Intelligent scheduling approach for steelmaking-continuous casting interface using on improved Q-learning

The Steelmaking-Continuous Casting (SCC) process is a critical stage in steel production, where scheduling directly affects energy consumption, production continuity, and product quality. To address the complex and dynamic scheduling environment of the SCC operations, this study proposes an intelligent scheduling model based on an improved Q-learning (IQL) algorithm. By integrating a dual-agent architecture and an experience replay mechanism, the model simultaneously optimises heat sequencing and ladle scheduling while satisfying production constraints and energy-saving objectives. Industrial validation demonstrates that the IQL reduces the overall SCC production cycle by 8.0% and decreases cumulative temperature drop by 45 °C during the EAF-LF transition and by 109 °C during the LF-CC transition. The IQL outperforms heuristic methods and standard Q-learning (QL), reducing total production time by 3.1% and 5.6%, respectively, and achieving higher equipment utilisation. Compared with manual scheduling, the IQL reduces electricity consumption by 39593.4kWh, resulting in an economic benefit of approximately 25735.7 CNY and a corresponding reduction in CO2 emissions of roughly 21.0 t per casting sequence. The proposed framework provides an adaptive real-time scheduling strategy for intelligent and low-carbon SCC production.