Page - 417 - in Short-Term Load Forecasting by Artificial Intelligent Technologies

energies Article AHighPrecisionArtificialNeuralNetworksModel forShort-TermEnergyLoadForecasting Ping-HuanKuo1 ID andChiou-JyeHuang2,* ID 1 ComputerandIntelligentRobotProgramforBachelorDegree,NationalPingtungUniversity, Pingtung90004,Taiwan;phkuo@mail.nptu.edu.tw 2 SchoolofElectricalEngineeringandAutomation, JiangxiUniversityofScienceandTechnology, Ganzhou341000, Jiangxi,China * Correspondence: chioujye@163.com;Tel.:+86-137-2624-7572 Received: 14December2017;Accepted: 9 January2018;Published: 16 January2018 Abstract:Oneof themost important research topics in smartgrid technology is load forecasting, becauseaccuracyof loadforecastinghighly influencesreliabilityof thesmartgridsystems. In the past, loadforecastingwasobtainedbytraditionalanalysis techniquessuchas timeseriesanalysis andlinear regression. Since the loadforecast focusesonaggregatedelectricityconsumptionpatterns, researchershaverecently integrateddeep learningapproacheswithmachine learning techniques. In thisstudy,anaccuratedeepneuralnetworkalgorithmforshort-termloadforecasting(STLF) is introduced. Theforecastingperformanceofproposedalgorithmiscomparedwithperformancesof fiveartificial intelligencealgorithmsthatarecommonlyused in loadforecasting. TheMeanAbsolute PercentageError (MAPE)andCumulativeVariationofRootMeanSquareError (CV-RMSE)areused asaccuracyevaluation indexes. Theexperimentresults showthatMAPEandCV-RMSEofproposed algorithmare9.77%and11.66%,respectively,displayingveryhighforecastingaccuracy. Keywords: artificial intelligence; convolutionalneuralnetwork;deepneuralnetworks; short-term loadforecasting 1. Introduction Nowadays, thereisapersistentneedtoacceleratedevelopmentof low-carbonenergytechnologies inorder toaddress theglobal challengesof energysecurity, climate change, andeconomicgrowth. The smart grids [1] are particularly important as they enable several other low-carbon energy technologies [2], including electric vehicles, variable renewable energy sources, and demand response. Due to thegrowingglobal challengesof climate, energysecurity, andeconomicgrowth, accelerationoflow-carbonenergytechnologydevelopmentisbecominganincreasinglyurgentissue[3]. Amongvariousgreentechnologies tobedeveloped, smartgridsareparticularly importantas theyare keyto the integrationofvariousother low-carbonenergytechnologies, suchaspowerchargingfor electricvehicles,on-gridconnectionof renewableenergysources,anddemandresponse. The forecast of electricity load is important for power systemscheduling adoptedby energy providers [4]. Namely, inefficient storage and discharge of electricity could incur unnecessary costs,whileevenasmall improvement inelectricity loadforecastingcouldreduceproductioncosts and increase tradingadvantages [4], particularlyduring thepeakelectricity consumptionperiods. Therefore, it is important forelectricityproviders tomodelandforecastelectricity loadasaccuratelyas possible, inbothshort-term[5–12] (onedaytoonemonthahead)andmedium-term[13] (onemonth tofiveyearsahead)periods. Withthedevelopmentofbigdataandartificial intelligence(AI) technology,newmachinelearning methodshavebeenapplied to thepower industry,where largeelectricitydataneedtobecarefully Energies 2018,11, 213;doi:10.3390/en11010213 www.mdpi.com/journal/energies417