Fiber-Optic Enabling Data-Driven State Estimation for Intelligent Battery Management

The development of digital twins and intelligent management for lithium-ion batteries urgently requires extensions beyond existing sensing dimensions. While artificial intelligence methods are widely adopted, their effectiveness is fundamentally constrained by the limited measurable parameters (voltage, current, and temperature) that cannot capture mechanical responses during lithium-ion intercalation. Here, we introduce a new sensing dimension through fiber Bragg grating (FBG)-based optical strain monitoring, providing insights into battery internal states correlated with graphite anode dynamics. By developing a spatiotemporal feature fusion algorithm, the FBG-derived strain characteristics are synergistically integrated with electrical signals to train a long short-term memory (LSTM) network, establishing a novel multi-physics state estimation framework. Experiments covering wide temperature ranges $\left(5^{\circ} \mathrm{C}\right.$ to $\left.40^{\circ} \mathrm{C}\right)$ under Federal Urban Driving Schedule (FUDS) and Dynamic Stress Test (DST) profiles demonstrate that the optical-electrical fusion approach reduces root mean square error by 44.38% and reduces mean absolute error by 43.26% compared to conventional electrical-signal-only LSTM models. This work not only validates the fundamental role of optical sensing in battery modeling but also established a paradigm for multi-physical-field digital twins, with potential applications extending to state-of-health estimation and other smart battery management scenarios.