Prediction of emissions from gas turbine power generation on GWO-XGBoost-Sobol

To mitigate the environmental pollution caused by gas turbine power generation and control its nitrogen oxide (NOx) emissions to meet the standards, this study proposes a NOx emission prediction model based on the gray wolf optimization (GWO) algorithm and optimized extreme gradient boosting (XGBoost) by utilizing the operational data of gas turbines. To assess the performance of the XGBoost model and evaluate whether the GWO algorithm significantly improves the XGBoost model, we use five indicators, namely, root-mean-square error (RMSE), mean squared error (MSE), mean absolute error, mean absolute percentage error, and the coefficient of determination (R2), to compare the unoptimized and hybrid optimization models. Experimental results reveal that the XGBoost model exhibits the highest RMSE, MSE, and R2 of 0.278, 0.077, and 0.928, respectively. Five indicators of GWO-XGBoost surpass those of other hybrid optimization models. Finally, Sobol global sensitivity analysis is performed to determine first-order and total sensitivity. The analysis explores the effects of variables on NOx emissions, offering valuable insights for controlling emissions to meet standards.