Abstract: Reliable and accurate wind speed prediction is beneficial to maintain the safe operation of power system. In order to improve the prediction accuracy,a short-term wind speed forecasting model is proposed based on residual,variational mode decomposition（VMD）,extreme learning machine（ELM）and long short-term memory（LSTM）. Firstly,the VMD algorithm is used to decompose the wind speed sequence into several sub-sequences to reduce the complexity of the original data. Secondly,the ELM network is employed as the initial prediction engine to extract the features of each wind speed sub-sequence. Then,all the sub-sequences are reconstructed to obtain the preliminary prediction results. To further mine the unstable characteristics of the raw wind speed time series,the LSTM is utilized for modelling the residuals of the preliminary prediction results. Finally,the resulting prediction wind speeds are obtained by integrating the predicted residuals and the preliminary results. Experiments are carried out on a real wind farm dataset and the predicted results are compared with other models. Experimental results show that the proposed model can significantly improve the prediction performance.