Robust state of charge estimation for wide-temperature-range lithium-ion batteries: integration of H∞ observer and adaptive kalman filter

Abstract

The State of Charge (SOC) is required for the safe and stable operation of lithium-ion batteries in electric vehicles, and thus, a high-precision method for obtaining SOC by the Battery Management System (BMS) is needed. However, lithium-ion batteries have a strong non-linear characteristic over a wide temperature range of −20°C to 55°C, and both parameter drift and external interference occur; thus, traditional SOC estimation methods are not very robust and have a large estimation error. To address the above issues, a relatively stable State Observers using H-infinity and adaptive Kalman filters (AKF) is proposed. First, a temperature-coupled second-order RC equivalent circuit model is constructed, and then its parameters are obtained through Hybrid Pulse Power Characterization (HPPC) tests and the recursive least squares (RLS) method. H∞ observers and AKFs are used to address interference and model mismatch, respectively, and the noise covariance matrices are adaptively adjusted to optimize estimation accuracy. Finally, the two algorithms are combined in a weighted-fusion mode to use the strengths of both. According to the experiment results, the maximum SOC estimation error of the proposed method in the full-temperature range is less than 0.9%, and it has strong robustness against temperature fluctuations and external noise, thus ensuring the reliability of SOC estimation for a vehicle-mounted BMS. This study has certain theoretical and engineering application value in improving BMS performance at high and low temperatures.