A hybrid method for short-term electricity price forecasting based on BPNN and GSM-SVM

Accurate price forecasting is of great importance to the pricing of the market for all participants. This paper proposes a hybrid method of electricity price forecasting in a practical market. The results of Back Propagation Neural Network (BPNN) and Support Vector Machine (SVM) are combined in this hybrid method according to their different strengths. Wavelet transform (WT) is applied to pre-process the data and reconfigure the forecasts. To optimize the parameters of the model in SVM approach, the two-layer Grid Search Method (GSM) is adopted. The proposed method is applied and examined by the information and data of New South Wales (2012)in Australian National Electricity Market. Case study indicates that the hybrid method can better capture the patterns of price data and predict the electricity price more accurately than individual methods.