Research on short-term load forecasting model based on wavelet decomposition and neural network

This paper gives a method which bases on the wavelet decomposition and the neural network to predict the short-time load. Using wavelet transform, the load sequence is decomposed into sub-sequences on different scales, then using appropriate artificial neural network models the sub-sequences of forecasting date are predicted. Finally, by means of restructuring from the sub-sequences, the final forecasting results of the load sequence are obtained. The actual load data of electric network in Yichang, Hubei, China are applied to build the model. The instance shows that the proposed method is possessed of higher forecasting accuracy and better adaptability than back propagation (BP) neural network forecasting methods.