Wind speed forecasting is important for the security of wind power integration. Based on the theories of wavelet, wavelet packet, time series analysis and artificial neural networks, three hybrid models [Wavelet Packet-BEGS, Wavelet Packet-ARIMA-BFGS and Wavelet-BEGS] are proposed to predict the wind speed. The presented models are compared with some other classical wind speed forecasting methods including Neuro-Fuzzy, ANFIS (Adaptive Neuro-Fuzzy Inference Systems), Wavelet Packet-RBF (Radial Basis Function) and PM (Persistent Model). The results of three experimental cases show that: (1) the proposed three hybrid models have satisfactory performance in the wind speed predictions, and (2) the Wavelet Packet-ANN model is the best among them. (C) 2013 Elsevier Ltd. All rights reserved.