Abstract: As affected by the movement of atmospheric fluid,short term wind speed is characterized by volatility,nonlinearity etc,which severely influences the wind speed’s prediction precision. Aiming at this problem,the wind speed prediction method of optimizing the EMD decomposition process with Lorenz equation is proposed to solve the problem. Since Lorenz equation can be used to characterize chaotic effect caused by atmospheric fluid on wind speed,the one-dimensional Lorenz fluctuation sequence is adopted to optimize EMD decomposition process of wind speed,improve the modal aliasing phenomenon in the direct EMD decomposition process of wind speed and weakens chaos degree of high frequency components by EMD decomposition of wind speed. To confirm universality of the method,the short-term wind speed data of three kinds of time scales are analyzed and researched. The research results show:compared to EMD-BP neural network model,MAPE can be reduced about 5%, RMSE can be reduced about 0.1 and the qualified ration of wind speed prediction results is effectively enhanced by using the optimized EMD-BP neural network by Lorenz equation.