Abstract: In view of excavating the non-stationary and nonlinearity of wind power，a hybrid model based on ensemble empirical mode decomposition (EEMD) and Volterra neural networks(VNN) is introduced into forecasting short-time wind power. Firstly, the end issue of EEMD is dealt with by using the largest Lyapunov prediction method. Secondly, the new gained wind power time series is decomposed into a series of sequences of different time scale by ensemble empirical mode decomposition to reduce its non-stationary. Then Volterra neural networks of each component is established on the basis of important parameters including embedded dimensions ,delay time and maximum Lyapunov exponent ,after mining the sequences chaotic characteristics by means of phase space reconstructed. Finally, the predicting results of each subsequence are superimposed to gain the final estimating result. Calculation example results shows that the proposed model is able to excavate power time series features effectively and obtain higher prediction accuracy.