An Intelligent Control-Based Charging Strategy for Lithium-Ion Batteries

This paper proposes an intelligent framework for the efficient fast charging of Li-ion batteries while strictly following physical constraints and thermal safety. Using an electrothermal coupled model, a MIMO PID observer is first designed to estimate the battery state of charge and core temperature accurately. These insights along with the measurable battery variables are then provided to a deep reinforcement learning (DRL) framework to learn an internal dynamics-aware multi-constrained charging strategy. Besides priority experience replay, a dynamic reward function is formulated to aid efficient learning and convergence. The effectiveness of the proposed DRLguided strategy is demonstrated experimentally using Tesla 21700 battery cells. Results show that the proposed strategy can charge 68 % faster than the recommended 0.3 C CCCV and even surpasses extreme 3C CCCV (which violates critical temperature limits) while perfectly following all physical constraints, highlighting its potential for practical applications.