Based on discharge, water level, and precipitation data at the Guajingkou Station and water level data at surrounding stations from 1966 to 2014(with data missing from 1989 to 2005), a model for estimation of discharge in Taihu Lake was established with multiple linear regression. The annual sequence matrixes of the impact factors of the discharge were classified using clustering analysis according to the similarity, the years similar to the objective years were determined, and, finally, the regression model parameters were calibrated based on the data from similar years. Compared with the regression model based on the common series data, the regression model based on the data from similar years has a higher accuracy.