The spatio-temporal statistical downscaling of precipitation has always been a difficult research point in intelligent grid forecasting. By comparing several spatio-temporal downscaling methods, we obtained the optimal scheme suitable for the Beijing Region in this study. In terms of spatial statistical downscaling, spatial distribution characteristics of five methods are compared, namely inverse distance weighting, ordinary Kriging, Nearest, Bilinear, 3-D ordinary Kriging method, and the results show that Bilinear interpolation method has the best application effect in Beijing Region, with the smallest error and highest ETS. For temporal downscaling, two kinds of allocation schemes based on regional numerical model are compared, including hourly allocation schemes (RMAPS and CMA-MESO) and average allocation. The results show that there are no significant differences in RMSE and MAE for the three methods. The hourly allocation by RMAPS is more preponderant than by observation in ETS score, and it performs also better in heavy rainfall cases, which means the allocation by RMAPS has better advantages from the perspective of forecast accuracy. The schemes of bilinear interpolation and RMAPS hourly allocation are taken as the spatio-temporal downscaling schemes in the objective forecasting technique of Beijing Meteorological Observatory, and can support intelligent grid forecasting for providing refined forecast products. The results can provide some references for forecasting and associated researches.