Abstract: Wind power prediction usually concludes determined and uncertainly prediction. The former puts important on prediction accuracy and the later focus on the the risk of prediction results. For increase the prediction accuracy, an improved differential evolution algorithm was applied to widely search the optimal solution of wavelet neural network in different directions. By calculating joint conditional probability of wind power predictions about prediction error and power fluctuation, the risk of prediction results could be more fully assessed. The effectiveness of the proposed method was verified by simulation experiments.

Key words: wind power prediction, improved differential evolution algorithm, wavelet neural network, joint conditional probability