Intelligent Control of the Dry Slag Discharge System in Power Plants Based on the Random Forest Algorithm

The random forest algorithm was applied to the dry slag discharge control system in power plants to enable intelligent control of thermal slag transportation, discharge, and damper adjustment devices in boiler systems. Based on the theory of cooling air volume calculation, the relationships among steel belt speed, boiler slag discharge rate, and cooling air volume were analyzed. The random forest algorithm was trained using actual operation data from the power plant, resulting in the development of a regression formula and proportional coefficient. Furthermore, model corrections were conducted to refine the relationships between power generation load and steel belt speed, steel belt speed and slag temperature, as well as cooling air temperature and slag temperature. A comparison between the corrected and original models was performed. The experimental results demonstrate that the introduction of the Random Forest algorithm significantly improves the prediction accuracy of cooling air volume and slag temperature, while substantially reducing the fitting error. According to the results presented in this study, under Leave-One-Out Cross-Validation (LOO-CV), the Random Forest model reduces the RMSE of the strip velocity–unit load relationship from 24.88 MW to 14.43 MW, and simultaneously increases the R² value from 0.896 to 0.965.Experimental results indicate that the data-driven control system, supported by the integrated random forest algorithm, significantly improves the accuracy of cooling air volume and slag temperature control, providing an innovative solution for the intelligent adjustment of slag discharge equipment in power plants.