Abstract: Due to existence of the randomness and intermittence of wind speed and the dispersion of wind speed and power of each unit,it is difficult to predict the wind power in a wind farm. On the basis of calculating the weights of the linear correlation of wind speed,a wind speed model based on the improved FCM（fuzzy c-means clustering）clustering algorithm is proposed in this paper,and the relationship function between the equivalent wind speed of the wind farm and the wind speed of the improved FCM is established. The measured data of a certain wind farm shows that the proposed method for wind power prediction is simple in algorithm and the prediction accuracy improved by 71.35%,and the validity and universality of the prediction method are verified under different daily cycles of the wind farm.