sensors Article Towards Stochastic Optimization-Based Electric Vehicle Penetration in a Novel Archipelago Microgrid QingyuYang1,*,†,DouAn2,†,WeiYu3,*,†,ZhenganTan2,† andXinyuYang2 1 SKLMSELab,SchoolofElectronic&InformationEngineering,Xi’anJiaotongUniversity, Xi’an710049,China 2 SchoolofElectronic&InformationEngineering,Xi’anJiaotongUniversity,Xi’an710049,China; [email protected](D.A.);[email protected](Z.T.);[email protected](X.Y.) 3 DepartmentofComputerandInformationSciences,TowsonUniversity,Towson,MD21252,USA * Correspondence:[email protected](Q.Y.);[email protected](W.Y.); Tel.:+86-159-2972-3079(Q.Y.);+1-410-704-5528(W.Y.) † Theseauthorscontributedequallytothiswork. AcademicEditors:DongkyunKim,HoubingSong,Juan-CarlosCano,WeiWang,WaleedEjazandQingheDu Received:18April2016;Accepted:1June2016;Published:17June2016 Abstract: Due to the advantage of avoiding upstream disturbance and voltage fluctuation from apowertransmissionsystem,IslandedMicro-Grids(IMG)haveattractedmuchattention. Inthis paper,wefirstproposeanovelself-sufficientCyber-PhysicalSystem(CPS)supportedbyInternet of Things (IoT) techniques, namely “archipelago micro-grid (MG)”, which integrates the power gridandsensornetworkstomakethegridoperationeffectiveandiscomprisedofmultipleMGs whiledisconnectedwiththeutilitygrid. TheElectricVehicles(EVs)areusedtoreplaceaportionof ConventionalVehicles(CVs)toreduceCO emissionandoperationcost.Nonetheless,theintermittent 2 natureanduncertaintyofRenewableEnergySources(RESs)remainachallengingissueinmanaging energyresourcesinthesystem. Toaddresstheseissues,weformalizetheoptimalEVpenetration problemasatwo-stageStochasticOptimalPenetration(SOP)model,whichaimstominimizethe emissionandoperationcostinthesystem. UncertaintiescomingfromRESs(e.g.,wind,solar,and load demand) are considered in the stochastic model and random parameters to represent those uncertaintiesarecapturedbytheMonteCarlo-basedmethod. Toenablethereasonabledeployment ofEVsineachMGs,wedeveloptwoschedulingschemes,namelyUnlimitedCoordinatedScheme (UCS)andLimitedCoordinatedScheme(LCS),respectively. Anextensivesimulationstudybased onamodified9bussystemwiththreeMGshasbeencarriedouttoshowtheeffectivenessofour proposedschemes. Theevaluationdataindicatesthatourproposedstrategycanreduceboththe environmentalpollutioncreatedbyCO emissionsandoperationcostsinUCSandLCS. 2 Keywords:archipelagomicrogrid;electricvehicles(EVs);two-stagestochasticprogramming;scheduling 1. Introduction With the development of the smart grid, microgrids (MGs) are becoming miniaturized, independentandcommunity-basedcontinuously. Asatypicalenergy-basedCyber-PhysicalSystem (CPS)supportedbyInternetofThings(IoT)techniques[1],MGsintegratebothphysicalelementsin thepowergridandcyberelements(sensornetworks,communicationnetworks,andcomputation core)tomakethepowergridoperationeffective. Suchacomplexsystemcallsfornovelapproachesfor systemdesignandmodelingandnewtechniquestomanageenergyresourceefficiently. Duetohighlyintermittentanduncertaintiesfrombothphysicalpowergridandcybernetwork components,newtechniquesforMGmodelingandmanagementshouldbedevelopedtoenhance energyefficiency[2]. Specifically,theemergedIslandedMicro-Grids(IMG)[3]havereceivedgrowing Sensors2016,16,907;doi:10.3390/s16060907 www.mdpi.com/journal/sensors Sensors2016,16,907 2of22 attentionduetotheself-balanceandstabilityundervoltagefluctuations. TheswitchinIMGconnected totheutilitygridcanturnofftheconnection,incasethemalfunctionresultedfromthemaingrid, includingupstreamdisturbances,voltagefluctuation,etc.[4].TheoperationofIMGisabletoeffectively mitigatethedetrimentalsituationsraisedbytheutilitygrid,reducetheoperationcostandenhancethe reliabilitywhenapowershortageoccurs. DuetoflexibilityofcharginganddischargingElectricVehicles(EVs)inIMG,energyefficiency andenvironmentalfriendship,EVs,asakeycomponentinIMG,havedrawnmuchresearchattention inrecentyears. WiththeintegrationofEVsinIMG,effectiveschedulingschemesarerequiredtohelp smooththeloadcurveandreducetheenvironmentalemissions.Tothisend,alargenumberofresearch effortshavebeenconductedonEVs. Forexample,anobjectiveofminimizingtheoperationcostand maximizing the profit of IMG was studied by scheduling the EV charging and battery swapping stationwithafuzzycontrolapproachin[5]. SaberandVenayagamoorthyin[6]investigatedtheoverall scaleofEVsundertheobjectiveofreducingtheCO emissionandoperationcost. Nonetheless,their 2 proposedschemeonlyusedEVstoreplaceallConventionalVehicles(CVs)whilethelocalunitcapacity andtheinvestmentcostswerenotconsidered. InanIMG,theinstalledlocalunits’generationneedtomeettheloaddemandduringthepeak timeduration,combiningwiththegenerationofRenewableenergysources(RESs). Incomparison withtheMGsconnectedtotheutilitygrid,alargercapacityshouldberequiredinanIMG,which leadstoanexpensiveinvestmentcost. Inaddition,iftherearesufficientRESsinanIMG,thesurplus energywillbewastedaftersatisfyingtheloaddemandandfillingupthestoragedevice,leadingtoan inefficientenergyusage. Tothisend,asanessentialloadandenergystoragedevice,EVsoffergreat flexibilitytoinfluencetheenergyefficiency,stability,andenvironmentbenefitsofIMGsignificantly. Thus,itiscriticaltodevelopeffectiveschedulingschemesforEVsinanIMG. Nonetheless,theintermittentandrandomnessofRESs(e.g.,windandsolar)willcreategreat challenges to the stability and reliability of IMG operation [7]. In particular, the forecast accuracy of the uncertain parameters are essential to the energy efficiency. For instance, a lower prediction accuracywillleadtomorereverserequirementstocompensatetheerrorsinthecasethattheforecast generation is more than the actual generation. In addition, the precise prediction of parameters (electricityprices,etc.)helpstheconsumersmakerationaldecisionsofEVcharginganddischarging, thusavoidinggeneratingnewpeakloadforthesystem. Therefore,howtotackletheuncertaintiesin anIMGisacriticalissue. Toaddresstheseissues,inthispaper,wefirstproposeanovelIMGsystem,named“archipelago micro-grid(IMG)”,whichisdisconnectedfromtheutilitygridandcomposedofseveralconnected MGs. In such a system, the energy demanded is fully satisfied by distributed energy generations. In the proposed IMG, we formalize the optimal scheduling problem, which aims to minimize the emissionandoperationcostbyarrangingtherationalcharging/dischargingplananddeployinga reasonablenumberofEVswhileconsideringvariousuncertainties. Thekeycontributionsofthispapercanbesummarizedasfollows: • To improve the energy efficiency and ensure the stability of IMG, we propose a novel multi-microgrid system concept named “archipelago microgrid”. In our proposed system, the required power will be fully provided by the distributed power generation in local MGs. ThroughtheenergytransmissioncontrolledbyaMicroGridCenterController(MGCC,aggregator) among the MGs, the emission and operation cost created by local units can be mitigated. Theproposedsystemishelpfulinreducingoperationcostandenhancingthestabilityofthegrid. • WeproposeaStochasticOptimalPenetration(SOP)model,aimingtominimizethecostofthe proposed“archipelagomicrogrid”system. ToaddresstheuncertaintiesfromRESsandelectricity prices,theoptimizationproblemisformalizedasatwo-stagestochasticprogrammingproblem. Intheformalizedproblem,theuncertainparameterssuchaswind,solargenerationcapacityare capturedbytheMonteCarlo-basedmethod[8]. Forthesakeofcomparison,theDeterministic OptimalPenetration(DOP)modelbasedonourpriorwork[9]isintroducedasabaselinescheme. Sensors2016,16,907 3of22 In the optimization problem, the emission cost created by CVs and local units, the operation costofstartup/shutdownexpenseofunits,tariffcompensation,batterycapacitydegradation, andpowerlossesareconsideredintheoptimizationmodel. Theproposedstochasticmodeloffers adesiredflexibilitybetweentheenvironmentalandeconomicbenefitsbyseekingtheoptimal numberofEVswiththeconsiderationsoftheRESsandunitsgenerationlimits. • To achieve the minimization of emission and operation cost, we propose the following two schemes to schedule the optimized scale of EVs: (i) Unlimited Coordinated Scheme (UCS); and (ii) Limited Coordinated Scheme (LCS). In UCS, we consider that all the surplus energy is utilized to charge as many EVs as possible. In this ideal scenario, the mutual transmission amongMGsisalsoallowedtoavoidenergywasted. Nonetheless,peakloadlimitsandresidence preferenceofenergyusageshouldbeconsideredinpractice. Tothisend,weproposetheLCSto achieveamorerealisticnumberofpenetratedEVstominimizethetotalcost. • WecarryoutanextensivesimulationstudyonamodifiedIEEE9-bussystemtodemonstrate theeffectivenessoftheproposedSOPmodelusingthetwoschedulingschemes. Thesimulation resultsshowsthat,afteraddressinguncertainties,theemissionsandoperationcostarereducedin comparisonwiththedeterministic-basedoptimizationmodel. Inaddition,withrespecttothetwo scheduleschemesinSOP,theLCShasbeenproventobemoreeffectiveonarrangingthescaleof EVsandrealizingtheemissionandoperationcostreductionthantheUCS.Ourexperimentaldata showsthat,incomparisonwithabaselinenon-coordinatedaveragescheme,15.2%ofemissions canbereducedand11.2%costcanbesavedintheLCS,weconductsensitivityanalysistovalidate theimpactofdifferentparametersontheoptimalsolution. Partofthisworkwaspublishedin[9]. Basedonthemuchshorterconferenceversion,wehave madesubstantialextensionsinthisjournalsubmission. Therestofthepaperisorganizedasfollows, wepresentsystemmodelsinSection2. InSection3,weshowtheformalizationoftheSOPproblem andproposeour schedulingschemes. InSection 4, wegivethe performanceevaluationresultsto demonstratetheeffectivenessofourproposedschemes. WepresenttheliteraturereviewinSection5, andweconcludethepaperinSection6,respectively. 2. SystemModels Inthissection,wefirstpresentthesystemmodelofourarchipelagomicrogridandthendescribe themarketmodelusedinthepaper. 2.1. ArchipelagoMicrogridModel In this paper, in order to refrain from the voltage fluctuations and upstream disturbances caused by the utility grid, while increasing the reliability of local MGs, we propose the concept of an independent and self-sufficient system, called “archipelago microgrid”. The basic structure of“archipelagomicrogrid”isillustratedinFigure1. Asshowninthefigure,thesystemisatypical energy-basedCyber-PhysicalSystemsupportedbyInternetofThings(IoT)techniques,whichconsists ofbothcyberandphysicalcomponents. (cid:48)(cid:76)(cid:70)(cid:85)(cid:82)(cid:74)(cid:85)(cid:76)(cid:71)(cid:3)(cid:21)(cid:3) (cid:36)(cid:74)(cid:72)(cid:81)(cid:87)(cid:21) (cid:36)(cid:74)(cid:74)(cid:85)(cid:72)(cid:74)(cid:68)(cid:87)(cid:82)(cid:85) (cid:87)(cid:85)(cid:68)(cid:81)(cid:86)(cid:80)(cid:76)(cid:86)(cid:86)(cid:76)(cid:82)(cid:81)(cid:3)(cid:79)(cid:76)(cid:81)(cid:72) (cid:70)(cid:82)(cid:80)(cid:80)(cid:88)(cid:81)(cid:76)(cid:70)(cid:68)(cid:87)(cid:76)(cid:82)(cid:81)(cid:3)(cid:79)(cid:76)(cid:81)(cid:72) (cid:36)(cid:74)(cid:72)(cid:81)(cid:87)(cid:3)(cid:20) (cid:36)(cid:74)(cid:72)(cid:81)(cid:87)(cid:3)(cid:81) (cid:48)(cid:76)(cid:70)(cid:85)(cid:82)(cid:74)(cid:85)(cid:76)(cid:71)(cid:3)(cid:20) (cid:48)(cid:76)(cid:70)(cid:85)(cid:82)(cid:74)(cid:85)(cid:76)(cid:71)(cid:3)(cid:81) Figure1.Structureofthearchipelagomicrogrid. Sensors2016,16,907 4of22 ThephysicalcomponentsarecomprisedoflocalMGs,theinterconnectiontransformerwhich ensures the power transmission among local MGs. In each MG, the diesel unit and various RES generationdevicesareusedtoproducetherequiredenergy. Theinstalledcapacityofalldistributed generatorswillmeettheloaddemandofeachMGinordertoavoidtheoccurrenceofpoweroutage. Inaddition,RESgenerators,whichconsistofsolarpanelsandwindturbines,areintegratedineachMG. TheloadsinMGsaredividedintotwotypes: EVsandnon-EVloads. ThroughG2V(Grid-to-Vehicle) andV2G(Vehicle-to-Grid)technologies,theEVscanbemanagedtochargeordischargeatdifferent timeslotsaccordingtothecurrentpowerconsumption,whilethenon-EVloadsareunabletoschedule. In addition, the cyber components guarantees the information transmission in the system, whicharetypicalsensornetworkscomposedofAdvancedMeteringInfrastructure(AMI)—inwhich, the data measured by sensors, including timely energy consumption, load forecasting and RES relatedinformationwithintheMG,willbecollectedbytheMGagentsandtransmittedtotheMGCC. The entireregionwillbemonitoredbythesystemMGCCthatcansendorreceivevariousinformation fromortotheagentssupportedbyAMI(sensornetworks). 2.2. MarketModelandAggregator InordertopenetratethereasonablenumberofEVswhileminimizingtheoperationandemission cost(accordingtothetimelysupplyanddemandinformation),weconsiderthatwhenthereexists energysurplusinsomeMGs,inordertoreducetheCO emissionoperationcost,wecandevelop 2 aneffectivepowerschedulingschemeforthelocalunitandarrangethemostchargedEVstofully utilizethedistributedpowergeneration. Then,theremainingpowercanbedeliveredtootherMGsfor thepurposeofloadoffsetandEVcharging. Inaddition,weneedtodischargeEVswhenthepower shortage occurs. The discharged power can not only address the need of supply, but also reduce substantialgenerationcost. Asthe“centralnervous”ofthewholesystem,theresponsibilityoftheMGCC(MicroGridCenter Controller)istominimizetheCO emissionandoperationcost. WeassumethatallMGsarehave 2 the same interests and the surplus power transport among them is free. Nonetheless, because of thetransmissionvoltageandresistanceofthelinebetweentheMGs,powerlossesareunavoidable whentransmittingpowerinMV(MotorVehicle)networks[10]. Obviously,theaggregatorisableto preferentiallyallocatethetransportedpowerbetweentheMGswithlesspowerlosses.Accordingtothe collectedinformationassociatedwiththeelectricalquantityandRESgenerationfromMGs,theMGCC cancomputetheoptimizedschedulingplanfortheEVsanddeployanumberofEVsforcharging. When the RES generation in some MGs is not enough for the load demand (including EVs), they will first send a request signal to the MGCC to ask for help. If there is any surplus energy providedbyotherMGs,theMGCCwilldispatchtheremainingpowertotherequesterautomatically tomeettheloaddemandandEVchargingactivities. Ifthereisnoremainingenergyleft,theMGCC willfirstexploitthefullychargedEVtodischargewithoutaffectingitsnormaloperation,andthen, startupthelocalunittomeetthepowershortage. WeconsiderEVsasthestoragedeviceinthesystem, thustheunusedenergyafterschedulingwillnotaffectthescheduleinthenexttimeslot. Inaddition, weassumethatEVsonlychargeanddischargeonceaday. Therefore,theschedulingateachperiod willbeindependent,meaningthattheschedulingresultsattimeslottwillnotaffecttheresultatany other time slot. Once the schedule is complete, the historical data will be stored. The MGCC will monitortheoperationofthesystemagainandpreparefortheschedulingatthenexttimeslot. 3. OurApproach In this section, we first present the basic idea of our approach. We then give the formulated optimizationmodelandintroducetheproposedschedulingschemes. Table1showallkeynotations usedinthepaper. Sensors2016,16,907 5of22 Table1.Notations. NG,NM,NT,NS: Thenumberoflocalunits,MGs,timeslots,andscenarios φ : TheminimumbatteryenergystoredforhandlingEV’snormaldrivingactivities min ω: Weightingfactor µ: Compensationfactorofpricegaps λ ,λ : Penaltyfactorofbatterycapacitydegradationandpowerlosses 1 2 η ,η : Charging/dischargingefficiencyofstoragebattery cha dis a ,b : FuelconsumptioncoefficientsofDGjinMGi i,j i,j α ,β : TheoperationcostcoefficientsofDGjinMGi i,j i,j E(P ): CostofunitrtocompensateRESpredictionerrorsinMGiattimet i,r,t ncha,ndis: ThenumberofEVscharged/dischargedattimetinMGi i,t i,t r ,r : Minimumandreal-timeelectricitypriceduringtheday($/kWh) min it R : LineresistancebetweenMGiandl(Ω) i,l V: TransmissionvoltageamongMGs(kV) Q : TransportedpowerbetweenMGiandl(kW) i,l N ,N ,N : ThenumberofEVs,CVsandtotalvehiclesinMGi EV,i CV,i V,i P : Thepowerlossesduringpowertransmission(kW) loss P : Powergenerationoflocalunitj(kW) i,j,t P : Non-EVloadinMGiattimet(kW) load,i,t P ,P : PowergenerationofPVandwind(kW) Pv,i,t Wind,i,t SU ,SD : Startupandshutdowncostofunitj($) i,j i,j Ru ,Rd : Ramp-up/downlimitofunitj(kW) i,j i,j Pcha ,Pdis : Charging/dischargingpowerofthekthEV(kWh) EV,i,t,k EV,i,t,k Pmin,Pmax: Minimum/maximumpowergenerationofunitj(kW)inMGi i,j i,j Pmax,Pmin: ThemaximumandminimumcapacityofEV’sbattery(kWh) EV EV Charging/dischargingstatusofEV,where0indicateschargingand1indicates I ,U : i,t i,t discharging,respectively Operationstatusofunitj,where0meanstheunit’sstopstatusand1indicates x i,j,t: itsoperationstatus,respectively Startupandshutdownstatusofunitj,where0and1meansstartupand y ,z : i,j,t i,j,t shutdown,respectively SlopeCoefficientandInterceptCoefficientoffuelconsumptionper SC,IC: unitgeneration 3.1. BasicIdea ToimprovetheenergyefficiencyandreducethecostofIMG,inthispaper,weproposeanovel system,named“archipelagomicrogrid”structure. Theproposedsystemhastheabilityofenabling theinteractionamongMGsandisolatingthemfromtheutilitygird,aswellastheconstantandlow electricitypriceoftransportedpoweramonglocalMGs. Notice that the diesel generators used in the archipelago microgrid will lead to lots of CO 2 emissions.TheexistenceofCVsareasourceofpollutionaswell.Withmoreattentiontoenvironmental issues, emission cost has become an essential aspect in MGs. In addition, the startup/shutdown associatedcost,thelifelossoflithiumbatteries,andtheamountoffuelutilizedforpowergeneration leadtoagreateroperationcost. Thus,itisrequiredforanoptimizedpenetrationofEVstoprotectthe Sensors2016,16,907 6of22 environmentandrealizeagreatreductiononCO emission. Inaddition,asthestoragedeviceisinthe 2 archipelagomicrogrid,wecanmakeabetterresponsetothetimelyloadlevelsbyregulatingthenumber of EVs to be charged at low demand intervals or discharged during the power shortage intervals. Bydoingso,areductionofemissionsandgenerationcostcanbeachieved. Nonetheless,anincrease ofothercostssuchasbatterycapacitydegradationandpowerlosseswillbeintroduced. Tothisend, anoptimizationproblemaimstominimizetheemissionandoperationcostassociatedwiththescale (number)ofEVsisproposedinthepaper. Theobjectivecontainstwoaspectsofcosts: emissionand operation. Inordertoincreasetheflexibilityintheschedulingprocess,wesetthedifferentweighted ratiosfortheenvironmentalandeconomicbenefits. Thelargerweightedfactorofemissioncostset, themoreattentionisfocusedonenvironmentalissues,andviceversa. Nonetheless,variabilitiesanduncertaintiesraisedbyRESsmaketheenergyresourcemanagement inthesystemwithMGsremainachallengingissue. Onthesupplyside,unlikethetraditionalenergy resources,windandsolarpoweroutputarehighlyuncertainandunpredictable. Evenasmallerrorin thepredictioncanresultingreaterrorsinreal-timeoperations. Onthedemandside,factors(natural disasters,plug-invehicles,personalhabitsofusingenergy,weatherandtemperature,etc.) makeit difficulttoaccuratelypredicttheusageofenergy. TheeffectivenessofintegratingMGscanbeaffected by those uncertainties, including the local generator output, reverse requirement, etc. To this end, inthispaper,weformalizetheoptimalEVpenetrationproblemasatwo-stagestochasticprogramming problem. Intheformalizedproblem, wefirstconsidertheparametersthatcouldcapturedifferent uncertainties,includingtherandomnessofrenewableenergyresourcepredictionsuchaswindandPV. WethenleveragetheMonteCarlo-basedmethodtoconstructscenariosaccordingtothedistribution functionsofthoseparameterstocaptureuncertainties. Tominimizetheexpectedcost,weproposetwoschemestoarrangethenumberofEVchargingor discharging,whicharedenotedasUnlimitedCoordinatedScheme(UCS)andLimitedCoordinated Scheme(LCS),respectively. Inourschemes,thesurplusenergyisallowedtobetransportedamong the local MGs and the power can be used for charging the EVs and offsetting the non-EV load. Particularly,intheUCS,allthesurplusenergyisallowedtochargeasmanyEVsaspossible,while tariff compensationfortheresidentialuserswhocannotchargeEVsintheirpreferenceintervalsandthe systempeakloadlimitsareconsideredinLCS. 3.2. ProblemFormulation Wenowformalizetheoptimalemissionandoperationcostproblemasatwo-stagestochastic programming problem. We first present the objective function and then introduce the constraints. Generallyspeaking,thestochasticprogrammingprocessisamethodthataimstominimizethecostina numberofscenariosconstructedbyMonteCarlo-basedmethod,whilebeingobligatedtouncertainties intheproblem. Thebasicideaofthetwo-stageprogrammingprocessistoconductarecursionprocess tomakeacorrectivedecisionaftertheoccurrenceofrandomevents. Inourtwo-stagestochasticprocess,theinputstotheunderlyingproblemincludethescenarios generatedbytheMonteCarlo-basedmethod,whichrepresenttheuncertaintiesinrenewableenergy outputs,includingwindandsolar,aswellaselectricityprices.Theoutputsoftheoptimizationproblem arecomposedofthefirst-stagedecisionsandthesecond-stagedecisions. Thefirst-stagedecisions includethecommitmentstatusesofallconventionalunits, powerdispatchofallgenerationunits, andlossescreatedbythepowertransmissionamongMGs. Thesecond-stagedecisionsconsistofEV chargingdecisionsandreversetocompensatethepredictionerrorswhentheactualRESgeneration is lower than the predicted amount. In addition, for the purpose of comparison, a Deterministic OptimalPenetration(DOP)modelisconsideredasabaselinescheme,inwhichtheparameterssuchas renewableenergygenerationarebasedonthepersistforecast,andtheelectricitypricesareabletobe predictedbytechniques(e.g.,Auto-RegressiveandMovingAverage(ARMA)[11]andSupportVector Machines(SVM)[12]). Sensors2016,16,907 7of22 3.2.1. StochasticOptimalPenetration(SOP)Model SOPmodeling: Inordertominimizetheemissionandoperationcostwithconsideringthevariousuncertainties, theobjectivefunctionoftheStochasticOptimalPenetration(SOP)modelisasfollows: (cid:40) (cid:34) (cid:35) NMNT NG ∑ ∑ ∑ (cid:0) (cid:1) 2min w E P +E (N ) +(1−w) i,j,t CV CV,i i=1t=1 j=1 (cid:34) NG NG NS NR ∑(cid:0) (cid:1) ∑ (cid:0) (cid:1) ∑ ∑ SU +SD + Cost P + ρ E(P ) i,j i,j i,j,t s i,r,t j=1 j=1 s=1 r=1 ndis i,t (1) +λ1ηchan∑ic,htaPEchVa,i,t,kIi,t+ k∑=1PEdηVisd,iis,t,kUi,t k=1 NS ni,t (cid:35)(cid:41) +λ Ploss+ ∑ ρ µ ∑ Pcha (r −r ) 2 i,t s EV,i,t,k t min s=1 k=1 WenowexplaintheSOPmodelindetail. Theemissioncostconsistsofthelocalunitemission andCVsemission,respectively. Thestartup/shutdownandgenerationcostoflocalunits,thepenalty ofbatterycapacitydegradation,thepowerlosscreatedbytransmission,andtariffcompensationare theexamplesoftheoperationcostaccordingtoEquation(1). (cid:0) (cid:1) The detail of the objective function is that, the first part E P and second part E (N ) i,j,t CV CV,i inourmodelcapturetheemissioncostscausedbythelocalunitsandCVs, respectively. TheCO 2 emissionofthelocalunitsdependsontheconsumptionoffuel,andtheCVemissionsaredirectly basedontheirnumberandtravelingdistance. BecausethePVandwindarecleanenergyresources, therewillbenopollutioncausedduringthegenerationofsuchresources. (cid:0) (cid:1) The third part SU + SD and fourth part Cost P in the SOP model represent the i,j i,j i,j,t startup/shutdownandoperationcostoflocalunits,respectively. InourMGmodel,thedistributed generation is composed of diesel generators, PV panels and wind turbines as shown in Figure 1. Notice that as the installation cost for PV panels and wind turbines integration is associated with long-termissues,weonlyconsidertheoperationcostofdieselgeneratorsinthepaper. Inthefifthpart, E(P )isthecostassociatedwithutilizingreversestocompensatetheerrors i,r,t whenthepredictionoutputislargerthantheactualRESgeneration. Wedefineρ astheprobabilityof s scenariosanddenoteNSasthenumberofscenarios. ThesixthpartincludestheequationonthethirdlineoftheEquation(1),whichrepresentsthe penaltyoflithiumbatterydegradationinEVs. Theseventhpartλ Ploss presentsthepenaltyofthe 2 i,t powerlossraisedbythepowertransmissionamongMGs. Duetothelineresistance,powerlossis inevitablewhilethesurplusistransportedfromoneMGtoanother.Toalleviatethepowertransmission loss,aMediumVoltage(MV)betweentwogridsisused. Accordingto[10,13],thepowertransmission lossescanbeobtainedthrough R Q2 P = i,l i,l,t (2) loss V2 whereR isthelineresistancebetweenMGiandl,andV indicatesthetransfervoltagebetweenthe i,l interconnectedMGs. Inordertofacilitatethemanagementofthesystem,alltransfervoltagesbetween differentMGsareconsideredtobethesamevalue. The eighth part, which is located after the last plus sign, represents the tariff compensation, enablingarealisticconsiderationofresidents’behaviors. Inreal-worldpractice,alltheconsumers tendtochargeEVsatalowerelectricitypriceinterval. Nonetheless,duetothelimitationofloadand Sensors2016,16,907 8of22 generation capacities, it is impossible to depose all the EVs in the same time slot. To this end, the aggregatorcandeferaportionofEVstoothertimeslotswithmoreRESsandlowloads. Noticethat thisbehaviorviolatestheconsumer’saspirationandincursanextracostsincethechargingoccurs athighpriceslots. Tothisend,itisessentialtointroducetariffstomakeupfortheresidentialusers’ additionaloverheadraisedbynotbeingchargedattheirdesiredslots. Inparticular,withtheincrease ofµ,morecompensationcanbepaidandthenumberofEVscanbesignificantlyreducedtoo. SOPconstraints: Intheoptimizationproblem,theconstraintsneedtobeconsideredaswell. Thefirstconstraint is to balance power. For each MG in the system, the following power balance constraint needs to be satisfied: The first constraint is to balance the power. For each MG in the system, the following power balanceconstraintneedstobesatisfied: ni,t NG P + ∑ Q + ∑ Pcha ×I = ∑ P + load,i,t i,l,t EV,i,t,k i,t i,j,t l(cid:54)=i k=1 j=1 Pi,r,t+PPv,i,t+PWind,i,t+ ∑ Ql,i,t− ∑ Qloss,l,i,t (3) l(cid:54)=i l(cid:54)=i ni,t + ∑ Pdis I ∀i,t EV,i,t,k i,t k=1 where the left parts in the equality indicate the the local units generations, RES generations (e.g., solar and wind), the discharged power from EVs, and the transported power from other MGs,therightpartspresenttheamountofdemand,whichconsistsofnon-EVloads,theamountof powerchargedtotheEVs,powerlossesresultedfromlineresistance,andtheamountofthesurplus powertransmittedtootherMGs. Second,thelocalunitsconstraintsshouldbesatisfied. Recallthatthefuelconsumptionfeatures ofdieselgeneratorsarevariedduetodifferentlevelsofpowerrated. Ingeneral, thefuelusedfor theunitpowergenerationwillreduceastheincreaseoftheratedlevelofpowerusedbygenerators. To describethecharacteristicofdieselgenerators,weusealinearmodelforthefuelconsumptionthat considersdifferenttypesofgenerationsby[14]: FC = a P +b P ,∀i,j,t (4) i,j i,j i,j,t i,j rated,i,j where FC is the fuel consumption of diesel generator j in MG i. For the operation cost of diesel i,j generators,weusethecostmodeldevelopedby[15]asfollows: (cid:0) (cid:1) Cost P = α x +β P , ∀i,j,t (5) i,j,t i,j i,j,t i,j i,j,t wherethe Cost(.) representstheoperationcostofdieselgenerator j inMG i. Inaddition, theunit operationconstraintscanbeexpressedas: Pminx ≤ P ≤ Pmaxx i,j i,j,t i,t,j i,j i,j,t (cid:0) (cid:1) Pi,j,t−Pi,j,t−1 ≤ Rui,j (cid:0)1−yi,j,t (cid:1)+Pim,jinyi,j,t (6) Pi,j,t−1−Pi,j,t ≤ R(di,j 1−zi,j,t +Pim,jinzi,j,t yi,j,t+zi,j,t ≤1,yi,j,t−zi,j,t = xi,j,t−xi,j,t−1 where the first constraint indicates the power limit of the diesel generator, the second and third constraintsshowtheramp-up/downlimits,andthefourthconstraintindicatestherelationshipamong thedifferentoperationstatusesofgenerators. In addition, notice that the unlimited integration of EVs is unrealistic. Thus, the scale of EVs should be constrained. The number of vehicles, including CVs and EVs, can be estimated by the Sensors2016,16,907 9of22 analysisbasedonthenumberofresidentsineachMG.Weleveragethemodeldevelopedin [6,16]to conducttheestimation. Then,wehave R Q D N = V2G REC Total (7) R D Average whereD indicatesthetotalloaddemandineachMGperday. D istheaveragedailyelectricity total Average consumptionofresidents,whichisequalto2.0833kWaccordingtheresultsin[6]. Q isthevehicle REC ownershiprateofresidents,andR istheproportionofresidentswhoparticipateintheV2Gplan. V2G Apparently,withthepopularityofthevehicles,itisreasonabletoassumethatQ =1. Inthispaper, REC weconsiderthatalltheresidentsuseV2GtechnologysothatR = 100%. Thus,therelationship V2G betweenvehiclescaleandthenumberofresidentsisderivedasfollows: NT N = ∑ ncha EV,i i,t (8) t=1 N = N = N +N ,∀i V,i R,i EV,i CV,i wherethefirstequationindicatesthatthenumberofEVsiscomputedbythesumofEVsineachtime slot. ThesecondequationsmeansthatthenumberofvehiclesiscomposedofCVsandEVs. Withconsiderationofloadlimits,wealsoassumethatthesumofthenewlyaddedloadcaused byEVchargingandnon-EVloadshouldnotbelargerthanthemaximumloadcapacity(Pmax )at load,i eachtimeslot. Tothisend,thefollowingconstraintisproposedtoprotectthesystemfromgenerating newloadpeaks: ncha i,t ∑ Pcha I +P ≤ Pmax (9) EV,i,t,k i,t load,i,t load,i k=1 where the left parts in the inequality are the consumptions of EVs and non-EV loads, and Pmax load,i indicatesthelimitationsofloadinMGi. Finally,thecharginganddischargingconstraintsofEVsshouldbeconsideredaswell. NT 0≤ ∑ Pcha I ≤ PMax EV,i,t,k i,t EV t=1 NT 0≤ ∑ Pdis U ≤ (1−φ )PMax EV,i,t,k i,t min EV (10) t=1 (cid:18) (cid:19) φ PMax ≤ N∑T ηchaPcha I − PEdVis,i,t,kUi,t ≤ PMax min EV EV,i,t,k i,t ηdis EV t=1 I +U =1 i,t i,t Here,thefirsttwoconstraintsillustratethecharginganddischarginglimits,andtheφ refers min totheminimumbatteryenergystoragefortheregulardrivingofEVs. Thethirdoneguaranteesthe storedpowerinEVs. Thelastconstraintisusedtoavoidthesimultaneousoperationofchargingand dischargingatthesametimeslot. 3.2.2. DeterministicOptimalPenetration(DOP)Model IncomparisonwiththeperformanceoftheSOPmodel,aDeterministicOptimalPenetration(DOP) modelisformulatedasabaselinescheme,inwhichtheparametersassociatedwiththeuncertainties (ambienttemperature,windspeed,etc.) arebasedonapersistpredictionandtheelectricitypricesare Sensors2016,16,907 10of22 forecastedbypredictiontechniques[11]. TheconstraintsfromEquations(2)to(10)shouldalsobe satisfied. Tothisend,weintroducethefollowingDOPmodelasthebaselinescheme: (cid:40) (cid:34) (cid:35) NMNT NG ∑ ∑ ∑ (cid:0) (cid:1) 2min w E P +E (N ) i,j,t CV CV,i i=1t=1 j=1 (cid:34) NG NG ∑(cid:0) (cid:1) ∑ (cid:0) (cid:1) +(1−w) SU y +SD z + Cost P i,j i,j,t i,j i,j,t i,j,t j=1 j=1 ndis i,t (11) +λ1ηchan∑ic,htaPEchVa,i,t,kIi,t+ k∑=1PEdηVisd,iis,t,kUi,t k=1 NR ni,t (cid:35)(cid:41) + ∑ E(P )+λ P + µ ∑ Pcha (r −r ) i,r,t 2 loss EV,i,t,k t min r=1 k=1 Here,thefirstandsecondpartsintheDOPmodelindicatetheemissioncostraisedbythelocalunits andCVs. Thethirdandfourthpartsarethestartup/shutdownandoperationcostoflocalgenerators. ThefifthandsixtharethepenaltyofbatterydegradationofEVsandthepowerlosscreatedbypower transmission among local MGs. The seventh part part is the tariff compensation to make up the overheadoftheusersforcharingEVsatthetimeslotswithhigherprices. Notice that, in comparison with the SOP model, the uncertainties of the renewable energy resources are not considered and addressed. In addition, as essential energy resources in the "archipelago microgrids", the uncertainties raised by RESs could affect the operation of system. Therefore, the DOP model will result in a higher operation cost than the SOP model. Because of this,moretraditionalenergyresourcescouldbeutilizedtocompensateforthepredictionerrorsofthe generationfromRESs(i.e.,theactualgenerationissmallerthantheprediction). Tothisend,theDOP modelisproposedasabaselineschemetoevaluatetheeffectivenessoftheproposedSOPmodel. 3.3. ProposedSchedulingSchemes We now present the two proposed scheduling schemes to carry out the optimal scale of EVs effectively. Todemonstratetheeffectivenessofthetwoschemes,abaselineallocationschemeforEVs, namedUncoordinatedAverageScheme(UAS),isalsoconsidered. Inthefollowing,wedescribethese threeschedulingschemesindetail. 3.3.1. UncoordinatedAverageScheme(UAS) Sincethenon-EVloadsareinflexible,theaverageallocationmethodforEVsisafeasiblewayto reducetheemission. IntheUAS,thesurplusofpowerisusedtosupplyloadspreferentially,andthen, theremainingpowerisusedtochargeEVs. ByaccumulatingallthesurpluspowerineachMG,the maximumallowablenumberofEVscanbederived. Inparticular,ifthereisnotenoughRESpowerto meettheaveragedemandofEVsatthetimeslott,theexceededEVswillbedeferredtothenexttime slot. Byusingthisscheme,wecanplanthemaximumnumberofEVsineachMGtoreachtheupper limitofvehiclescale. NoticethattheUASistheleastdesirablescenariosincethesurpluspoweris unabletotransportamonglocalMGs. Itmeansthatalloftheexcessenergyafterfulfillingtheload andEVschargingwillbedissipated. Inthisscheme,eachMGoperatesasanindependentindividual withouttheinterconnectionwitheachother,whichwilldirectlyresultinlessaccommodatedEVsand lowenergyefficiency.
Description: