Abstract: Referring to raw data of the wind farm and based on the characteristics of wind speed,this article proposed a short-term wind speed prediction algorithm based on EMD-GRNN combination of probability and statistics．The paper firstly preprocessed the original data,then took advantage of empirical mode decomposition (E-MD) on the original wind speed,to get all the basic raw wind modal component (IMF).Nextly,this paper proceeded to reconstruct the data of each component to get the training samples and forecasting samples and utilized the train-ing sample to train GRNN network to make the network predictive capability.Finally this paper made a probability statistics on the output data of the various components．Simulation results show that the EMD-GRNN prediction model of combination of probability and statistics has good predictive accuracy and stability.

Key words: Short-term wind speed prediction,EMD,GRNN network,Probability statistics