Bypass Equalization Strategy of Li-Ion Battery Based on LSTM Prediction of State of Charge

The inconsistency caused by battery aging significantly reduces the available capacity of battery packs, highlighting the necessity of battery equalization systems to maintain consistency. However, conventional bypass equalization strategies only activate the equalization process after the state of charge (SOC) difference exceeds a preset threshold, leading to delayed regulation and limited dynamic performance. To overcome this limitation, this paper proposes an optimized bypass equalization strategy for lithium-ion batteries based on SOC prediction using a long short-term memory (LSTM) network. First, the mechanism of SOC inconsistency is analyzed, and a firstorder equivalent circuit model is adopted to model the battery, incorporating aging effects to construct a heterogeneous battery pack model. Then, a bypass equalization strategy is designed, leveraging LSTM to predict the SOC evolution of the battery pack and improve equalization control. Finally, simulations and result analysis are conducted in MATLAB/Simulink. The results demonstrate that the proposed strategy can trigger bypass equalization proactively, maintaining SOC differences within the threshold and significantly enhancing the consistency of the battery pack.