Advanced state-of-health estimation integrating partial capacity-based initialization and resistance-informed covariance correction in adaptive extended Kalman filter for lithium iron phosphate batteries

Accurate state-of-health (SOH) estimation is a cornerstone of battery management systems (BMSs) for lithium iron phosphate (LiFePO4; LFP) batteries, yet remains challenging because of the flat open-circuit voltage (OCV) profile of LFP cells and the sensitivity of Kalman filter-based estimators to inaccurate initial states and error covariance settings. This study proposes an advanced SOH estimation framework that couples three mutually reinforcing components with an adaptive extended Kalman filter (AEKF) core: (i) partial capacity-based SOH initialization exploiting the high-sensitivity 3.56–3.58 V voltage region of LFP cells, (ii) resistance-informed dynamic correction of the initial error covariance matrix driven by the deviation between RLS- and DEKF-estimated internal resistance, and (iii) online equivalent circuit model (ECM) parameter identification via multiple adaptive forgetting factor recursive least squares (MAFF-RLS), continuously feeding the AEKF. Together, these components address the high initialization sensitivity and slow convergence of conventional AEKF methods, with β = 7 and α = 1 × 106 identified as optimal tuning values. Validation on 116.5 Ah LFP pouch cells under multiple degradation levels, temperatures, and driving profiles shows that the proposed framework reduces the SOH mean absolute error (MAE) by over 20% compared with a fixed-covariance DEKF baseline, while accelerating early-cycle convergence and maintaining robustness against initialization errors.