Page - 405 - in Short-Term Load Forecasting by Artificial Intelligent Technologies

Energies 2018,11, 242 • For theELM, therewere100neurons in thehidden layer,andthehardlimfunctionwaschosenas theactivationfunctionforconvertingtheoriginal features intoanotherspace. • For theSVR, thepenaltycoefficientwasset tobe80,andtheradialbasis functionwaschosenas thekernel functiontorealize thenonlinear transformationof input features. D E 7LPH RI DOO WKH VDPSOLQJ GD\V RQH XQLW PLQXWHV 7LPH RI WKH GD\ RQH XQLW PLQXWHV Figure 8. Periodicity knowledge and the residual time series of the retail store data set: (a) thedaily-periodicenergy-consumingpattern; (b) theresidual timeseries. 405