State of Charge Estimation for Lithium-Ion Battery Pack With Selected Representative Cells

Electric vehicles (EVs) are instrumental in driving the transition toward transportation electrification, achieving carbon peak targets, and striving for carbon neutrality. Within the EV ecosystem, battery packs serve as vital energy storage systems. However, existing research has primarily concentrated on modeling and estimating the state of individual battery cells, posing challenges when applying these models directly to battery packs due to their inherent complexity and the variability among cells within them. Consequently, limited efforts have been made to explore alternative models and methods to improve estimation accuracy while reducing complexity. Here, we propose a novel data-driven and filter-fused algorithm for estimating battery packs’ state of charge (SOC). First, representative cells are selected to minimize data redundancy and system complexity while accurately representing the pack’s state. Then, the long–short-term memory (LSTM) network is used to establish a mapping between SOC and electrical measurements from the pack. Finally, we integrate the extended Kalman filter to smooth the output, creating a closed-loop structure that enhances estimation accuracy. Experiment results demonstrate the efficacy of the proposed method in accurately estimating the SOC for battery packs. Furthermore, the method exhibits robustness and generalization ability, which indicates its potential for practical application in real-world scenarios.