Page - 320 - in Short-Term Load Forecasting by Artificial Intelligent Technologies

Energies2018,11, 1449 and BA is that they easily fall into local optimums, leading to reduction in prediction accuracy. Wolfpackalgorithm(WPA),asanewmetaheuristicapproach, is introducedin thispaper tooptimize theparameters inLSSVM.This techniquepossesses goodglobal convergence and computational robustnessdueto insensitivityof thechangeofparameters inWPA[24]. As a result of the complexity and diversity of the influential factors for load forecasting in quick-changee-buschargingstations, it isofgreatnecessity toselectproper inputs for theprediction, sothatredundantdatacanbereducedandcomputingefficiencycanbeimproved[25]. Fuzzyclustering (FC) isamathematical techniquethatclassifiesobjectsaccordingto theircharacteristics [26]. Inview of the fact that thedaily loadcurveswithsimilar influential factorsofchargingstationsarebasically consistent,goodpredictionresultscanbeachievedbytheuseofsamplesonsimilardays.Consequently, a transitiveclosurealgorithmgroundedonafuzzyequivalentmatrix inFCisselected in thispaper, whichcanextract samplessimilar to thepredictedday. It cannotonlyavoid theblindnessofchoosing similardaysbyexperience,butalsoovercometheadverseeffectsofunconventional loaddatacaused bysuddenchangeof factorsonLSSVMtraining. Therefore, the influential factors for the loadinquick-changee-buschargingstationsareanalyzed in this paper, and a load forecastingmodel combining FCwith LSSVMand optimized byWPA (FC-WPA-LSSVM)isestablishedhere. Therestofpaper isorganizedas follows: Section2conducts ananalysisof thedaily loadcharacteristics forquick-changee-buschargingstationsbasedonrelated statisticaldataandstudiesvarious influential factors includingdaytypes,meteorological conditions and bus dispatch; Section 3 provides a brief description of FC, LSSVMandWPA, aswell as the completepredictionframework;Section4 introducesanexperimental studytovalidate theproposed method;andSection5makes furthervalidation. InSection6, conclusionsareobtained. 2.AnalysisofLoadCharacteristicsofE-BusChargingStations The loadofa largequick-changee-buschargingstation inBaoding,China, isprovided in this paper.Whenthebuscomes into thestation, thebatterywithelectricitydepletion ischangedbythe quick-changerobot,which is furtherconnectedto thechargingplatform.Then,abatteryfilledwith electricity is installed in thebus. After that, the e-busgoes intoa specificarea towait fordispatch instructions.Accordingto thedispatch, thee-busappearsat thechargingstationafter8:00a.m. each day,which leads toarise in load. Thechargerswillnot stopworkinguntil thebatterychargingof the laste-bus iscompleted.At that time, the loaddecreases to the lowestpoint. A typical daily load curveof the e-bus charging station is shown inFigure 1,whichdisplays theactivepowerperhour inaday. Incommonwith the traditional loadcurve, thereexistobvious crestsandtroughs.However, thecurveof thee-buschargingstationfluctuatesgreatly,andapparent distinctionsappearamongdifferentcurves,wherebythe load inwinterandsummer ishigh,while the load in spring andautumn is low. All of these characteristics createdifficulties for thedaily load forecastingof thechargingstation. Figure1.Typicaldaily loadcurveofane-buschargingstation inBaoding. 320