Page - 271 - in Short-Term Load Forecasting by Artificial Intelligent Technologies

Energies2018,11, 1605 Table1.Summaryof relatedstudiesonforecastingOCbetween2009and2017. Reference Method Type Duration Region Horizon [36] ANN Single 1965–2010 Turkey Long-term [37] MLP Single 1992–2004 Greek Long-term [46] FTS1,RTS2 Hybrid 1965–2012 MalaysiaandIndonesia Long-term [45] FTS,RTS Hybrid 1965–2012 Malaysia Long-term [40] ABCLM3 Hybrid 1981–2006 Jordan,Lebanon,Oman, andSaudiArabia Short-term [41] ABCNN 4,CSNN5, GANN6 Hybrid 1980–2006 MiddleEast region Short-term [22] GANN,ABCNN Hybrid 1980–2006 OPEC10 Short-term [34] GM7 Hybrid 1990–2002 China Short-term [44] ANFIS8 Hybrid 1974–2012 U.S. Short-term [42] GA,GNNM9 Hybrid 2000–2010 China Short-term SMLE* SVR,BPNN,LR Ensemble 1965–2016 GOC11 Long-term 1FuzzyTimeSeries; 2 RegressionTimeSeries; 3ArtificialBeeColonyAlgorithm; 4ArtificialBeeColonyNeural Network; 5 CuckooSearchNeuralNetwork; 6GeneticAlgorithmNeuralNetwork; 7GreyMarkov; 8Adaptive Neuro-FuzzyInferenceSystems; 9GeneticAlgorithm—GrayNeuralNetwork; 10Organizationof thePetroleum ExportingCountries; 11GlobalOilConsumption; *ProposedMethod. Figure1.StackingMulti-LearningEnsemble (SMLE)Framework. 2.2. EnsembleGeneration In the original data set, the initial training data, represented as D, hadm observations and n features, so that it ism×n. Themodelingprocedurecanberealizedbysettingdifferentparameters of thebase learners. In this level, someheterogeneousmodelswere trainedonDusingonemethod 271