Page - 23 - in Short-Term Load Forecasting by Artificial Intelligent Technologies

energies Article AHybridSeasonalMechanismwithaChaotic CuckooSearchAlgorithmwithaSupportVector RegressionModelforElectricLoadForecasting YongquanDong,ZichenZhangandWei-ChiangHong* ID SchoolofComputerScienceandTechnology(SchoolofEducationIntelligentTechnology), JiangsuNormal University/101,ShanghaiRd.,TongshanDistrict,Xuzhou221116, Jiangsu,China; tomdyq@jsnu.edu.cn(Y.D.); zzcpkzzw@126.com(Z.Z.) * Correspondence: samuelsonhong@gmail.com;Tel.:+86-516-8350-0307 Received: 24March2018;Accepted: 18April2018;Published: 20April2018 Abstract: Providing accurate electric load forecasting results plays a crucial role in daily energy managementof thepowersupplysystem.Duetosuperior forecastingperformance, thehybridizing support vector regression (SVR)modelwith evolutionary algorithmshas receivedattention and deserves tocontinuebeingexploredwidely. Thecuckoosearch(CS)algorithmhas thepotential to contributemore satisfactoryelectric load forecasting results. However, theoriginalCSalgorithm suffers from its inherent drawbacks, such as parameters that require accurate setting, loss of populationdiversity, andeasy trapping in local optima (i.e., premature convergence). Therefore, proposing somecritical improvementmechanismsandemployingan improvedCSalgorithmto determinesuitableparametercombinations foranSVRmodel isessential. Thispaperproposes the SVRwithchaoticcuckoosearch(SVRCCS)modelbasedonusingatentchaoticmappingfunction toenrich the cuckoosearch spaceanddiversify thepopulation toavoid trapping in local optima. In addition, to dealwith the cyclic nature of electric loads, a seasonalmechanism is combined with the SVRCCSmodel, namely giving a seasonal SVRwith chaotic cuckoo search (SSVRCCS) model, to produce more accurate forecasting performances. The numerical results, tested by using the datasets from theNational ElectricityMarket (NEM,Queensland, Australia) and the NewYork Independent SystemOperator (NYISO,NY,USA), show that theproposedSSVRCCS modeloutperformsotheralternativemodels. Keywords: support vector regression; tent chaoticmapping function; cuckoo search algorithm; seasonalmechanism; loadforecasting 1. Introduction Accurate electric load forecasting is important to facilitate the decision-making process for powerunit commitment, economic loaddispatch,powersystemoperationandsecurity, contingency scheduling,andsoon[1,2].Asindicatedinexistingpapers,a1%electric loadforecastingerror increase would lead toa£10millionadditional operational cost [3], on the contrary,decreasing forecasting errorsby1%wouldproduceappreciableoperationbenefits [2]. Therefore, lookingformoreaccurate forecastingmodelsorapplyingnovel intelligentalgorithmstoachievesatisfactory loadforecasting results, tooptimize thedecisionsofelectricitysuppliesandloadplans, to improve theefficiencyof the power systemoperations, eventually, reduces the system risks towithin a controllable range. However,dueto lotsof factors, suchasenergypolicy,urbanpopulation, socio-economicalactivities, weatherconditions,holidays,andsoon[4], theelectric loaddatadisplayseasonality,non-linearity, andachaoticnature,whichcomplicateselectric loadforecastingwork[5]. Energies2018,11, 1009;doi:10.3390/en11041009 www.mdpi.com/journal/energies23