Abstract: Wind speed has characteristics of non-linearity,non-stationarity,intermittent and randomness. In order to improve the accuracy of short-term wind speed prediction,this paper proposes a combination prediction model based on PAM clustering,singular spectrum decomposition（SSD）and LSTM neural network to predict short-term wind speed to solve the problems mentioned above.Firstly,PAM algorithm is used to perform similar daily clustering on the original wind speed data to improve the learning efficiency of the neural network. Secondly,with the advantage of suppressing model aliasing and false components,SSD is used to decompose the wind speed series and extract multi-scale regular pattern. Finally,with a strong ability to capture the fluctuation law of long-term dependent series,LSTM neural network is used to predict the decomposed wind speed components. The prediction results of each component are accumulated to obtain the wind speed prediction results. Experiment results show that the combined prediction model based on PAM-SSD-LSTM can effectively improve the accuracy of short-term wind speed prediction.