In February 2026, the TRANS Research Group published a study on real-world battery health management in the Journal of Energy Storage. To address the complexity of battery packs, this work constructs a thermoelectric dynamic graph attention network that captures internal cell inconsistencies and spatial heterogeneity. Furthermore, an interpretable temporal decomposition mechanism is introduced to disassemble degradation processes into traceable, physically meaningful components. Validated on large-scale real-world electric vehicle data, this method ensures high-precision estimation across different vehicles and provides a quantitative basis for analyzing user driving behaviors.
