{E88 Tanancion on Power Dever Vo 5,101 January 190 ne IMPLEMENTATION OF HEURISTIC SEARCH STRATEGIES FOR DISTRIBUTION FEEDER RECONFIGURATION “Tim Taylor, Member IEEE" David Labeemn, Member TEEE Eteettic Power Research Center “School of Engineering [North Carolina Stas Unive Raleigh, North Carolina 27695-7911 ABSTRACT ‘This peper describes « method for feeder reconfiguration wih the poteatial for handling falc operating. constraints, ‘The eppedach taken isto sot up a decision fre 10 represent the Yeiou! switching operations available, A bestfirst tee ‘carhing strsegye Dated on heuristics, i used 0 evaluate the ‘Variour tlemaves. Switching options which could result in Cverlonds, voltage problems, oF other operating constraints, Such ns those assocated with system protection, are quickly ‘minted es possiblities,” "The development of « rule-based {tem aimed At the redaction of the search space is presented 2'S"mneans of implementing the above suaegy. An example [8 deed wo iasuate the concepts deseribed above. Keywords: Distribution Automation, Expert, Systems, Distribution System Planning, Artificial Intligence. INTRODUCTION opening sod co Spjecuves are met." Under normal” operating conditions, the obfetves we to avoid tarformer overloads, feeder thermal SeStonds, snd abnormal voltages while simultaneously SSisimisimg feal power loscs. ih an emergency state, DE System can be renged so. that «maximum number of ‘ttomess retain ‘leetiel service. Altwough it har been mary eieueed ae + reasime como tol in the pas ¢ Roder itcomliguration program could also prove to be very felpfal ‘ir ihe planning’ and design phases of « system, Not fnly would it help in te planing of te layout of cet, But SMould be ale to predict the desired configuration of the tystom for diferent contingency cases. Mr as. flee. fooder ceconfignrion, tey aes advantage of the large degre ‘iverscy that oxi on Some abicbuton spstemes That", cach ditibution foder Sara ‘diferent combination of commercial, industrial, and ‘ll uh “Pet ote end ry ine ie fd ‘wosk and asason thi hey require re suply.Pesdst ae ae ee eid aioe for te inter of ead from Jewviljloaded portions of the dstrbution eygtem network 10 Senet erie Hay ne Tu won ot nly improve the operaing conditions of the system, but it Sead so nate ihe fll ulzaon of wystem. hardware Cepabiltes, “This could) result in the. deferral of capital Ependitures end reduced operaing expenses. "A number of papers have appeared on the genera opi of feeder ‘reconfiguration tr it applies to. notmal operating Cadi "ian ey ai etn ‘cconfigurtion strategy in" which the Topology ofthe network "tchicned newt tables, A search algorithm was used loads in ‘order to distribute feeder” loads to" canter Presently wk Ganaral Blac, Schenectady, New York 29 59 779-3 PDA paper recommeded and approved ty the Litt Trananseaton sed Distribution Connittce of Goo st tow Engioweriag Soctety for presentation at {he 1ERE/PES 1960 Sumunr Heating, Long beach, Calitorata, July 9 ity 1989. namecripe scbattted Februsey 1, 1983; SRE Nacisine’for printing kpett 27, 1988: oportionstely and stleviste overloads. Reeeited's program for detemining te best normal operating fonfigerainn of radial feodr.Ths progam ulized reach ore varour ond tnaer opportntn. The tush sploked margins tha represened the. additional load that SSE be added wo a point without vilming constrains. “An alternative saegy was described by Aoki eta. (] for minimizing power Yrses on disuibuton systems trough the ranirer Seacelonaizing switcher on fesers irconnecing ‘Wierensobsauons The problem wat classifieds a dare Spuimizaton problem, but because. of the large number of SEctonaliing switches that could be selected, the vartblet {Scoping th Teatonn of normaly open iwi wee imate continuous on, per & [scscrbed 's tnsformer load balancing approsch. Using fnoiher approach, Wong a3) described. lnic Sami approach to aesting load song subsites, Kh fit seach was sed 1 fd yf of séding loin Ne Sang operating ome Ut a csribed an expert system operational id for restoration ites redaction of disuibouon systems. ‘A number of recent papers, have focused upon the development of optimization algorithms aimed atthe delmmtnatton “of The. best” system configuration, for Sietanion of lasts Reveal, Liv etal (7) developed 240 meio ihe duermmaion of ihe minim ont [eguraon of redial disibution systems using opumization fSchulques. "Alt, Baran ctl (6) described a reconfiguration IRetiodslogy for lose reson td Toad balancing based pon Sesterng bench change pe ewachings Taal Stirmhsruion’ of dsuibuten networks to. decrease their enue Tine fosees and incioded results pertaining to Ierge- Fale system examples “The work described in this paper some of the eure teach mthodology developed by Grange, Cian Ten and Yin. A comprehensive feeder recon felted paper, Coveslar et al (11) descr elma te foes change reuling from the trans of & SEDs fom one fer tomo eer, Ths done tough the cain e's single sich a the opening of» snl etinalizing Swit ‘The contribution of this paper is to focus on the implementation ef heuristic seuch strategies for feeder ‘eofigraion ener onal operating conaions. A common ‘Beme eich has appeared in most of the past work on feeder feconfiguration tthe need for some type of search procedure ‘Fevevdute every possible configuration of the feeder is not recital ince too many combinations would exist. To reduce [fete ‘of ths search « number of heuristic procedures have been proposes. "The approach taken is to construct a decision tee to represent the svalable switlung operations, A Dest fst wee feliching staeyy, bused on hewstcs described in this paper {fuged 0 evsiate the variou aleratves. In developing the Heute tee-seatch strategy, roles of thumb were created to (scree the search apace of the exhaustive tee search sat fod stil vive af an opsimal of near-optimal solution, ‘velopment of acrle-used system simed a the redoction of the sean space is then presented as a means of implementing the search ‘rategy. "Am example is sed to illustrate the ‘ectiveness of ths approach. (245 $977/90/0100423901.00 © 1969 IEEE20 DECISION TREE MODEL FOR HEURISTIC SEARCH Distribution System Reconfiguration Model ion suateay described here, it is fon systems ae redial in pate that ‘Greach toad leserved through only one source. Al loadsa the dstrbation system are served by subiation wansormers through dintnbuon feeder. "The loude are modeled at het ‘Deak valves, and ti tnumed that each Une secon contains alte ‘whote tats may be opened or closed. "Ax shown in Figure 1 the boundary swish ate normally open witcher between two diferent focders which may be Cesed for the ‘urpose of transferring loads to diferent feeders. The iol frie are normally lowed ewithes along, eer wick ing felts of ‘the Figure 1. Dissibution System Example ‘The primary goal of the heuristic strmgy discussed in hie paper is to efficiently eliminate all tsssformerverlosd Problems ‘and. vols Aetermining the ‘openclosed stain of all-svitches in We System: "A'secondy consideration isthe minimization of tine Iosser The tansformer overload roblem can be defined in he Simplest way as having Xi last one substaton wanstrmer in {he nynem that caries ota connected KVA Toad greater han the mominal capacity of the Wansformer, The lead on he overlonded transformer is. deceared by transferring loads 10 ther wansformers. As soon as the total connected Toad i ‘leqeased Below the nominal transformer capaci the overoad problem is assumed to be removed. If more than one esformer overload exits, the most severe problem is solved ‘est and the eat are solved in order of decreasing level of sven ‘A feeder consteit problem is sid to exist if there re any simotmal voltage wolaions or ermal line overionds ont Daricular feeder” If'more thn one volaon exist the most Severe robiem is solved tit and the rest are solved in order of decreming level of severeness. "A violation is solved by ‘esfring Lode ther fetes, ol meer her transformers), Dung this process ana Yd flow analysis can be performed (0 evaluate the effects of reconfiganng the Stem. Tree Search Approach to Feeder Reconfiguration Determination of « soltion 1 the feeder reconfiguration problem ‘involves searching through «tee. of possible Ecnigrayons." Consider the sytem in, Fgh 1 Sad the corresponding search wee in Figure 2. If wansformer 2.3 ‘xperiencing an overload, then to eliminate the problem the {mount of everioad on tanaformer 2 murt be transferred to ‘ansforsers 1 andlor 3 without creating tn overload on esther (of these transformers. Originally, the wansfer of one load fom Uansformer 2 to another uansformer ie considered. For instance, the load af node 11 may be tansfered to tansformet 1 by closing switch 10 and opening switch 14. Similarly. the Joad si no 12 may be umnlered to tesformer 3 by Fitch 21 and opening switch 15. The loads at nodes 16 {Simay te wantienel lewis. If te snus of performing these oad wansfers are examined then the ee of Figure 2 Chiained.‘The tee i creed when i is necessary to tanalr Sor ten Sve fond cman he overond coon {istance, suppose closing switch 22 and opening swt oes nol remove the ovetloed from wanafooner but it doee Sot cree an overload on transformer. Wik te system ia 1s Stu de effect of porforming father Yoad transfers can be ‘xamined, in thi manner, 2 wee in consvcted in which cach ‘ode of tie wee i the source of number of branches equal 0 the number of open switches connected 19 the feeders Of the Ltansformer experiencing the problem. The final goa! is to Scrch the wee for s configuraton that will elimiawe overloads {tad constrains from the system, If more than one solution ‘inthe ne switching sion hich detibtr the tramfSmer loadings most proporuonalely with respect to thet ‘nominal capaci i Selected, [Figure 2 Search Tree for Elimination of Overload {Legal Operations forthe Removal of Overloads and Constraints ‘The procedure outlined below effectively implements proaches to constraint removal suggested by Grainger, Cimiar, bes wed Yin (10,11) The fist ask inthis met Snvolves fading the fenbie swithings that will remove one Transformeroverloed problem walnut sealing mother problem liewhere. A feasible solution may regute the change of state (of one or # number of boundary switches Wihin the Subsystem, {nN pcific order “The eps of te procedure ares flows 1+ Start with the most overloaded transformer (OLN), Determine the transformers which have direct links othe OLT. ‘These wanaformers are refered ax the fist level ‘eanaformers FLTS). 2. Identify all boundary ewitehes between the OLT and the Furs. 3. Eliminate the candidae boundary switches for which the oad tansfer il emuse 4 thermal problem in the tie fection. “This invclves a simple aclimetic comparison ‘Sctwech dhe EVA thormal capac of te normally open Section (on which the boundary twitch is Toca) ih {Se EVA oad to be tested trough this sich 4 Consider each candidate boundary switch; close that switch and open the isolating. switch to transfer the ‘connected load from the OLT to the FLT; update the total EVA‘on the OLT and FLT due to the load faster apply the following rales: RULE()) IF dve to this overload is elimi icing action, the wanaformer ted and not ereued on the FLTTHEN this ig a candidate solution for the transformer-overoad problem. RULE) IF due to this switching action, the transformer ‘overioud is not eliminated ‘bul e new transformer. Gverloed problem is not erested on the FLT THEN Spay te ehing ct a he poo of PY and FLT repeat RULES(1)<(@) forall newly ‘rented candidate switching options. RULEG) IF due to this switching action, the transformer ‘overioud is ot eliminated ‘bu s new ansformet- Grerlond problem ie created on the FLT THEN ‘iscard thir twitching option. RULE Few hi rig so, ee tei eae wn cee eee SE on SOE HUT rai et na Soatacied lof n't Ricco oes Sy Scat SEN pon The pa Gh at (cove he casos ofa ate Tae lr 2) a (ce ce Pe ee ieee ogi eben, ee BU ete Seer alan aati te te ee copes Stelan mince ‘The determination of solution 10 & feeder-constrant problem is similar to the above procedure. Loed transfers are Pesformed among feeders to sliminate.abpormal voltage PElmons and fooler thermal overlouds, If « feeder has either of these, problems, it iy referted tg as a Consrsined feeder (CSE), the fist level feeders (ELF) are those feoders which ffeve direct open links to the CSF.” During the intermediate fieps of te solution proces, an ac londflow analysis can be pefomed to accurately te the change in the feder section Poprents and voltages. Problemt astociated with severe Sdcroltages ee handled fire Next the CSF with the everest thermal overload Gf there is any) is considered. end, the boundary pitches Between the constrained feeder (CSF) andthe First level feeders (FLF) will be identified; the boundary Switches that cannot cary the wansfered load or those that ‘ilerate a tansormer overload wil be eliminated before the eeaon rules te applied. In the event that more than one Slaton exists for the removal. of any. feeder-constrant problem, the configuration that reduces real power loses by [le resist amount is selected as tbe optimal Slaton. Hewistc vs. Exhaustive Search Strategies ‘The above strategy is efficient for wees that are not too er idm vech tue Hoes ail nt of time and. energy can be spent searching for the Spuimal solution of eaplonng he toe fet the optimal oF SBetopcimal solation has teen found. The somber of Selick acne that can be considered forthe relief of & {ten of ean dnb sysem Can Eaton, ie are. Also, ifthe abate isto be implemented ina realaime Tage einen t's gecestty w bonnes he amount of {ine spent searching for the Sout. ‘To. guarantee an optimal solution, an exhaustive, tse seach sfuld be used all nodes ofthe wee evaluated, On ‘e"oaher hand by utlzing heuristics during the search, i Taste sree he mater af pt hats rire cal, a ge suategy tha would lead ively tothe optimal BoM LTE RIA, Wh probly not alae. Hotere Piheune and’ effort stved by finding a near-optimal solution ulekiy could compensate for not obiaining the optimal ticn ‘every ‘time, then such an’ approach would be benet Spe elosest to 4 goal That node is then exp ‘Buernine iy cildren nodes. I none of the resulling children Ten, eon they are evaluated heuristically 10 determine thelr emsining distance tos goa The cycle then begins again by expanding the tee ode on the bottom level with the Peeatchpotetil wo feed to's foul This ts continted wnt « ral ius been found. Such search techniques are commonly Spiele pobins commonly amend wih eet a rence. ficial intel ‘As an example, consider Figure 3, where A is the root node tnd'O is t geal node. Iatlly, since A if not a goal ode, ris expended into te children modes B and C. Since the TRinmated distance from B to a goal node is less than the Shlmated distance from € 10 2 goal node, B is seleced for Expmsion. Since neiter D nor B are goal nodes, the estimated ‘lances from each of these nodes 10 a goal tre calculate, ‘The cuiinated distance from E Yo e god! node ie leas than the ‘Stunaied distance from D to a goal node, so E is selected for SSpmsion. Upon expending it iy found that node G is & "s {foul node, eo fie seach i terminate. Figure 3. An Example of BestFirt Seach ‘The advantage ofthe bes-ret serch technique is thatthe path taken to find the goal i usually, but not always, shorter Bien with other methods, The disadvantage is thatthe optimal Scluon is not always found, The bestfist search will be ‘Sivantageous only ifthe time and energy saved by finding & Solutions pethape not the optimal ope, compensates for Mot ‘Shuaning fhe bes solution everytime ‘A bestfirst search qualifies as a good approach to the recoafigaration proces decribed hove due to te reduction of “The value in decreasing the search space becomes apparent in considering the complexity of # Gineibution system and” all” the potential effects & Yanhave on the fystem, Fist, « poteatial fon will rogue idles related. to protection, Sipetde seitching. and voltage regulation. Determining he ‘FR of reconigaration on ll fees of ditibuton operation tein ceo, 29 dale int att STopuons studied. ‘Second, emergency mode reconfiguration fie Boviows tne constraints” is tmportant to find a feasible Solution. which meets all system constraints in a re ‘Smount of time, "Third, toe teaches tend to become very large Jnrthe minimization of losses upon a system when no esi lao evi Tee ned a etn of euriatics for the normal-sate reconfiguration problem may Mart fome insight on the ways tat addtional heuristics Could be developed. for the reconfiguration of circuits in Emergency mode or fr the minumizaion of lores