Lithium Battery Electrode Coating Process Control Based on Deep Reinforcement Learning

The electrode coating process is critical in lithium-ion battery manufacturing, yet its inherent time-delay challenges traditional Proportional-Integral-Derivative (PID) control in balancing stability and dynamic response. This paper proposes a Deep Reinforcement Learning (DRL)-based control method by integrating Gated Recurrent Unit (GRU) and Twin Delayed Deep Deterministic Policy Gradient (TD3) to enhance long-term temporal dependency modeling. A reward mechanism combining boundary constraints that penalize out-of-limit states to ensure operational safety and dead zone constraints that suppress redundant actions within predefined tolerance bands optimizes precision-stability trade-offs. Simulations show the proposed algorithm reduces adjustment time by 51.3 and 24.0 percent compared to PID and fuzzy PID under ideal conditions. With Gaussian white noise disturbances, it exhibits enhanced robustness with a standard deviation of 0.952, lower than the conventional PID (1.232) and fuzzy PID (1.036) approaches. Moreover, it achieves a Process Capability Index (CPK) of 2.04, surpassing fuzzy PID’s 1.82 by 0.22, demonstrating its superior capability in enhancing coating process performance.