ABSTRACT Liquid metal batteries (LMBs), owing to their high capacity, dendrite‐free operation, and scalability, hold significant promise for large‐scale energy storage applications. For the industrialization of LMBs, accurate state of charge (SOC) estimation is crucial for their battery management systems. However, as an emerging electrochemical energy storage technology, LMBs lack extensive test datasets, and their flat voltage plateau complicates SOC estimation. To address these challenges, this study introduces a novel SOC estimation framework based on a self‐attention‐enhanced bidirectional gated recurrent unit (Self‐attention‐BiGRU) network. The model is validated using a 200 Ah LMB cell under diverse test profiles, including variable C‐rate cycles, HPPC, DST, and FUDS. Results show that the proposed method significantly outperforms baseline models, achieving a mean absolute error below 1.5% and demonstrating enhanced robustness across various operating conditions. This work provides an effective data‐driven solution for precise SOC estimation in LMBs.
Su et al. (Fri,) studied this question.