A Comparison of Machine Learning Models for Predicting Coal-Fired Power Plant Performance

Cofiring and biomass firing have become programs in coal-fired power plants towards net zero emissions. Cofiring with biomass and other solid waste fuel in a coal-fired power plant will undoubtedly impact the power plant's performance. This study evaluates machine learning algorithms for modeling thermal efficiency in a coal-fired power plant. The model is a starting point to explore the potential of machine learning (ML) for thermal efficiency prediction and reducing emissions toward net zero emission. The comparison of various ML algorithms, including Neural Networks (NN), Random Forest (RF), AdaBoost (AB), and Gradient Boosting (GB), in the context of combustion performance modeling has been examined. NN excels, achieving R2 scores> 0.951 in predicting thermal efficiency. Furthermore, a study of biomass cofiring can be undertaken as a feasible and environmentally friendly approach to mitigating emissions. The NN model for coal combustion can be utilized to predict the performance of biomass cofiring as a solution to enhance sustainability and reduce carbon emissions through the transfer learning process. Further validation with diverse data is needed for various coal-biomass compositions and operating conditions.