Interpretable machine learning model for reservoir outflow water temperature prediction

The construction of reservoirs changes the water temperature of rivers and further affects aquatic ecological environments. Ecological operation is an effective way to improve the outflow water temperature of reservoirs, and it uses accurate outflow water temperature prediction as a basis. Compared to numerical models, machine learning models have the advantages of high efficiency and nonlinear fitting; hence, they can be used to predict the outflow water temperature of reservoirs. However, most machine learning models from previous studies exhibit poor interpretability. In this study, we focused on the Pubugou Reservoir and developed an interpretable machine learning model to predict reservoir outflow temperature. By selecting input variables based on the physical mechanisms of the numerical model and using the Shapley Additive Explanations (SHAP) method, we identified the impact of feature variables on model outputs, thus enhancing the model's interpretability. In addition, we use validated numerical models to expand the water temperature dataset for machine learning training and optimize the model parameters to improve the model performance. The results indicate that Genetic Algorithm-optimized Support Vector Regression (GA-SVR) performs very well in predicting the outflow water temperature of reservoirs, with Mean Absolute Error (MAE) and Root Mean Square Error (RMSE) values of 0.106 °C and 0.136 °C, respectively, and a maximum prediction error of 0.41 °C. Using the SHAP method to explain the machine learning model revealed that the inflow water temperature is the main feature impacting the model output, and other operation conditions are secondary features. The research frame and results can provide a reference for reservoir ecological regulation and watershed ecological environment management.