.42 ec umber of load flow teraton 26 The optmal radal confguraton of the network after all the wtchng operaton hown n Fgure 3b). Table 2 how the mulaton reult of the bae confguraton and the optmal confguraton. The mnmum and the maxmum voltage of the two confguraton are depcted n Fg. 4. The power lo before reconfguraton kw and reconfguraton kw. From the reult t oberved that reducton n power lo kw whch approxmately 39. %. The number of all load flow run requred for the entre proce 26 Table.3. The optmum reult of determntc and tochatc tuaton Cae Power loe [kw] Optmum te Swtche DR Stochatc ,4,8,32,28 conderng Determntc ,4,8,32,28 DFIG DR wthout Determntc ,4,8,32,28 DFIG 34 By comparng the obtaned reult n Table and Table 3, t clear that there are dfference between the reult of determntc and tochatc cenaro. In other de, the probablty of the determntc cenaro about 4.% (among 000 produced cenaro). Th mean that the determntc oluton may happen wth the low probablty of 4.%. Conequently, the determntc cenaro cannot be an acceptable oluton by telf. On the other hand, ung the obtaned aggregated cenaro, all 000 cenaro contrbute nto determnng the ADR reult accordng to ther probablty value. So, the ADR reult of the tochatc framework are more realtc than the determntc one. In other word, the mot mportant advantage of tochatc cenaro that ndvdual cenaro problem become mple to nterprete and wth aggregaton the 000 cenaro the underlyng problem tructure preerved D: The voltage profle The voltage profle before and after reconfguraton hown n from Fg.5. It oberved that the mnmum voltage before reconfguraton 0.93 p.u for determntc DR wthout DFIG and after reconfguraton takngh nto conderaton DFIG under both determntc and tochatc cenaro,the mnmum voltage p.u. Th how that the mnmum voltage n the network mproved by 2.78 % after the reconfguraton tochatc DR COSIDERIG DFIG determntc DR conderng DFIG Determntc DR wthout DFIG Fg.4: The profle voltage of three tuaton mentoned n Table 3

10 Internatonal Journal of Emergng Technology and Advanced Engneerng Webte: (ISS , ISO 900:2008 Certfed Journal, Volume 3, Iue 7, July 203) E: Comparon wth other method The HBFDE method compared wth the method propoed by Chou et al [2] and Jngng Zhao et al [20] for the ame 33-bu tet ytem. For effectve comparon, the reult of the propoed method along wth other method are hown n Table 4.The avng n total lo by the propoed method hgher than all other method. The number of te wtch operaton obtaned by the HBFDE and all other method 5.The CPU Tme for the HBFDE method the bet compared to the other method. Table.4. A Comparon of reult obtaned by optmzng actve power lo Method HBFDE Jnglng Zhao et al [20] Chou et al [2] Fnal open wtche 33,4,8, 32,28 7,9,4, 32,37 7,9,4, 32,37 VI. COCLUSIO Total lo avng (%) CPU tme In th paper, a hybrd evolutonary algorthm baed on the combnaton of BF and DE algorthm called HBFDE ha been propoed to obtan the optmal confguraton of radal dtrbuton ytem, when the obectve functon power lo and holdng all the other obectve of ADR a contrant. An ADR ha been tuded conderng the wnd baed DFIG. To reach a more realtc oluton than the determntc tuaton, a tochatc framework ha been condered to model the tochatc behavor of the wnd peed and load. Addtonally, the mulaton reult have hown the effectvene of HBFDE algorthm and the tochatc approach whch can lead to a more effcent utlzaton of energy n the dtrbuton ytem and permt decon maker to elect the poble acton to cope wth the wnd uncertanty n the modern power ytem. REFERECES [] Maurzo Delfant, Davde Falabrett, Marco Merlo Dpered generaton mpact on dtrbuton network loe Electrc Power Sytem Reearch, Volume 97, Aprl 203, Page 0-8. [2] Sath Kanal, Vhal Kumar, Bareev Tyag, Optmal placement of dfferent type of DG ource n dtrbuton network, Internatonal Journal of Electrcal Power & Energy Sytem, Volume 53, une 203, Page [3] Merln, A., and H. Back. "Search for a mnmal-lo operatng pannng tree confguraton n an urban power dtrbuton ytem." In Proc. of the Ffth Power Sytem Conference (PSCC), Cambrdge, pp [4] R.J. Sarf, M.M.A. Salama, A.Y. Chkhan, A urvey of the tate of the art n dtrbuton ytem reconfguraton for ytem lo reducton, Electrc Power Sytem Reearch, Volume 3, Iue, October 994, Page 6-70 [5] A. Aaa, F.D. Galana, "Dtrbuton network reconfguraton for lo reducton ung MILP," gt, pp.-6, 202 IEEE PES Innovatve Smart Grd Technologe, 202 [6] guyen Tung Lnh, guyen Quynh Anh, "Applcaton Artfcal Bee Colony Algorthm (ABC) for Reconfgurng Dtrbuton etwork," ccm, vol., pp.02-06, 200 Second Internatonal Conference on Computer Modelng and Smulaton, 200 [7] D. Pal, S. Kumar, B. Tudu, K. K. Mandal,. Chakraborty Effcent and Automatc Reconfguraton and Servce Retoraton n Radal Dtrbuton Sytem Ung Dfferental Evoluton Proceedng of the Internatonal Conference on Fronter of Intellgent Computng: Theory and Applcaton (FICTA) Advance n Intellgent Sytem and Computng Volume 99, 203, pp [8] Qwang L, We Dng, Janquan Zhang, Anhu Lu, "A ew Reconfguraton Approach for Dtrbuton Sytem wth Dtrbuted Generaton," ICEET, vol. 2, pp.23-26, 2009 Internatonal Conference on Energy and Envronment Technology, 2009 [9] K. J. Ruell, R. P. Broadwater, "Model-baed automated reconfguraton for fault olaton and retoraton," gt, pp.-4, 202 IEEE PES Innovatve Smart Grd Technologe, 202 [0] guyen Tung Lnh, guyen Quynh Anh, "Applcaton Artfcal Bee Colony Algorthm (ABC) for Reconfgurng Dtrbuton etwork," ccm, vol., pp.02-06, 200 Second Internatonal Conference on Computer Modelng and Smulaton, 200 [] J. Torre-Jménez, J. L. Guardado, F. Rva, S. Maxmov, E. Melgoza, "Reconfguraton of Power Dtrbuton Sytem Ung Genetc Algorthm and Spannng Tree," cerma, pp , 200 IEEE Electronc, Robotc and Automotve Mechanc Conference, 200 [2] J-PyngChou; Chung-Fu Chang; Chng-Tzong Su;, "Varable calng hybrd dfferental evoluton for olvng network reconfguraton of dtrbuton ytem," Power Sytem, IEEE Tranacton on, vol.20, no.2, pp , May [3] uno Tenaznha, Suana Vnga, "A Survey on Method for Modelng and Analyzng Integrated Bologcal etwork," IEEE/ACM Tranacton on Computatonal Bology and Bonformatc, vol. 8, no. 4, pp , July-Aug. 20, do:0.09/tcbb

11 Internatonal Journal of Emergng Technology and Advanced Engneerng Webte: (ISS , ISO 900:2008 Certfed Journal, Volume 3, Iue 7, July 203) [4] Y-xong Yan, Yan Wang, Yong-heng Sh, "The Reearch on Fault Retoraton Conderng Dtrbuted Generaton Baed Partcle Swarm Optmzaton," appeec, pp.-4, 20 Aa-Pacfc Power and Energy Engneerng Conference, 20 [5] M. E. Baran and F. F. Wu, etwork reconfguraton n dtrbuton ytem for lo reducton and load balancng, IEEE Tran. Power Del.vol. 4, no. 2, pp , Apr [6] Janmng Yu; Fan Zhang; Feng ; Yuanhe Ma;, "Improved Genetc Algorthm wth Infeable Soluton Dpong of Dtrbuton etwork Reconfguraton," Intellgent Sytem, GCIS '09. WRI Global Congre on, vol.2, no., pp.48-52, 9-2 May [7] Shuang Ma; Hao Zhang; Gang Xu, "Dtrbuton etwork Reconfguraton wth Wnd Turbne Baed on Breakng Crcle," Engneerng and Technology (S-CET), 202 Sprng Congre on, vol.,no.,pp.,4,27-30may 202 [8] Zhou Suwen; Chen Xngyng; Lu Jan; Dong Xnzhou; Lao Yngchen, "Dtrbuton network reconfguraton conderng the random character of wnd power generaton," Advanced Power Sytem Automaton and Protecton (APAP), 20 Internatonal Conference on, vol.2, no., pp.95,200,6-20oct.20 [9] HE Yuqng et al, Scenaro Model an Algorthm for the Reconfguraton of Dtrbuton etwork Wth Wnd Power Generator College of Mechatronc & Control Engneerng, Shenzhen Unverty. [20] Jngng Zhao et al Jont optmzaton algorthm for network reconfguraton and reactve power control of wnd farm n dtrbuton ytem WSEAS Tranacton on Crcut and Sytem Volume8Iue2,February 2009 Page [2] Lu, Y., and K. M. Pano. "Bommcry of ocal foragng bactera for dtrbuted optmzaton: model, prncple, and emergent behavor." Journal of Optmzaton Theory and Applcaton 5.3 (2002): [22] Bwa, Art, et al. "Synergy of PSO and bacteral foragng optmzaton a comparatve tudy on numercal benchmark." Innovaton n Hybrd Intellgent Sytem. Sprnger Berln Hedelberg, [23] Huae Dng, Zechun Hu, Yonghua Song Stochatc optmzaton of the daly operaton of wnd farm and pumped-hydro-torage plant Renewable Energy, Volume 48, December 202, Page [24] Al-Awam, A.T.; El-Sharkaw, M.A., "Stattcal characterzaton of wnd power output for a gven wnd power forecat," orth Amercan Power Sympoum (APS), 2009, vol., no., pp.,4, 4-6 Oct [25] H. Q. Zhou, Z. P. Song, J. P. Wang, Y. Xue, "A Revew on Dynamc Equvalent Method for Large Scale Wnd Farm," appeec, pp.-7, 20 Aa-Pacfc Power and Energy Engneerng Conference, 20 [26] S. Dutta, R. Sharma, "Optmal torage zng for ntegratng wnd and load forecat uncertante," gt, pp.-7, 202 IEEE PES Innovatve Smart Grd Technologe, 202 [27] Goletan, S.; Raoofat, M.; Farah, E., "An mproved nteger coded genetc algorthm for ecurty contraned unt commtment problem," Power and Energy Conference, PECon IEEE 2nd Internatonal, vol., no., pp.25,255, -3 Dec [28] MuhamadRazal..M and Hahm. A. H, Backward Reducton Applcaton for Mnmzng Wnd Power Scenaro n Stochatc Programmng, The 4th Internatonal Power Engneerng and Optmzaton Conf, Malaya.,pp, 23-24, June 200. [29] Hetch, Holger, and Werner Römch. "Scenaro tree reducton for multtage tochatc program." Computatonal Management Scence 6.2 (2009): [30] Yao, D.L.; Cho, S.S.; Teng, K.J.; Le, T.T., "Determnaton of Short-Term Power Dpatch Schedule for a Wnd Farm Incorporated Wth Dual-Battery Energy Storage Scheme," Sutanable Energy, IEEE Tranacton on, vol.3, no., pp.74,84, Jan. 202 [3] Snha, dul, et al. "Hybrd bactera foragng-de baed algorthm for economc load dpatch wth non-convex load." Machne Learnng and Cybernetc (ICMLC), 200 Internatonal Conference on. Vol. 6. IEEE, 200. [32] Pand, V. Ravkumar, Art Bwa, Sambarta Dagupta, and B. K. Pangrah. "A hybrd bacteral foragng and dfferental evoluton algorthm for congeton management." European Tranacton on Electrcal Power 20, no. 7 (200):PP [33] B-chang Zou, Qng-wu Gong, Xun L, Dao-un Chen, "Reconfguraton n Dtrbuton Sytem Baed on Refned Genetc Algorthm for Improvng Voltage Qualty," appeec, pp.-4, 20 Aa-Pacfc Power and Energy Engneerng Conference, 20 36