Battery Pack Cell Balancing using Topology Switching and Machine Learning

Battery packs are widely used in electric vehicles. Imbalance in the state of charge (SoC) of battery cells results in suboptimal operation of these packs due to their early stop of charging and discharging processes. This in turns leads to an increase in the number of charging cycles, which degrades the performance of the battery pack and results in shorter lifetime. To address cell imbalance, battery management systems (BMS) must employ cell balancing or equalization methods. In this paper, we propose a novel battery pack balancing technique, which uses a reconfigurable switching network to periodically change the pack topology in order to achieve cell balancing. The periodic reconfiguration is based on a machine learning (ML) algorithm that points the next topology that will get the pack into an operation mode in the next control period where cell imbalance is reduced. The proposed technique is verified with a custom simulation framework that integrates state-of-the-art extended Kalman filtering (EKF) models for state estimation for increased accuracy. Simulation results demonstrate that the proposed technique can successfully reduce battery pack cell imbalance.