Research on Multi-objective Optimization Algorithm for Coal Blending

Coal blending optimization presents a complex and challenging task, as it encompasses multiple competing objectives, such as quality, cost, and environmental considerations. This study specifically targets coal blending for coking applications and offers a comprehensive approach to tackle these challenges. We employ Miceforest to accurately fill in missing data and introduce an enhanced lightweight Transformer network architecture for constructing a coke index prediction model. Based on the predicted outcomes, we develop a multi-objective and multi-constraint coal blending optimization model, allowing us to calculate the maximum allocation for individual coals and minimize the algorithm’s search space. Furthermore, we suggest a universal repair method that corrects unreasonable solutions while preserving the algorithm’s evolutionary trend throughout the iterative process. To assess the efficacy of our approach, we integrate these technological advancements with traditional multi-objective algorithms. Experimental results reveal that these modified algorithms yield superior performance in generating more efficient and effective coal blending solutions for coking plants, as compared to their original counterparts.