In order to enhance the significance of numerical weather prediction (NWP) for short-term wind power prediction and take into account the influence of transitional weather processes on power prediction, a short-term wind power prediction method considering identification and testing of transitional weather processes is proposed. The samples with NWP interval 15 min of the time series are identified using a gated recurrent unit (GRU)based classifier for transitional weather processes. Based on the identification results, the wind speed series of transitional weather processes are tested by method for object-based on diagnostic evaluation (MODE), and the NWP forecasting regularity is explored. Based on the results of weather process identification for the time period to be predicted, matching weather processes, different models are selected for short-term wind power prediction. The proposed method is applied to a wind farm in Jilin, China, for arithmetic validation. The results show that the transitional weather process identification method has a high identification accuracy. The average reductions of RMSE value by 2.77% and MAE value by 2.46% for all types of weather process conditions prove the effectiveness of the method.