Aging-Aware Adaptive Battery State Estimation for LiFePO₄ Modules: Full-Lifespan Validation and Kalman Filter Comparative Analysis

This paper presents an aging-aware adaptive SOC/SOH estimation framework for LiFePO4 battery modules, validated through 3,574 constant-current charge–discharge cycles at 1C rate on a 4S4P module, covering the full lifespan from 91% to 35% SOH. A third-order polynomial capacity fade model (R2 > 0.99) is first established from the measured degradation data to dynamically update the rated capacity Qₙ used in Coulomb counting. A second-order RC equivalent circuit model coupled with an extended Kalman filter (EKF) then provides voltage-based SOC correction.

Monte Carlo validation (50 trials across five aging states) demonstrates that the adaptive Coulomb counting maintains an RMSE below 2.5% throughout the entire lifespan, while the adaptive EKF achieves an RMSE of 3.2%–9.3% over the SOH range of 54%–91%. Furthermore, innovation sequence analysis reveals that the flat open-circuit voltage characteristic of LiFePO4 causes sigma-point degeneracy in the unscented Kalman filter (UKF), leading to estimation divergence, whereas the EKF exhibits superior numerical stability under this condition. The complete framework is implemented and verified on a Nuvoton M487 embedded platform.