Page - 3 - in Short-Term Load Forecasting by Artificial Intelligent Technologies

Energies2018,11, 2226 inequalities [20,21]. Focusedon the advantages of theLS-SVRmodel todealwith suchproblems, thispapertriestosimulatethenonlinearsystemoftheMELtimeseriestoreceivetheforecastingvalues andimprovethe forecastingaccuracy.However, thedisadvantagesof theSVR-basedmodels in load forecastingare thatwhenthesamplesizeof the load is large, the timeofsystemlearningandtraining ishighly time-consuming,andthedeterminationofparametersmainlydependsontheexperienceof theresearchers. Thishasacertaindegreeof influenceontheaccuracy in loadforecasting. Therefore, exploringmoresuitableparameterdeterminationmethodshasalwaysbeenaneffectivewaytoimprove the forecastingaccuracyof theSVR-basedmodels. Todeterminemoreappropriateparametersof the SVR-basedmodels,Hongandhiscolleagueshaveconductedresearchusingdifferentevolutionary algorithmshybridizedwithanSVRmodel[22–24]. Inthemeantime,Hongandhissuccessorshavealso applieddifferentchaoticmappingfunctions(includingthelogisticfunction[22,23]andthecatmapping function[10]) todiversify thepopulationduringmodelingprocesses,andthecloudtheorytomake sure the temperaturecontinuouslydecreasesduringtheannealingprocess, eventuallydeterminingthe mostappropriateparameters toreceivemoresatisfactory forecastingaccuracy[10]. The fruitflyoptimizationalgorithm(FOA) isanewswarmintelligentoptimizationalgorithm proposed in 2011, it searches for global optimization based on fruit fly foraging behavior [25,26]. The algorithmhas only four control parameters [27]. Comparedwith other algorithms, FOAhas theadvantagesofbeingeasytoprogramandhavingfewerparameters, lesscomputation,andhigh accuracy [28,29]. FOAbelongs to thedomainofevolutionarycomputation; it realizes theoptimization ofcomplexproblemsbysimulatingfruitflies tosearchfor foodsourcesbyusingolfactionandvision. It has been successfully applied to thepredictive control fields [30,31]. However, similar to those swarmintelligentoptimizationalgorithmswith iterativesearchingmechanisms, thestandardFOA alsohasdrawbacks such as apremature convergent tendency, a slowconvergent rate in the later searchingstage,andpoor local searchperformance [32]. Quantumcomputinghasbecomeoneof the leadingbranchesof science in themoderneradueto itspowerful computingability. Thisnotonlypromptedus tostudynewquantumalgorithms,butalso inspiredus to re-examine some traditional optimizationalgorithms fromthequantumcomputing mechanism. Thequantumcomputingmechanism(QCM)makes fulluseof the superpositionand coherence of quantum states. Compared with other evolutionary algorithms, the QCM uses a novel encodingmethod—quantumbit encoding. Through the encoding of qubits, an individual can characterize any linear superposition state, whereas traditional encodingmethods can only representonespecificone.Asaresult,withQCMit iseasier tomaintainpopulationdiversity than with other traditional evolutionary algorithms. Nowadays, it has become ahot topic of research thatQCMisable tohybridizewithevolutionaryalgorithms to receivemore satisfactory searching results. The literature [33] introducedQCMintogeneticalgorithmsandproposedquantumderived geneticalgorithm(QIGA).Fromthepointofviewofalgorithmicmechanism, it isverysimilar to the isolatednichesgeneticalgorithm.HanandKim[34]proposedageneticquantumalgorithm(GQA) based onQCM.Comparedwith traditional evolutionary algorithms, its greatest advantage is its betterability tomaintainpopulationdiversity.HanandKim[35] further introducedthepopulation migrationmechanismbasedonure [34], andrenamedthealgorithmaquantumevolutionalgorithm (QEA).Huang[36],LeeandLin[37,38],andLietal. [39]hybridizedtheparticleswamoptimization (PSO)algorithm,Tabusearch(TS)algorithm,geneticalgorithm(GA),andbatalgorithm(BA)with the QCMandthecatmappingfunction,andproposedtheCQPSO,CQTS,CQGA,andCQBAalgorithms, whichwere employed to select the appropriate parameters of an SVRmodel. The results of the application indicate that the improvedalgorithmsobtainmoreappropriateparameters,andhigher forecastingaccuracy is achieved. Theaboveapplications also reveal that the improvedalgorithm, byhybridizingwithQCM,couldeffectivelyavoid localoptimalpositionandprematureconvergence. 1.3. Contributions Considering the inherent drawback of the FOA, i.e., suffering frompremature convergence, thispaper tries tohybridize theFOAwithQCMandthecat chaoticmapping function tosolve the 3