Abstract: In order to solve the problems that the traditional clustering algorithm is seldom used to mine and exploit the wind power sequence trend characteristics, and the performance of adjustable design for different wind farms is insufficient, a wind power prediction method based on novel multi-dimensional power trend clustering is proposed. Firstly, a novel measurement method for multi-dimensional power trend similarity distance is proposed, including wind power sequence fluctuation degree, fluctuation time, and the numerical dimension of fluctuation, which can mine the trend characteristics of wind power data. Secondly, the proposed strict coefficient is used to adjust the degree of participation in each dimension to adapt to the better clustering effect of different wind farm data. Finally, the proposed multi-dimensional power trend distance measurement is combined with the traditional fuzzy C-means soft clustering algorithm and Elman neural network group to build a complete prediction model. The results of the study show that the trend characteristics of power sequence is effectively mined and the prediction accuracy of wind power is improved by the proposed method.