Short-term forecasting for wind speed based on wavelet decomposition and LMBP neural network

Abstract

In this paper, a wind speed forecasting method based on wavelet decomposition and BP neural network with Levenberg–Marquardt algorithm (LMBP) is proposed. Firstly, original wind speed seires is decomposed into one low-frequency component and several high-frequency components by wavelet decomposition method. Then different LMBP neural networks are built for the forecasting of every component respectively. Finally the predictions of components are reconstructed to obtain the prediction of original wind speed. As for the problem that the convergence rate is limited by the inversion calculation of large-scale matrix in the training process of LMBP network, super-memory gradient algorithm solving large linear equations is introduced to adjust weights and thresholds of the network. Meanwhile the structure of hidden layer neurons is optimized by least squares network pruning method. At the end of the paper, actual wind speed data from certain wind farm is used to verify the forecasting model and the results indicate that the model develops the precision of wind speed forecasting effectively.