Hybrid PSO-BP Neural Network Approach for Wind Power Forecasting

Abstract

Due to the intermittence and fluctuation of wind power generation, accurate forecasting of wind power is of great significance for the safe and stable operation of wind power integrated system. In the paper, a new wind power forecasting method combining particle swarm optimization (PSO) and back-propagation (BP) neural network is proposed. BP neural network structure was constructed according the number of input data, and PSO algorithm is used to get the optimal initial weights and biases of BP neural network, which can effectively overcome the shortcoming of BP neural network that it is easy to fall into local optimal solution, and increase the convergence speed of BP neural network. Considering the integrity of training data, the PSO-BP neural network was trained to use full year data in 2011. The original BP, PSO-BP, GA-BP and wavelet-BP neural network is applied for 6-h, 1-day, 3-day wind power forecasting in May and December of 2012, respectively. The compared results show that the mean absolute error(MAE) and the root mean square error(RMSE) of wind power forecasting based on PSO-BP neural network are clearly less than that based on original BP, GA-BP and wavelet-BP neural network.