Abstract: In order to solve the problem that wind speed in wind farms is difficult to be predicted accurately and efficiently due to its randomness,this study proposes a wind speed prediction method for wind farms based on density peaks clustering. Utilizing this method,wind turbines are firstly clustered based on density peak algorithm. Subsequently,the long short-term memory network models are used to predict the wind speed of similar wind turbines. Considering the unequal importance of each indicator during actual clustering,the data are preprocessed based on weighting theory. At the same time,the dimension reduction of data is realized through the principal component analysis in order that a poor effect of density peak clustering can be avoided when facing high-dimensional data.Finally,the effectiveness of the proposed method is verified based on the measured data of the wind farm. Experimental results show that the method has relatively high predicting accuracy.