An integrated method for state of charge estimation, lifetime prediction, and reliability assessment of Lithium-ion batteries under thermal and dynamic conditions

Abstract

Accurate estimation of the state-of-charge (SOC), remaining useful life (RUL), and reliability of lithium-ion batteries is essential for renewable energy storage and electric mobility systems. This paper proposes a unified experimental–analytical framework that systematically integrates temperature-dependent SOC estimation, degradation modeling, and probabilistic reliability assessment within a single validated pipeline. Two A123 LiFePO 4 pouch cells were experimentally characterized using low-current open-circuit voltage (OCV) protocols and representative dynamic driving cycles (DST, US06, and FUDS) across multiple temperatures. A dual-estimator structure combining Coulomb counting, OCV correction, and an extended Kalman filter (EKF) was developed to enhance both steady-state accuracy and transient responsiveness under thermal variations. In parallel, temperature-aware degradation kinetics were modeled using Arrhenius-based relationships directly linked to cycle-based RUL extrapolation. To explicitly account for inter-cell variability, Weibull survival analysis was incorporated into the same computational framework, enabling probabilistic life prediction rather than purely deterministic estimation. Sensitivity analysis further quantified the propagation of parameter uncertainty into SOC and RUL predictions. Experimental results demonstrate voltage estimation errors below 0.02 V (corresponding to approximately 1–2% SOC deviation under nominal conditions), clear temperature-driven acceleration of aging, and significant life divergence between nominally identical cells (≈1000 vs. 200 cycles). The primary innovation of this work lies not merely in incremental accuracy improvement, but in the coherent integration of estimation, degradation, and reliability modeling under realistic multi-temperature dynamic operation, providing a practical decision-support architecture for real-world battery management systems.