Intelligent Management of Distributed Generators Reactive Power for Loss Minimization and Voltage Control

Transcription

1 Intellgent Management of Dstrbuted Generators Reactve Power for Loss Mnmzaton and Voltage Control Mohd Zamr Che Wank 1, Istvan Erlch, and Azah Mohamed 3 Department of Electrcal Power System, Unversty of Dusburg Essen Dusburg, Germany 1 mohd.zamr@un-due.de stvan.erlch@un-due.de Dept. of Electrcal, Electronc and System Eng.,Unverst Kebangsaan Malaysa Bang, Malaysa 3 azah@eng.ukm.my Abstract Ths paper presents an ntellgent approach for managng reactve power from a group of dstrbuted generators () connected to low voltage (LV) bus. The objectve s to mnmze actve power losses and keep the voltage profles n the network wthn specfed lmt. Ths approach fnds optmal reactve power and optmal tap changer poston for man substaton transformer every ffteen mnutes. The effectveness of the proposed approach s demonstrated n fndng twelve s reactve power and optmal tap poston n a test network. Optmzaton results shows that ths approach manages to reduce power losses and at the same tme keepng the voltages wthn requred lmt. Intellgent management approach presented n paper s sutable to be ntegrated nto energy management scheme under smart grd concept. Index Terms Dstrbuted Generaton, Intellgent Reactve Management, Voltage Control, Smart Grd, Partcle Swarm Optmzaton I. INTRODUCTION Smart grd concept s expected to become a backbone n Europe future electrcty network [1]. In achevng a Smart Grd concept, a large number of dstrbuted generators () are requred nsde dstrbuton network whch s prognosed to supply up to 40% of the dstrbuton network s load demand. Ths large number of s requred to partcpate n mprovng the securty, relablty and qualty of electrcty supply by provdng actve power and other ancllary servces such as regulatng the voltage by provdng ther reactve supply to the network. One of the features of future electrcty network under smart grd concept s to have an effcent transmsson and dstrbuton network that wll reduce lne losses []. Mnmzng losses nsde electrcty transport networks wll result n lower combuston of fossl fuel thus reduce emsson of ar pollutant and green house gasses. Integratng of nsde dstrbuton network already reduces power losses because some porton of the requred load current from upstream s substantally reduce whch result lower losses through lne resstance. Further reducton of losses can be acheved by ntellgently managng reactve power from nstalled [3]. Eventhough current grd code restrcts the reactve power njecton by, t s expected ths practces wll be changed n the future electrcty network. Ths expectaton led to the proposals of reactve power management n a number of publcatons. In [4], reactve power s coordnated locally wth local voltage controller due to poor communcaton nfrastructure n dstrbuton network. In ths method, there s no coordnaton among s as well as no voltage control devces remotely located. Advance communcaton facltes however wll be ntegrated nto dstrbuton network under concept of Smart Grd [1], and ths wll permt the mplementaton of a new strategy n managng a group of under a new concept such as vrtual power plan (VPP). In VPP, a group of s wll be centrally controlled [5]. The central system wll gather data from sensors, through remote communcaton system, and calculates control varables and fnally send the calculated control varables va communcaton remotely. There are stll a lot of dscussons on how to perform ths management task but t s envsoned that ntellgent tools s needed for an effcent management. In ths paper, reactve power management approach ncorporatng optmzaton algorthm for a group of unts s proposed. The management objectve s to mnmze power losses whle mantanng voltage profle throughout the dstrbuton network wthn the requred lmts. Requred reactve power s are calculated and dspatched every 15 mnutes based on forecasted load. Ths approach s sutable to be ntegrated nto energy management system of dstrbuton network operator under Smart Grd concept. The next secton of ths paper wll frst envson future dstrbuton network operaton and control ssues. reactve power supplyng capablty s then brefly dscussed. Method of determnng /10/$ IEEE 685

2 optmal reactve power to be dspatched from each s s then descrbed. Fnally the smulaton result s dscussed and commented. II. FUTURE DISTRIBUTION NETWORK Future dstrbuton network s a smart grd whch wll be ntegrated wth advance communcaton facltes whch lnkng all, sensors and actuators wth network control center as depcted n fg. 1. Ths ntegraton wll enable effcent decsons on how to operate the network n real tme. It s envson that all voltage controller and nsde dstrbuton network wll be coordnately controlled remotely by one control center. Network wll be scanned by data acquston devce and the nformaton be used to determned the optmal operaton of all s and voltage controllers and optmal setpont wll be adjusted remotely through communcaton lnk. In a day, assumng the network s scanned every 15 mnutes, for twenty-four hours operaton, the network s scanned for 96 tmes. Load demand n dstrbuton system vares wth tme. Ths varaton s due to human actvtes, needs and desres vary by the tme of the day and season of the year. A typcal load demand curve s depcted n fg. for workng days n summer, wnter, fall and sprng. A peak load n wnter season s nearly double the peak load n fall. For dstrbuton network voltage profle, t wll vary accordng to load demand. Durng lght load, voltage wll rse and n contrast decrease durng heavy load condton. A voltage control scheme wll try to keep the voltage wthn ts voltage constrant followng the changes of load demand. A voltage control scheme works conventonally based on local measurement and are usually operated ndependently. The prmary voltage controller n a radal dstrbuton system s a transformer equpped wth onlne load tap changer (OLTC) n man substaton. Conventonal OLTC control method works based on local voltage measurement at the substaton feeder and the taps s changed up or down to return the voltage wthn specfed range whch typcally close to nomnal voltage. HV GRID OLTC Control P ( MW ) :00 01:15 0:30 03:45 05:00 06:15 07:30 08:45 10:00 Fg.. Typcal load demand curve It s however n future dstrbuton network, due to coordnaton of OLTC wth other voltage control devces as well as, the conventonal control method probably s not sutable. As the objectve of conventonal OLTC control s to return the voltage nsde deadband range, transformer s tap wll probably postoned unoptmally. It s therefore recommended that tap poston on substaton transformer s forced to an optmal poston remotely by control center. In current utlty practce, power dspatch s performed for one day ahead based on the one day ahead load forecast. Ths load forecast are determned usng conventonal method whch are based on the relaton of load power and factor nfluence t [6] or more ntellgent method whch based on smlarty [7]. In the load forecastng method, weather condton has a great nfluence on the projected load. In the future, t s expected wth the clmate change, weather condton s unpredctable. The weather could change suddenly and ths wlls defntely changng the load demand. If the dspatch of power performed based on one day ahead plannng power, network wll not be operated at optmal operatng pont and n a worst case can destablzed the whole system f the load demand suddenly ncreased beyond the planned power dspatch. In future electrcty network wth s term Smart, nformaton about system profles s expected to be avalable every 15 mnutes. It s therefore suggested that nstead of takng one day ahead dspatch plannng, optmzaton s run every 15 mnutes to determne next 15 mnutes power to be dspatched for each based on the projected forecast. Wth ths approach any sudden changes n the weather condton can be consdered to make sure power system s operated optmally. Reactve power setpont gven by optmzaton process can be send remotely by network operator manually or automatcally. Fall 11:15 Summer Sprng tme ( hours ) Wnter 1:30 13:45 15:00 16:15 17:30 18:45 0:00 1:15 :30 3:45 Management System Data Aquston Fg.1. Future dstrbuton network Network nformatons from sensors III. DISTRIBUTION NETWORK POWER LOSSES Future dstrbuton grd need to be effcent where losses are reduced to mnmum value. Actve power losses occur n the process of transportng electrcal power, due to lne resstance. An actve power loss n the lne depends on magntude of the current flows through the lne and resstance of the lne. Subjected to the number of lnes avalable n the 686

3 network, total actve power losses s gven by P loss total = 3I j Rj j N L j where N L s the number of the lnes avalable n the network. In ac dstrbuton crcut, due to electrc and magnetc feld produce by the flow of tme varyng current, nductance and capactance can be sgnfcant. When current flow through these two components, reactve power whch transmt no energy s produced. Reactve current flow n the lne contrbutes to extra power losses n addton to actve power losses menton prevously. Integraton of already reduced actve power losses because some porton of power from upstream s already reduced. Losses reducton can be further reduced by controllng the voltage profles n the network. In conventonal practce, capactor banks are added n the dstrbuton network to control the flow of ths reactve power. These capactor banks can be swtched n and out usng voltage regulatng relay to delver reactve power n steps. The swtchng s however lowered power qualty delvered to the customer as t leads to step changes n busbar voltage. Eventhough ths power qualty problem can be reduced by swtchng smaller amount of capactance at each step, ths acton however ncrease mantenance cost of swtch gear and complexty of control equpment. Reactve power however can be control contnuously usng based power electronc converter and prevously dscussed power qualty problem can be mtgated. As s capable of provdng reactve current, coordnatng wth other voltage controller devces n the network for voltage control s a strategy that can be appled. Losses reducton n dstrbuton level wll also reduce the losses n transmsson network. Ths wll lead a reducton of fossl fuel combuston, ar pollutant and green house gases. The need for provdng spnnng reserves can also be reduced. IV. OPTIMAL REACTIVE POWER FROM connected to electrcal grd through power electronc converter can be set to nject reactve power by changng ts operatng power factor or ncreasng ts reactve current output. Its reactve maxmum supply s gven by Q, ( t) = S, P ( t) max MAX (1) () How much reactve power can be suppled by a certan unt depends on ts actve power setpont. The dagram llustratng reactve power capablty s shown n Fg. 3. If actve power s set close to ts rated apparent power ratng, capactve or nductve reactve power range that can be suppled s small. Reactve supply capablty of can be ncreased by reducng ts actve power generaton. If actve power setpont s reduced, more reactve supply s guaranteed. In ths case, actve power output has to be curtaled whch s not favorable to the economc pont of vew because actve power from s expected to be cheaper than transported actve power from upstream. But f ths acton can beneft the dstrbuton network operaton, ths actve power curtalment s acceptable. In some cases, s ntentonally overszed to ncrease ts power supply capablty. Actve power supply can be set at unt rated value whle at the same tme there are stll reactve power supply avalable up to ts overszed ratng. S P P(t) Q IND Q IND,MAX Q CAP,MAX Q CAP Fg. 3. Apparent, actve and reactve power from By allowng to nject reactve power, the dependency of the dstrbuton system on the reactve power from hgh voltage grd s reduced. Eventhough the amount of reactve power that can be suppled by s lmted and s not enough to supply the entre reactve power requrement; ths correctve acton n consequence can mnmze the power losses nsde the dstrbuton network. As unt wll be dstrbutedly ntegrated nsde dstrbuton network. Dependng on the locaton, reactve power amount requred from unt wll be dfferent for dfferent operatng condton. To much reactve power requested from wll lmt ts actve power generaton but too low supply wll probably not enough to regulate the voltage at the connecton ponts. Optmal reactve power should be calculated because ths nformaton can be used to determne how much reactve power should be provded by each. Determnng the power capacty of to be allocated n a certan locaton can also be done through ths method. In fndng optmal reactve power, optmzaton algorthm s requred. Partcle swarm optmzaton algorthm (PSO) [8] s a populaton based search algorthm used to fnd the global optmal soluton by a process motvated by socal behavor of swarm such as fsh and brds searchng for food. Algorthm s ntalzed by a populaton of random potental solutons called partcles and a group of partcles s called swarm. Each partcle n the swarm moves n over the search space wth a certan velocty. Ths velocty s nfluenced by both partcles own flyng experence and ts neghbor flyng experences. Adaptve partcle swarm optmzaton (APSO) [9] s an extenson of PSO whch s free from parameter tunng. Ths parameter free technque stochastc optmzaton algorthm has been shown capable of solvng many power system optmzaton problems whch are nonlnear and dscontnuous n nature [10]. For ths algorthm only search space, objectve functon and stoppng crteron need to be defned. In ths algorthm, Trbes whch each comprses a group of partcles moves together n search space to fnd local mnmum. There are lnkages amongst trbes where they exchange the nformaton on ther local mnmum and decde global mnmum. The trbes whch contan more good partcles compared to bad partcles wll be consdered as good trbes 687

4 whle they are consdered as bad trbes f otherwse. Each bad trbe wll generate a new partcle smultaneously and these new partcles wll form new trbes. Ths adaptaton contnues untl a certan requrements are met. Not all technologes s however sutable n meetng ths crtera for an example from ntermttent resources such as Photovoltac s not relable because the requred servces probably cannot be delvered when t s needed. The technology that ft ths requrement perfectly s mcroturbne generaton system and fuel cell. These two technologes are already avalable commercally and has been demonstrated capable of provdng actve power and other added value servces such as njectng reactve current when requred[11]. V. TEST SYSTEM AND SIMULATION RESULTS The proposed technque s appled to the dstrbuton network shown n fg. 4. Ths radal dstrbuton network s fed by a HV grd. The man substaton transformer steps down the 110 kv voltage to 10 kv medum voltage. There are three feeders branchng out from the substaton wth each feeder havng three, four and fve 10kV/0.4kV dstrbuton transformers, respectvely. On each LV sde of a dstrbuton transformer there are dstrbuted generators wth ratng capablty of 33% of the maxmum local load. Wth ths magntude of generaton capacty all the actve power supply by s s consumed by local load. technology consdered n the study s converter based n whch generaton can be controlled by settng ts power set pont values remotely. technologes that ft ths requrement for example are mcroturbne generaton system and fuel cell. These two technologes are already avalable commercally and are the most prospectve power generators for dstrbuted generaton applcaton. These two technologes can be used to generate both electrcty and thermal power, but n ths study t s assumed they are used to generate only electrcal power. In ths dstrbuton network, the man substaton transformer s equpped wth OLTC whle all the other dstrbuton transformers are equpped wth off load tap changers. Wthout reactve supply, the man voltage control n the network s manly acheved by usng OLTC at the man substaton. For solvng optmal reactve power, the problem s mathematcally formulated as follows: Mn Subject to P loss (x) a) Transformer tap settng lmts mn max a a a b) All bus voltage lmts u mn u u max c) reactve power lmts mn max Q Q Q _ N N B (3) (4) (5) HV Grd 110/10 kv T1, T1 = 10/0.4 kv P 1 3 T1 T T3 T4 T5 T6 T T8 T9 T10 T Fg. 4. Dstrbuton network ntegrated wth T1 1 where N B s the number of buses and N s the number of unts. Vector x contans control varables (3) and (5). The transformer tap settng s treated as a dscrete decson varable whle the reactve power lmt s treated as a contnuous decson varable. reactve power and transformer tap settng are treated as decson varable to be optmzed wth the objectve of fndng mnmum power losses subjected to the constrant descrbed n equaton (3), (4) and (5). The dstrbuton network shown n fg. 4 s modeled n power flow program wnlf8 [1]. For every 15 mnutes nterval hstorcal load demand of a workng day n summer [13], the load s dstrbuted nsde the network. Power flow s run usng wnlf ncorporatng Newton Rapson method whle APSO menton n prevous secton s used to solve the optmzaton problem. APSO s ntegrated nsde wnlf8 n such a way that addton of decson varables can be done easly. The voltages at the secondary sde of transformer T1 from smulaton results s depcted n fg. 5. Even wthout reactve power the voltage could be mantan wthn ±5 % of the rated voltage. Ths s due to the facts that even only supplyng actve power, voltage drop n the lne s smaller due to smaller current flowng from upstream and the acton of man substaton transformer OLTC. The changes of voltage represent movement of transformer tap poston are shown n fg. 6. Percentage changes of voltage on vertcal axs represent the changes of tap poston. For one day operaton, wthout reactve power from, transformer need to change ts tap 15 tmes whle wth s reactve supply, only 6 movement s requred. Optmal reactve management manages to reduce the nearly 3 tme tap changes. In future dstrbuton system, t s expected that FACTS devces such as STATCOM wll be nstalled for reactve compensaton. STATCOM whch provde contnuous reactve supply can be located at the man substaton and by ts nstallaton the number of tap changes wll be substantally reduced. 688

5 Result from optmzaton smulaton for reactve power setponts of four s s depcted fgs. 7. Customer orented system s used n the smulaton where negatve reactve power s capactve whle postve value s nductve. Optmal reactve power of four from the results dsplay n fgure show that n meetng the objectve to mnmze actve power losses n some tme steps 4 supply nductve reactve power whle the other three s supplyng capactve reactve power to the network. These requrements dctate by where those s are located n the network. For example the farthest locaton n the network s where 1 s located. Ths locaton experence the lowest voltage profle compared to the other locatons. More capactve reactve power capacty n ths locaton s requred and szng should also be ncreased. It s expected that by guarantyng more reactve power capacty from 1, more actve power losses can be reduced. As the man objectve of centrally managng group of s to reduce the actve power losses, the comparson of actve power losses between the cases of reactve power s controlled centrally to the case there s no reactve power supply comng from s depcted n fg. 8. In most of the tme steps shown n the fgure, at least 0.05 MW actve power losses are reduced when reactve power form s centrally managed. In some tme steps, t can be seen up to 0.1 MW of losses be reduced. If all losses are accumulated, a substantal amount of reducton n actve power losses could be acheved n one day operaton. It also has to be kept n mnd ths losses reducton n dstrbuton network wll also reduce the losses n the transmsson network as well. If all dstrbuton networks connected to transmsson grd managng ts reactve power effectvely, a substantal amount of losses could be reduced n the whole power system. Changes of Voltage ( % ) Reactve Power ( MVar ) No reactve power from Optmal reactve power from tme steps Fg. 6. Correcton voltage by OLTC tme steps Fg. 7. Reactve power reference for four selected s Voltage ( p.u.) No reactve power from Optmal reactve power from Actve power losses ( MW ) Optmal reactve power from No reactve power from tme steps tme steps Fg. 5. Voltage at secondary sde of transformer T1 Fg. 8. Comparson of actve power losses 689

6 VI. CONCLUSION In ths paper, ntellgent approach for management of reactve power from a group of s presented. A group of reactve power and man substaton transformer tap poston s optmally calculated usng APSO algorthm. The effectveness of the approach s demonstrated on the test dstrbuton network contanng twelve s. The smulaton results ndcate that dstrbuton network power losses and at the same tme keepng all bus voltages wthn requred lmt can be acheved. Ths approach s sutable to be ntegrated nto energy management system applcaton under smart grd concept whch wll be appled to the Europe s electrcty networks n comng years. In addton to effectvely managng s reactve supples, dstrbuton network voltage control s at the same tme s performed. VII. REFERENCES [1] European Commsson, "European SmartGrds technology platform: vson and strategy for Europe s electrcty networks of the future, " 006, [onlne] [] Grd 030 A natonal vson for electrcty s second 100 years. [Onlne]. grd-030.pdf. [3] T. Ackerman and V. Knyazkn, " Interacton between dstrbuted generaton and the dstrbuton network: operaton aspects," n Proc. 00 IEEE PES Transmsson and Dstrbuton Conference and Exhbton, vol., pp [4] S. Toma, T. Senjyu, Y. Myazato and A. Yona, "Decentralzed voltage control n dstrbuton system usng neural network," n Proc. 008 nd IEEE Internatonal Conference on Power and Energy, pp [5] P. B. Andersen, B. Poulsen, M. Decker, C. Traeholt and J. Ostergaard, "Evaluaton of a generc vrtual Power Plant Framework usng Servce Orented Archtecture, " n Proc. 008 nd IEEE Internatonal Conference on Power and Energy, pp [6] IEEE Commttee Report, 1980, Load forecast bblography, Phase 1, IEEE Transacton on Power Apparatus and System, Vol. PAS-99, No. 1, pp [7] T. Senjyu, H. Sakhara, Y. Tamak and K. Uezato, Next-day load forecastng usng neural network based on smlarty, Electrc Power Component and Systems, 9: , 001. [8] Kennedy, J. and R.C. Eberhart, Swarm Intellgence, Morgan Kaufmann Publshers, 001. [9] Maurce Clerc, TRIBES, a parameter Free Partcle Swarm Optmzer: French verson: Presented at OEP 03, Pars, France. [10] V. S. Pappala, "Applcaton of PSO for optmzaton of power systems under uncertanty, " Dr.-Ing thess, Unverstät Dusburg-Essen, 009 [11] M. Z.C. Wank and I. Erlch, "Smulaton of mcroturbne generaton system performance durng grd fault under new grd code requrements," n Proc. of 009 IEEE PES Powertech, Bucharest. [1] I. Erlch, Analyse und smulaton des dynamschen verhaltens von elektroenergesystem (Analyss and smulaton of the dynamc behavor of electrcal power systems), (n German) Habltaton Thess, Dept. Elect. Eng., Tech. Unv. Dresden, Dresden, Germany, [13] AEMO, Aggregated prce and demand: , [Onlne]. Avalable: 690