A hybrid forecasting method of electricity consumption based on trend extrapolation theory and LSSVM

By mining electricity data characteristics and its relationship with environmental variables, this paper defines the electricity month-season ratio (EMSR) based on trend extrapolation and proposes a hybrid prediction method combining LSSVM and EMSR. The actual electricity consumption data of A Province is used for the case study. The prediction error of three methods are compared which are BP neural network, Elman neural network and EMSR-LSSVM hybrid method. It shows that the proposed algorithm can significantly improve the prediction accuracy.