In order to solve the problem that current machine learning-based dam safety monitoring models cannot give the explanation of the model prediction, the Shapley additive explanations (SHAP) value theory was introduced, and combined with the light gradient boosting machine (LightGBM) model, an interpretable safety monitoring model for concrete gravity dam deformation was established. The model can quantify the specific contribution of each impact factor. The verification results of an engineering example show that the model considers the complex nonlinear relationship between deformation and the environmental quantities, which is closer to the real situation, and it not only has good fitting accuracy and prediction accuracy, but also can interpret the model globally and locally.