Physics constrained state space modeling for battery state of ch 2027 Applie

Accurate estimation of the state of charge (SOC) is critical for the safe and efficient operation of lithium-ion batteries. However, SOC estimation based solely on external electrical signals struggles to capture internal battery states under dynamic loading and temperature variations, thereby limiting accuracy and robustness. To address this limitation, we propose a SOC estimation framework that integrates multiphysics sensing with electrochemistry-constrained sequence modeling. Specifically, an implantable fiber Bragg grating (FBG) sensing scheme is developed by combining an armored FBG with a bare FBG, enabling in situ decoupled measurement of internal thermo-mechanical states. Building on this sensing paradigm, a dual electrochemical-constrained Mamba framework (DEC-Mamba) is proposed. The framework employs an electrochemistry-constrained state-space model (SSM) as the sequential backbone, where electrochemical constraints are embedded into the state transition and update processes. In addition, physics-informed regularization during training and an uncertainty-aware adaptive calibration mechanism during inference are introduced to improve physical consistency and mitigate residual estimation bias under non-stationary operating conditions.