In the seepage monitoring index, there are serious correlations between reservoir water levels, reservoir water levels and rainfall. The general multiple linear regression is used to establish the seepage monitoring model, and the multiple correlations between the monitoring indicators affect the parameter estimation, expand the model error, and destroy the robustness of the model. In order to overcome the interference of multiple correlations on the model, a partial least squares regression that can distinguish system information from noise is introduced, and a program is compiled. Example analysis shows that the influencing components separated by the partial least squares regression model can make reasonable physical causes for the changes of the measured variables of the dam, and the prediction ability of the partial least squares regression model is far better than that of the ordinary least squares regression model, and the sum of the squared prediction errors of the former is only about one-twentieth of the latter.