Abstract: In order to raise the prediction precision and stability of winter wheat yield prediction, a prediction model built from the selected long-term nonlinear predictors is developed through the expanding correlation analysis between meteorological yield and meteorological parameters, and it performs well. The preliminary theoretical analysis is also conducted. The application of four years and the comparison with the other statistical methods indicate that the predicted results using the method are desirable and applicable to the yield prediction at the bottom-level weather stations.