Solar irradiance is closely related to the power generation of photovoltaic power plants, and the accuracy of irradiance forecast directly affects the accuracy of power generation forecast. Based on the reality of solar irradiance of the photovoltaic power plants and weather stations, WRF model irradiance forecast, the EC fine grid numerical forecast, and theoretical solar irradiance, a study of the revised solar irradiance forecast is carried out by means of stepwise regression. The conclusions are as follows: (1) the correlation between irradiance ratio and meteorological elements is better than that between irradiance and meteorological elements; (2) there are differences in meteorological factors that affect solar irradiance at different times, and a prediction model of solar irradiance at different times is established through the stepwise regression method; (3) in the case of nonclear weather, the RMSE of the regressed prediction irradiance is about 10% lower than that of the WRF model. The results have some application value in the new energy numerical forecasting service for photovoltaic power stations.