Seite - 314 - in Short-Term Load Forecasting by Artificial Intelligent Technologies

Energies2018,11, 1561 factor. A largerwidthcoefficientmay leadtosatisfactoryCPs, andasmallerwidthcoefficientmay result inasatisfactory intervalwidth. Therefore, inpractice, thedecisionmakerneeds toadjust the widthcoefficient forspecificdemands. Forexample,wechose thewidthcoefficientwithaminimum intervalwidth at the same time that theminimumdemandofCPwasguaranteed. (e)Nomatter howcomplex is thedataset, theproposedmodelalwaysprovides thebestperformancecomparedto benchmarkmodels.However,becauseof thecomplexityof thedata itself, someof theperformance is not remarkable. Ingeneral, theproposedmodelprovidedadesiredresult inmostcases. Furthermore, inapowergridoperator theproposedmethodhasastrongpracticalapplication significance. Ahighly accurate forecastingmethod is oneof themost important approachesused in improvingpowersystemmanagement, especially in thepowermarket [58]. Inactualoperation, for secure power grid dispatching, a control center has tomake a prediction for the subsequent load.Accordingtohistoricaldata, thedataset for thepredictivemodel involvedcanbeconstructed. Theresultsof thepredictivemodelareable toprovide theupperboundandlowerboundof the load at somepoint in the future. Dependingon theupper boundand lower bound, the control center canadjust thequantityofelectricityoneachcharging line. Therefore, suchahybridapproachwhich canprovidemoreaccurateresultscanensure thesafeoperationof thepowergridandimprovethe economicefficiencyofpowergridoperation. Author Contributions: J.W. carried on the validation and visualization of experiment results; Y.G. carried onprogrammingandwritingof thewholemanuscript; X.C.provided theoverall guideof conceptualization andmethodology. Funding:This researchwasfundedbyNationalNaturalScienceFoundationofChina(Grantnumber: 71671029) andGansuscienceandtechnologyprogram“Studyonthe forecastingmethodsofveryshort-termwindspeeds” (Grantnumber: 1506RJZA187). Acknowledgments:ThisworkwassupportedbytheNationalNaturalScienceFoundationofChina(No.71671029) and theGansu science and technologyprogram“Studyon the forecastingmethodsof very short-termwind speeds” (No. 1506RJZA187). Conflictsof Interest:Theauthorsdeclarenoconflictsof interest. Abbreviation STLF Short-termloadforecasting PI Prediction intervals PIW Prediction intervalswidth PINAW PInormalizedaveragewidth ENN Elmanneuralnetwork SNR Signal tonoiseratio IMF Intrinsicmodefunction Nstd Noisestandarddeviation Pop_num Totalpopulationnumber Maxiter Themaximumnumberof iterations CEEMDAN Thecompleteensembleempiricalmodedecompositionwithadaptivenoise NN Neuralnetworks CP Coverageprobability LUBE Lowerupperboundestimation PINRW PInormalizedroot-mean-squarewidth Dim Individualparameterdimension EMD Empiricalmodedecomposition MSE Meansquareerror NR Numberof realizations RP Recurrenceplot MOSSA Multi-objectivesalpswarmalgorithm E-LUBE LowerupperboundestimationwithENN 314