Seite - 28 - in Short-Term Load Forecasting by Artificial Intelligent Technologies

Energies2018,11, 1009 movements, it isalsocapable tofindout theglobaloptimum, i.e., it couldensure that thesystemwill notbe trappedina localoptimum[41]. 2.2.3. ImplementationStepsofCCSAlgorithm The procedure of the hybridCCS algorithmwith an SVRmodel is illustrated as followings. Therelevantflowchart is showninFigure1. ,QLWLDO WKH ORFDWLRQV RI UDQGRP Q QHVWV IRU WKH WKUHH SDUDPHWHUV 0DS WKH WKUHH SDUDPHWHUV LQWR FKDRWLF YDULDEOHV (YDOXDWH WKH ¿WQHVV YDOXH WR ILQG RXW WKH EHVW QHVW SRVLWLRQ E\ 0$3( LQGH[ 1R <HV 8VLQJ JOREDO VHDUFK (T DQG /p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igure1.Chaoticcuckoosearchalgorithmflowchart. Step1: Initialization. The locations of random n nests for the three parameters of an SVR model as x(i)k,j = [ x(i)k,1,x (i) k,2, . . . ,x (i) k,n ]T , k=C, σ, ε; i represents the iterationnumber; j represents thenumberof 28