Xgboost-Based Prediction of Net Electrical Energy Output of Combined Cycle Power Plants and Model Interpretability Study

Abstract

Combined cycle power plants occupy an important position in modern power systems due to their energy efficient and environmentally friendly characteristics. Accurate prediction of electrical energy output faces significant challenges due to complex changes in energy demand and environmental conditions. Existing methods are difficult to effectively capture the nonlinear relationship between the input variables and the output, and the interpretability of the model is poor. In this study, a prediction model based on the XGBoost algorithm is constructed and optimized for the prediction of net electrical energy output of combined cycle power plants. In addition, this study also compares two models, Random Forest and Support Vector Machine, and shows through experimental results that XGBoost is better than the other two models in terms of prediction accuracy, robustness and computational efficiency. Specific experimental results show that the root mean square error (RMSE) of the XGBoost model is 0.17, the explainable variance value (EVS) is 0.97, the mean absolute error (MAE) is 0.13, the mean square error (MSE) is 0.03, and the goodness-of-fit (R²) is 0.97. XGBoost performs better in all these metrics compared to the random forest and support vector machine models. To solve the opacity problem in the model prediction process, the SHAP framework is introduced to perform global and local interpretability analysis of the model, which effectively reveals the effects of ambient temperature (AT), ambient pressure (AP), relative humidity (RH) and exhaust vacuum (V) on the power output. The XGBoost model combined with the SHAP analysis not only improves the accuracy of the prediction, but also enhances the model's transparency and practicality, which provides strong decision support for the operation optimization of the power plant.