Seite - 372 - in Short-Term Load Forecasting by Artificial Intelligent Technologies

energies Article Short-TermLoadForecastingwithMulti-SourceData UsingGatedRecurrentUnitNeuralNetworks YixingWang ID ,MeiqinLiu* ID ,ZhejingBaoandSenlinZhang CollegeofElectricalEngineering,ZhejiangUniversity,Hangzhou310027,China;wangyixing@zju.edu.cn (Y.W.); zjbao@zju.edu.cn(Z.B.); slzhang@zju.edu.cn(S.Z.) * Correspondence: liumeiqin@zju.edu.cn;Tel.: +86-139-5800-7313 Received: 29March2018;Accepted: 1May2018;Published: 3May2018 Abstract: Short-termloadforecasting isan important task for theplanningandreliableoperation ofpowergrids. High-accuracy forecasting for individual customershelps tomakearrangements for generation and reduce electricity costs. Artificial intelligentmethods have been applied to short-termloadforecasting inpast research,butmostdidnotconsiderelectricityusecharacteristics, efficiency, andmore influential factors. In this paper, amethod for short-term load forecasting withmulti-sourcedatausinggated recurrentunitneuralnetworks isproposed. The loaddataof customersarepreprocessedbyclusteringtoreduce the interferenceofelectricityusecharacteristics. Theenvironmental factors includingdate,weatherandtemperaturearequantifiedtoextendtheinput of thewholenetworkso thatmulti-source information is considered. Gatedrecurrentunitneural networksareusedforextractingtemporal featureswithsimplerarchitectureandlessconvergence time in the hidden layers. The detailed results of the real-world experiments are shownby the forecastingcurveandmeanabsolutepercentageerror toprove theavailabilityandsuperiorityof the proposedmethodcomparedto thecurrent forecastingmethods. Keywords: short-termloadforecasting;artificial intelligence;gatedrecurrentunit; recurrentneural network;powergrid 1. Introduction Loadforecasting isanessentialpart forenergymanagementanddistributionmanagement in powergrids.With thecontinuousdevelopmentof thepowergridsandthe increasingcomplexityof gridmanagement, accurate loadforecasting isachallenge [1,2].High-accuracypower loadforecasting for customers canmake the reasonable arrangements of power generation tomaintain the safety andstabilityofpowersupplyandreduceelectricitycosts so that theeconomicandsocialbenefit is improved.Moreover, forecastingat individual customer level canoptimizepowerusageandhelp tobalance the loadandmakedetailedgridplans. Loadforecasting is theprocessofestimating the future loadvalueatacertain timewithhistorical relateddata,whichcanbedividedinto long-term load forecasting,medium-term load forecasting and short-term load forecasting according to the forecasting time interval. Short-term load forecasting,which thispaper focuseson, is thedailyor weekly forecasting[3,4]. It isusedfor thedailyorweeklyschedule includinggeneratorunit control, loadallocationandhydropowerdispatching.With the increasingpenetrationof renewableenergies, short-termloadforecasting is fundamental for thereliabilityandeconomyofpowersystems. Modelsof short-termloadforecastingcanbeclassified into twocategoriesconsistingof tradition statisticmodelsandartificial intelligentmodels. Statisticmodels, suchasregressionanalysismodels andtimesequencemodels, are researchedandusedfrequentlywerepreviously limitedbycomputing capability. Tayloretal. [5]proposedanautoregressive integratedmovingaverage(ARIMA)model with an extension ofHolt–Winters exponential smoothing for short-term load forecasting. Then, Energies2018,11, 1138;doi:10.3390/en11051138 www.mdpi.com/journal/energies372