A Comprehensive Investigation of Coal Blending Optimization: Methods, Applications, Challenges and Future Prospects

Coal blending optimization has become an important research topic, because it is able to improve fuel utilization, reduce operating costs and control pollutant emissions in both power generation and metallurgical production. This review aims to summarize representative coal blending optimization methods and clarify their applications in different industrial scenarios. To achieve this, the paper examines both traditional mathematical modeling approaches and recent Artificial Intelligence (AI)-hybrid methods, including MOPSO-based optimization, dynamic decision-making models, Support Vector Machine (SVM)-based prediction, XGBoost-SVR, hybrid residual prediction and reinforcement learning-enhanced evolutionary algorithms. Based on the comparison of these studies, this review discusses how the coal blending optimization gradually developed from static cost-oriented models to data-driven and multi-objective frameworks. This review also points out three main challenges faced by current studies which are AI interpretability, applicability and real-world complexity. It further outlines future prospects such as LLM-assisted explanation, transfer learning, domain adaptation and real-time optimization.