The rainfall predicting values from T106 and HLAFS of NWC,BP neuron and dynamic similar method of Shanxi Meteorological Observatory were used for prediction factors.By using of grey theory,the original sequence of the predicting factors and actual rainfall sequence were conducted for accumulating process,respectively,then the random weakened increasing sequence were obtained.The recurrence calculation with Kalman filter method was used to acquire the rainfall predicting values.The values were retrieved to initial state,and the rainfall prediction for every county was given.The experiment in Summer in 1998 was conducted and obtained a good result in rainfall prediction.