Voltage-Consistent SOC Trajectory Estimation and Concurrent Fault Decoupling of Lithium-Ion Batteries Based on Constrained Adaptive FFRLS-EKF

Reliable state-of-charge (SOC) estimation is essential for lithium-ion battery management, yet parameter drift, operating-profile variation, and sensor faults can compromise observer consistency. This study presents a reproducible constrained FFRLS-EKF framework in which online second-order RC parameter updates are subjected to resistance, capacitance, and time-constant feasibility constraints before being scheduled in the EKF. Estimator residuals and parameter variations are then reused for exploratory concurrent fault analysis. Because the dynamic driving-cycle datasets do not provide independently measured continuous reference SOC, SOC RMSE/MAE is not reported for DST, FUDS, UDDS, US06, or BJDST; Coulomb counting is treated only as a non-independent trajectory reference because it also contributes to the FFRLS regression target. A separate 21-checkpoint HPPC validation, with reference labels withheld from the estimator, yields SOC RMSE/MAE values of 2.24/1.76 percentage points for the constrained adaptive method, compared with 2.50/2.04 percentage points for the fixed EKF. A 270-run robustness study varies fault magnitude, onset time, voltage-noise level, and initial SOC. The results identify physical projection as the dominant stabilizing mechanism, with adaptive forgetting providing secondary transient-memory adjustment. An additional 243-run two-fault stress test shows that residual-sensitivity decoupling is not universally identifiable: exact-pair recovery degrades as noise increases and remains strongly dependent on the operating profile and fault pair. Accordingly, the concurrent fault module is presented as a transparent diagnostic baseline rather than a universally validated fault-isolation method.