A Novel Reinforcement Learning Balance Control Strategy for Electric Vehicle Energy Storage Battery Pack

Energy imbalance in electric vehicle energy storage battery packs poses a challenge due to design and usage variations. Traditional balancing control algorithms struggle to cope with large-scale battery data and complex nonlinear relationship modeling, which jeopardizes the stability of energy storage systems. To overcome this issue, we propose a reinforcement learning (RL)-based strategy for battery pack balancing control. Our approach begins with adaptive battery pack modeling followed by the employment of an active balancing control strategy to determine the duration of the balancing charge state and rank the balancing strength of individual battery pack cells. Subsequently, a RL network is employed to learn dynamic parameters that capture battery pack variations, enabling subsequent automatic learning and prediction of effective balancing strategies while simultaneously selecting the optimal control policy. Our simulation experiments demonstrate that our approach ensures an orderly charge and discharge process of battery pack cells, achieving an impressive balance efficiency of 91% when compared to other similar balancing control methods. Furthermore, the optimization of RL methods results in significant improvements in battery pack energy efficiency, stability, and operational costs. Notably, our method also outperforms other similar control methods in terms of energy utilization rates, establishing its superiority in this category.