Abstract: The paper presents a crop-weather model for yield prediction on provincial level and higher. First, North China area is divided into a number of subregions by the R-type cluster analysis of climatic yield. Based on different effective of forecast period (EFP),correlations between climate yield and weather factor for each area are computed to sieve out forecast factors. Finally, forecast equations are established using the method of nonlinear multi-regression of accumulated and factorial type. The model may be applied to different crops, the EFP can be adopted according to characteristics of crops development stage or preduction needs.The EFP of winter wheat yield forecast in Norther China is divided into three phases: before wintering, after wintering and before ripeness. The multiple correlation coefficients are larger than 0.70,0.90 and 0.95 for the forecasts of the three different EFP respectively. The correlation error is below 1500,8% and 4% respectively.