<![CDATA[Improving combustion efficiency and reducing flue gas emissions in coal-fired power plants (CFPPs) have become critical priorities amid growing global pressure to mitigate the environmental impact of the energy sector. Machine learning (ML) has demonstrated strong potential in predicting flue gas parameters, yet systematic reviews mapping algorithmic trends, implementation challenges, and integration opportunities with CFPP operations remain limited. This study presents a systematic literature review (SLR) of 31 selected articles published between 2019 and 2024 across Scopus, IEEE Xplore, and ScienceDirect databases, utilizing the PICOC framework for selection. The analysis shows that LSTM is the most frequently applied model for temporal prediction of flue gas temperature, while Random Forest is widely adopted for estimating NOx emissions. However, most studies are constrained to single-plant datasets, and real-time control system integration is still uncommon. These findings highlight the need for hybrid approaches that emphasize not only predictive accuracy, but also model interpretability via Explainable AI methods such as SHAP, and adaptability across diverse operational conditions. This study advocates future development directions by embedding predictive models within digital twin frameworks to enhance decision-making and optimize system performance sustainably. As such, the review contributes to bridging academic research with practical industrial demands in coal-based energy generation.]]>
Copyrights © 2026
