Page - 341 - in Short-Term Load Forecasting by Artificial Intelligent Technologies

Energies2018,11, 1282 2.4. IntroductionofFactorAnalysis-AntColonyClustering-BatAlgorithm-ExtremeLearningMachine (FA-ACC-BA-ELM)Model Since theELMhas lessability torespondtosamplesof the trainingset, itsgeneralizationability is insufficient. SoweproposeBA-ELM.In thispaper, theflowchartof the factoranalysis-similarday-bat algorithm-extremelearningmachine (FA-SD-BA-ELM)model is showninFigure3.Asdiscussedin part1,autocorrelationandthepartial correlationfunction(PACF)areexecutedtoanalyze the inner relationshipsbetweenthehistory loads. Basedonthe influencingfactorsof load, factoranalysis (FA) isusedforextracting inputvariables.Accordingto theresultof factorsanalysisandprevious load, theantcolonyclusteringalgorithm(ACC) isusedtofindhistoricaldays thathavecommonfactors similar to the forecastday. Part2 is thebatoptimizationalgorithm(BA)andpart3 is the forecastingof theextremelearningmachine (ELM). Start 22 influence factors Continuous load values auto correlation and partial autocorrelation Input variables The topological structure of ELM Initialize the weight and bias of ELM Optimum the weight and threshold Optimum matrix of the output weights £ Predict load values end Initialize the parameters of BA Terminal condition Update search pulse frequency and position of bats Calculate the fitness value of bats and hold the best postiion Calculate the fitness value of bats and hold the best position of bats NY Part 1rt Part 3rt Part 2rt Ant colony clustering Factor analysis Load values of the same time of each day Update the pulse frequency and the volume of BA Figure3. Theflowchartof the factoranalysis-ant colonyclustering-batalgorithm-extreme learning machine (FA-ACC-BA-ELM)model. 341