Uncertainty of wind power will cause the impact to power grid. Wind power real-time forecasting can ease the power grid and ensure the stability of the grid. Based on the analysis of the characteristics of two kinds of single forecasting models, proposes a combined forecasting model based on historical data and NWP data. The sequence feature of the historical data and the applicable forecasting method are analyzed. The decision tree classification model is established. The best forecasting method is selected by real-time data sequence feature analysis. The results show that the combined model can forecast the accuracy higher than that of the single model. Real - time sequence feature analysis and matching the best forecasting model can improve the prediction accuracy by the decision tree model.