In order to reduce the bad effects of wind power when connect to the power grid, it is necessary to predict the wind power accurately. According to the wind power of the non-linear and non-stationary characteristics, a short-term wind power combination forecasting method based on EEMD and Coupling SAE-BP is proposed. In this method, EEMD is used to decompose the wind power series into a series of relatively stable components to reduce the interaction between the characteristic information. Then, the prediction model is established by using SAE to learn the high frequency components while BP predicts the low frequency components. Finally, the prediction results of the wind power are obtained by superimposing the sequence prediction results. The prediction process and the results show that the proposed model can improve the prediction accuracy and have high utilization value.