Page - 361 - in Short-Term Load Forecasting by Artificial Intelligent Technologies

Energies2018,11, 1253 (3)NILAoptimization. SearchtheoptimalweightsandthresholdsofCNNonthebasisofNILA. If themaximumiterationnumber is reached, theoptimalparametersareobtained; ifnot, repeat the optimizationstepsuntil thecondition issatisfied. (4) CNN training. After initialization including the neuron numbers in the input layer, convolutional layer,andsubsamplinglayer,respectively, traintheCNNoptimizedbyNILA,andderive theoptimal forecastingmodel. (5)Simulationandprediction. Forecast theshort-termloadofEVchargingstationsbasedonthe trainedapproachandanalyze theresults. 3.AnalysisofLoadCharacteristics inElectricVehicle (EV)ChargingStation Thestudyofinfluentialfactorsthataffecttheloadinchargingstationcontributetoloadforecasting accuracy improvement. ThispaperselectsanEVchargingstation inBeijingasacasestudy. It canbe seenthat theloadisheavily influencedbymeteorologicalconditions, seasonalvariation,anddaytypes. 3.1. SeasonalVariation Seasonalvariationhasanobviouseffectonthe loadcharacteristics inEVchargingstation[27]. Therefore, the typical daily load curves in spring, summer, autumn andwinter are compared in Figure4. It shouldbenotedthat these fourdaysareallTuesday,andareall sunnydays. Figure4.Typicaldaily loadcurves in fourseasons. Aspresented inFigure 4, the loadof theEVcharging station is relativelyhigh inwinter and summer,mainlydueto increasinguseofairconditioning in these twoseasons,which leads tomore energyconsumption.Asaresult, air conditioning loadcanbeconsideredasavital influencingfactor. 3.2.MeteorologicalConditions The load inEVchargingstation isgreatlyaffectedbytemperatureandweather type,whilewind andhumidityplay insignificantroles [28,29].Here, take thedaily loadcurveson1June,8 Juneand 15June in2017asexamples. Theaveragedaily temperaturesare23.5 ◦C,27 ◦Cand31 ◦C,respectively. It canbeseenthat there isapositiverelationshipbetweentemperatureanddaily load,asshownin Figure5. Therefore, temperature is selectedas the influential factor in thispaper. 361