From Fiber-Optic Sensing to Mesoscale Modeling: A Spatial–Temporal Transformer-Enhanced Distributed Thermal Profiling Approach for Lithium-Ion Batteries

Accurate modelling and profiling of internal temperature distribution are critical to the thermal management and operational safety of Lithium-ion batteries. Due to the limited resolution in temperature distribution sensing, existing approaches generally rely on pre-calibrated models incorporating simplified assumptions of axial temperature uniformity or homogeneous heat generation, limiting their precision and reliability. This study proposed a novel spatial-temporal transformer (STT)-driven mesoscale heat generation-enhanced distributed thermal (MHGDT) model. Relying on the high-resolution spatial temperature information from emerging optical fiber technologies, the MHGDT model offers a detailed analysis of the non-uniform heat generation and dissipation distribution characteristics associated with operating conditions at a fine scale. On this basis, by actively integrating the STT network, the nonlinear spatiotemporal dependencies between historical operating conditions and heat generation distribution characteristics are concurrently captured, enabling online prediction and updating of key parameters in the MHGDT model as conditions vary. Combined with an Unscented Kalman Filter-based observer, experimental validation demonstrates that the proposed method achieves a 55% improvement in accuracy compared to the pure model-based approach, with a maximum absolute error of less than 0.33°C and error fluctuations under 15% under extreme ambient temperature. These results highlight significant improvements in both the precision and reliability of the proposed method in real-world battery applications.