Page - 354 - in Short-Term Load Forecasting by Artificial Intelligent Technologies

energies Article Short-TermLoadForecastingforElectricVehicle ChargingStationBasedonNicheImmunityLion AlgorithmandConvolutionalNeuralNetwork YunyanLi*,YuanshengHuangandMeimeiZhang DepartmentofEconomicManagement,NorthChinaElectricPowerUniversity,Baoding071000,China; 51850962@ncepu.edu.cn(Y.H.); 51851539@ncepu.edu.cn(M.Z.) * Correspondence: liyunyanbd@126.com Received: 20April2018;Accepted: 10May2018;Published: 14May2018 Abstract:Accurateandstablepredictionofshort-termloadforelectricvehiclechargingstations is ofgreat significance inensuringeconomicalandsafeoperationofelectricvehiclechargingstations andpowergrids. Inorder to improvetheaccuracyandstabilityofshort-termloadforecastingfor electricvehiclechargingstations,an innovativepredictionmodelbasedonaconvolutionalneural networkandlionalgorithm, improvedbyniche immunity, isproposed. Firstly,niche immunity is utilized torestrictoverduplicationofsimilar individuals, soas toensurepopulationdiversityof lion algorithm,which improves theoptimizationperformanceof the lionalgorithmsignificantly. The lion algorithmis thenemployedtosearch theoptimalweightsandthresholdsof theconvolutionalneural network. Finally, aproposed short-term load forecastingmethod is established. After analyzing the loadcharacteristicsof theelectricvehiclechargingstation, twocases indifferent locationsand differentmonthsareselectedtovalidatetheproposedmodel. Theresults indicatethat thenewhybrid proposedmodeloffersbetteraccuracy, robustness,andgenerality inshort-termloadforecastingfor electricvehiclechargingstations. Keywords: electricvehicle (EV)chargingstation; short-termloadforecasting;niche immunity (NI); lionalgorithm(LA); convolutionalneuralnetwork(CNN) 1. Introduction The development of the electric vehicle (EV) industry has attracted broad attention from governments, automanufacturers, and energy enterprises. Electric vehicles are regarded as an effectiveway to copewith the depletion of fossil energy and increasingly serious environmental pollution [1]. Charging stations, serving as the infrastructure, have been extensively built along with theadvanceofEVs. However, thevolatility, randomness, and intermittenceof the loadbring newchallenges to optimal dispatching and safe operation of power grids [2]. The establishment of a scientific and reasonable short-term load forecastingmodel for EVcharging stationswill not only improve thepredictionprecision for optimal dispatching, butwill alsopromote the rational constructionof charging stations, andboost thepopularity rate ofEVs. Accordingly, focuson the researchof loadforecastingforEVchargingstations isofgreat significance. Thecurrentmethodsof loadforecastingforEVchargingstationscanbedividedinto twoparts, namely: statisticalapproachesandartificial intelligentalgorithms. Statistical forecastingmodelsare basedon the theoryofprobability and statistics, suchas theMonteCarlomethod [3]. Concretely, onthefoundationofaresidents’ trafficbehaviordatabase, theMonteCarloapproachexploitsadefinite probabilitydistribution function tofit theusers’drivingbehaviors, andestablishesamathematical modelwithrandomprobability to forecast thecharging time, location,and loaddemandofEVs in the future [4]. Simple though it is, thiskindofmethod isnot suited toaddress load forecasting for Energies2018,11, 1253;doi:10.3390/en11051253 www.mdpi.com/journal/energies354