Research on Steel Production Scheduling Optimization based on Deep Learning

Steel production scheduling represents a pivotal aspect of the steel manufacturing process, encompassing the strategic allocation of resources and the optimization of production processes. This directly impacts the efficiency and cost-effectiveness of the production process. However, conventional scheduling optimization techniques are ill-equipped to address the intricate nuances of the steel production environment and the deluge of real-time data. In this work, a method of steel production scheduling optimization based on deep reinforcement learning was proposed and subsequently innovated and optimized. The Deep Q Network (DQN) was employed as the fundamental model, with the objective of enhancing the stability and convergence speed of the model. This was achieved through the design of the state space and action space of production scheduling, as well as the incorporation of experience playback and target networks. With regard to model optimization, an adaptive adjustment mechanism of the reward function is proposed, which enables the model to balance the optimization of multiple objectives, such as production efficiency and energy consumption, with greater accuracy. Furthermore, the network structure has been enhanced and a multi-head attention mechanism has been incorporated to augment the model’s capacity for scheduling decisions in intricate production scenarios. The optimized model was subjected to experimental verification on an actual steel production dataset, and its performance in terms of scheduling efficiency and resource allocation accuracy was found to be excellent.