This paper addresses a problem of flood forecasting with the self-memory function. Considering flood forecasting’s uncertainty and updating demand, a hybrid hydrological model, namely Differential Hy- drological Grey Model with self-memory function (DHGM-SM), is developed. The model has two fold features. One is to establish a self-memorization equation linked with DHGM, that could extract useful information from past data series and realize updating of hydrological dynamic process. The other is that this model has higher efficiency relative to original hydrological model without self-memory func- tion. This approach was applied to river flow forecasting of two representative basins in Tunxi of South China and Daqinggou of North China. It is shown that this hybrid method has satisfactory forecasting accuracy by examination of both calibration and validation. Considering the flood forecasting’s uncertainty and updating demand, a hybrid hydrological model, ie, Differential Hy- drological Gray Model with self-memory function (DHGM-SM), is developed. model has two fold features. One is to establish a self-memorization equation linked with DHGM, that could extract useful information from past data series and realize updating of hydrological dynamic process. The other is that that model has higher efficiency relative to original hydrological model This approach was applied to river flow forecasting of two representative basins in Tunxi of South China and Daqinggou of North China. It is shown that this hybrid method has satisfactory forecasting accuracy by examination of both calibration and validation.