A novel approach for electricity demand forecasting

Abstract

In this paper, a novel approach, WPLSSVM, has been proposed for electricity demand forecasting, which combines particle swarm optimization (PSO), least squares support vector machine (LSSVM), and wavelet transform (WT). Firstly, the wavelet transform method is used to decompose the original sequence in WPLSSVM. Secondly, the WPLSSVM models the series using LSSVM, in which the parameters have been optimized by particle swarm optimization. Lastly, WPLSSVM obtains the final prediction by wavelet reconstruction. To test the model, the half-hour electricity demand series of New South Wales (NSW) in Australia has been used. The results demonstrate the validity of the approach.