Page - 333 - in Short-Term Load Forecasting by Artificial Intelligent Technologies

Energies2018,11, 1449 FromFigure12, it canbeseenthatFC-WPA-LSSVMpresents the lowestRMSE,MAPEandAAE, withcorrespondingvaluesof2.07%,1.92%and1.97 in thespringtest, 2.29%,2.20%and2.11%inthe summertest, 2.39%,2.35%and2.25%intheautumntest, and2.08%,1.90%and1.84%inthewinter test. It canbeseenthat theoverallpredictionperformanceof the forecastingapproachwasoptimal dueto theadvantagesofFC,WPAandLSSVM.Inconclusion, the loadforecastingmodel fore-bus chargingstationsbasedonFC-WPA-LSSVMcanprovideaccuratedatasupport for theeconomical operationof thestation. Inaddition, theproposedmodelcanalsobeapplied to the loadforecastingof otherchargingstations,anditspredictionaccuracywillnotbeaffectedbychanges in thenumberof electricvehiclesandother factors. Since this forecastingmodel isbasedonMATLABdevelopment, if the transportationcompany wants touse thismodel topredict the load in the future, theycanalsouse it easily andobtain the forecast resultswithoutadditionalcosts. 6.Conclusions Inviewof the loadcharacteristics fore-buschargingstations, thispaperselectedeightvariables, includingdaytype,maximumtemperature,minimumtemperature,weathercondition, thenumber ofaccumulateddailynumberofchargede-busesandthe loadsat thesamemoment in theprevious threedays, as the input. Anovel short-termload forecasting technique fore-buschargingstations basedonFC-WPA-LSSVMwasproposed, inwhichFCwasusedtoextract similardatesas training samples,andWPAwasintroducedtooptimize theparameters inLSSVMtoimprovetheprediction accuracy. Twocase studieswere carriedout toverify thedevelopedapproach incomparisonwith WPA-LSSVM,LSSVMandBPNN.Theexperimental results showedthat the forecastingprecisionof theproposedmodelwasbetter thanthecontrastingmodels.Hence,FC-WPA-LSSVMprovidesanew ideaandreference forshort-termloadforecastingofe-buschargingstations. Theloadofe-buschargingstations isakindofpowerloadwithcomplexchangerulesanddiverse influential factors.With the large-scaleapplicationofelectricvehicles,moreandmoree-buscharging stationswill start tobeput intouse. At that time, researchonactualoperationofchargingstations willbemoreabundant. It isnecessarytomakefurtherefforts toseekmoresuitable loadforecasting approaches for e-bus charging stationsbasedon the studyof loadvariation rules and the internal relationshipsbetweenthe loadandinfluential factors. Funding: This work is supported by the Fundamental Research Funds for the Central Universities (ProjectNo. 2018MS144). Acknowledgments:ThanksforStateGridHebeiElectricPowerCompanyprovidingtherelevantdatasupporting. Conflictsof Interest:Theauthorsdeclarenoconflictof interest. References 1. Tan,S.;Yang, J.;Yan, J.;Lee,C.;Hashim,H.;Chen,B.AHolistic lowCarboncity IndicatorFrameworkfor SustainableDevelopment.Appl. Energy2017,185, 1919–1930. [CrossRef] 2. Yan, J.;Chou,S.K.;Chen,B.;Sun,F.; Jia,H.;Yang, J.Clean,AffordableandReliableEnergySystemsfor low CarboncityTransition.Appl. Energy2017,194, 305–309. [CrossRef] 3. Majidpour,M.;Qiu,C.;Chu,P.;Pota,H.;Gadh,R.ForecastingtheEVCharging loadBasedonCustomer ProfileorStationMeasurement?Appl. Energy2016,163, 134–141. [CrossRef] 4. Paparoditis, E.; Sapatinas, T. Short-TermLoad Forecasting: The Similar Shape Functional Time-Series Predictor. IEEETrans. PowerSyst. 2013,28, 3818–3825. [CrossRef] 5. Yildiz,B.;Bilbao, J.I.; Sproul,A.B.AReviewandAnalysisofRegressionandMachineLearningModelson CommercialBuildingElectricity loadForecasting.Renew. Sust. EnergyRev. 2017,73, 1104–1122. [CrossRef] 6. Cerne,G.;Dovzan,D.;Skrjanc, I. Short-termloadforecastingbyseparatingdailyprofileandusingasingle fuzzymodelacross theentiredomain. IEEETrans. Ind. Electron. 2018,65, 7406–7415. [CrossRef] 7. Ashtari,A.;Bibeau,E.; Shahidinejad,S.;Molinski,T.PEVChargingProfilePredictionandAnalysisBasedon VehicleUsageData. IEEETrans. SmartGrid2012,3, 341–350. [CrossRef] 333