Deep learning for intelligent energy optimization and low-carbon transformation in metallurgical systems: A review

The metallurgical industry is one of the largest industrial energy consumers and a major contributor to global carbon emissions. Therefore, improving energy efficiency while reducing emissions has become a key path towards sustainable industrial development and carbon neutrality. However, complex physicochemical interactions, strong process coupling, and dynamic operating conditions continue to limit the effectiveness of traditional optimization methods. This review systematically synthesizes the applications of deep learning in energy optimization, emissions control, and intelligent process management across the metallurgical production chain, including ore pretreatment, smelting, refining, and casting. Existing studies demonstrate the potential of deep learning to improve energy efficiency, support emissions monitoring and control, and enhance real-time process optimization, although the magnitude and evidence of these benefits vary across processes and validation conditions. Despite rapid progress, large-scale industrial deployment is still limited by data quality, model generalization, and industrial adaptability. The system framework developed in this review highlights the potential of deep learning in supporting energy-efficient production, emission reduction, and intelligent process management across the metallurgical system, while also providing scalable insights into the sustainable transformation of other energy-intensive industries.