By selecting of the ECMWF grid fields and using of such methods as differential method, synoptic diagnosis, factor combination, etc, the predictor database that can reflect the local characteristics of synoptic dynamics was established. The “0, 1” linearization was conducted on the predictors and predictands, and the initial selection was based on the PRESS (Prediction Residual Error Sum of Squares) criterion. The 0〖KG-*3〗〖CD*2〗〖KG-*6〗120 hour precipitation prediction equation was established with the optimal subset regression method, and the precipitation probability was predicted with the multifactor probability weighting regression method. The forecast system has been put into operation, and the verification proved that the forecasting performance and the provided prefecturelevel guidance objective forecasting products are satisfactory. The forecast system can be connected with MICAPS to realize the highlevel automation and assure the objective and quantitative forecast products.