Robust State of Charge Estimation of a Lithium Iron Phosphate Battery Module Using a Dual-UKF Under Various Operational Temperatures

Abstract

This paper proposes a robust framework to estimate the state of charge (SOC) of a lithium iron phosphate (LFP) battery module operating under diverse temperature conditions. The proposed framework integrates an enhanced equivalent circuit model (EECM) with a dual unscented Kalman filter (dual-UKF) to address SOC inconsistency and temperature-dependent nonlinearity in LFP modules. Bayesian optimization (BO) is employed to identify both model parameters and filter hyperparameters systematically, ensuring high fidelity and adaptability without manual tuning. The dual-loop UKF architecture simultaneously estimates SOC and updates time-varying parameters, effectively compensating for initialization errors of up to 10%. The proposed framework maintains an SOC root-mean-square error (RMSE) below 1% across a wide temperature range from −15 to 45°C, with a faster convergence observed at low temperatures caused by enhanced observability from increased potential variations. Comprehensive experimental validation using standard dynamic driving cycles and real-world stochastic battery energy storage system (BESS) data confirms the accuracy, robustness, and applicability of the proposed framework. Thus, the proposed framework offers a practical solution for battery management systems (BMSs) in LFP-based battery applications, as it provides accurate monitoring capabilities under diverse environmental and operational uncertainties.