Seite - 393 - in Short-Term Load Forecasting by Artificial Intelligent Technologies

Energies 2018,11, 242 consumptiondata. In this study,firstly, thestructureof theproposedhybridmodelwillbepresented, andhowtoextract theenergy-consumingpatternwillbedemonstrated. Then, the trainingalgorithm for themodifiedDBNmodelwillbeprovided. The learningof theDBNmodelmainly includes two steps,whichfirstlyoptimizes thehiddenparametersby thecontrastivedivergence (CD)algorithm inapre-trainway, and thendetermines theoutputweightingvector by the least squaresmethod. Furthermore, theproposedhybridmodelwillbeapplied to thepredictionof theenergyconsumption in two kinds of buildings that have different energy-consuming patterns (daily-periodicity and weekly-periodicity).Additionally, toshowthesuperiorityof theproposedhybridmodel, comparisons with fourpopularartificial intelligencemethods—thebackwardpropagationneuralnetwork(BPNN), the generalizedradialbasis functionneuralnetwork(GRBFNN), theextremelearningmachine(ELM), andthesupportvectorregressor (SVR)willbemade. Fromthecomparisonresults,wecanobserve thatall thepredictors (DBN,BPNN,GRBFNN,ELMandSVR)designedusingboth theperiodicity knowledge and residual dataperformmuchbetter than thosedesignedonlyby theoriginal data. Hence,we can judge that the periodicity knowledge is quite useful for improving the prediction performance in thisapplication. Theexperimentsalsoshowthat, amongall thepredictionmodels, theproposedDBNbasedhybridmodelhas thebestperformance. The rest of this paper is as follows. In Section 2, the deep belief networkwill be reviewed. InSection3, theproposedhybridmodelwill bepresentedfirstly, and then themodifiedDBNwill beprovided. InSection4, twoenergyconsumptionpredictionexperiments forbuildings thathave different energy-consumingpatternswill bedone. In addition, the experimental and comparison resultswillbegiven. Finally, inSection5, theconclusionsof thispaperwillbedrawn. 2. IntroductionofDBN The DBN is a stack of restricted Boltzmann machine (RBM) [11,30]. Therefore, for better understanding,wewill introduce theRBMbefore the introductionof theDBNinthissection. 2.1. RestrictedBoltzmannMachine Thestructureofa typicalRBMmodel is showninFigure1. TheRBMisanundirected,bipartite graphicalmodel,whichconsistsof thevisible (input) layerandthehidden(output) layer. Thevisible layerandthehiddenlayerarerespectivelymadeupofnvisibleunitsandmhiddenunits, andthere isabias ineachunit.Moreover, thereareno interconnectionwithin thevisible layeror thehidden layer [31]. 9LVLEOH/D\HU Y Y Y QY K K PK+LGGHQ/D\HU K 9LVLEOH%LDV D +LGGHQ%LDV E E E PE D D QD :HLJKW0DWUL[ : Figure1.Thestructureofa typicalRBMmodel. 393