Seite - 37 - in Short-Term Load Forecasting by Artificial Intelligent Technologies

Energies2018,11, 1009 Figure9.Theseasonal tendencyofactualhourlyelectric load inExample2. Table6.The24seasonal indexes forSVRCCSandSVRCSmodels forExample2. Time Points Seasonal Index(SI) Time Points Seasonal Index(SI) Time Points Seasonal Index(SI) Time Points Seasonal Index(SI) SVRCCS SVRCS SVRCCS SVRCS SVRCCS SVRCS SVRCCS SVRCS 00:00 0.9718 0.9317 06:00 1.0545 1.1043 12:00 0.9848 0.9911 18:00 0.9753 1.0242 01:00 0.9848 0.9670 07:00 1.0383 1.1133 13:00 0.9896 0.9959 19:00 0.9707 0.9743 02:00 0.9894 0.9960 08:00 0.9854 1.0833 14:00 0.9898 0.9960 20:00 0.9711 0.9754 03:00 0.9937 1.0001 09:00 0.9913 1.0259 15:00 0.9994 1.0058 21:00 0.9610 0.9674 04:00 1.0076 1.0140 10:00 0.9860 0.9951 16:00 1.0144 1.0208 22:00 0.9519 0.9435 05:00 1.0343 1.0407 11:00 0.9841 0.9903 17:00 1.0252 1.0441 23:00 0.9567 0.9245 Theforecastingcomparisoncurvesofsixmodels inExample2, includingSARIMA(9,1,10)×(4,1,4), GRNN(σ= 0.07), SSVRCCS,SSVRCS,SVRCCS, andSVRCSmodels andactualvaluesare shown as inFigure 10. It indicates that theproposedSSVRCCSmodel is closer to theactual electric load values than theother comparedmodels. Similarly, the enlargedfigures, Figures 11–14, fromeight peaks inFigure10areprovided todemonstrate the tendencycapturingcapabilityof theproposed SSVRCCSmodelandhowcloser theSSVRCCSmodelmatches theactual electric loadvalues than other alternativemodels. It is clear that for eachpeak, the red real line (SSVRCCSmodel) always followscloselywith theblackreal line (actualelectric load),whetherclimbingupthepeakorclimbing downthehill. 37