Fine grid forecast is the main service of the China Meteorological Administration, and also the future development direction of weather forecast. This system improves the spatial resolution (0025°×0025°), and at the same time, meteorological elements such as precipitation and temperature forecast quality. This article described the technical methods in the data products of this system, from four aspects: (1) established the technical framework for grid forecast, using the Dynamic Cross Optimal Elements Forecast (DCOEF) method to establish the background field of grid forecast, which means comparing different model’s element forecast results and selecting that with higher forecast quality in past 15 days as the base field for forecasters; (2) proposed the method of “stationrevised value transmitting to the grid field” for consecutive elements correction. The cross test shows that the accurate rate of 24hour minimum and maximum temperature (<2 ℃) are improved by 34% and 23%, respectively, by this method compared to the model downscaling data, and also, the method has better application value in the combination of the background field collaborative and subjective station forecast and objective grid element forecasts; (3) based on the Bias Correction method to correct grid precipitation; the results show that through calculating forecast bias to decrease light rain frequency and increase rainstrom frequency, the 24hour TS (Threat Score) improved by 25% and 482%, respectively, compared to the original model. (4) proposed the reverse deviation data normalization algorithm to deal the inconsistent problem of the objective or subjective correction data in the time series, which does not change the elements forecast trends of original models, and at the same time, the elements are coordinated in time, so to solve the problem of time coordination of grid elements.