An Adaptive Robust Square Root Unscented Kalman Filter for State of Charge Estimation for Lithium-ion Batteries

Abstract

Accurate estimation of the State of Charge (SOC) is essential for effective Battery Management Systems (BMS), as it directly affects the performance, safety, and lifespan of battery-powered devices. However, precise SOC estimation remains challenging due to the nonlinear dynamics of batteries and unpredictable process and measurement noise. This paper proposes a novel hybrid approach that combines an Adaptive Robust Square Root Unscented Kalman Filter (ARSRUKF) with Recursive Least Squares (RLS) to improve the accuracy and robustness of SOC estimation. The ARSRUKF enhances numerical stability by maintaining the semi-positive definiteness of covariance matrices and enables direct computation of their square roots, which improves both symmetry and computational efficiency. Unlike conventional filters such as the Extended Kalman Filter (EKF) and Unscented Kalman Filter (UKF), the ARSRUKF does not require prior knowledge of noise statistics, making it suitable for real-world applications with unknown or non-Gaussian noise. To further enhance filter performance, an Adaptive Neuro-Fuzzy Inference System (ANFIS) is integrated for real-time adjustment of noise covariances, allowing dynamic tuning of filter parameters based on system behavior. Experimental results under diverse operating conditions, including temperature fluctuations and varying noise levels, demonstrate that the proposed method consistently outperforms EKF and UKF. The presented framework offers a robust and accurate solution for SOC estimation, with broad applicability in electric vehicles, renewable energy storage systems, and portable electronic devices.