Short-Term Load Forecasting Based on Fuzzy Clustering Wavelet Decomposition and BP Neural Network

This paper proposes a composite method for short-term load forecasting, which is based on fuzzy clustering wavelet decomposition and BP neural network. Firstly, the similar-day´s load is selected as the input load based on the fuzzy clustering method; secondly, the wavelet method is applied to decompose the similar-day load into the low frequency and high frequency components, from which the feature of each load component can be captured. Finally, the separate neural network model is used to predict each load component, and the value of the forecasted load is obtained by superimposing the prediction value of each load component. The method proposed in this paper is tested on an actual power load in the year of 2010, and the results are compared with two other existing methods, which show that this method provides more accurate predictions.