State-of-Charge Estimation of Lithium-Ion Batteries in Parallel for Electric Vehicles Using Improved Square Root Unscented Kalman Filter

An accurate state of charge (SOC) estimation is crucial for battery applications. Traditional model-based methods have difficulty addressing parameter mismatches and measurement errors, whereas the Kalman filter method faces challenges in handling non-Gaussian noise. In this paper, an improved square root unscented Kalman filter (SRUKF) is proposed for the SOC estimation of lithium-ion batteries (LIBs). The improved SRUKF can effectively handle colored noise, and it is challenging to accurately model it owing to the mismatch between the measurement model and the theoretical reference model in the process of SOC estimation. Using this method, colored noise and white noise are separated by the characteristic that the mean value of colored noise is different from that of white noise. The Kalman gain was modified according to the amplitude of the colored noise in the current and historical states, and the estimated value was closer to the real SOC than that of the conventional SRUKF. To better simulate the characteristics of large-capacity batteries in electric vehicles, parallel battery cells with different initial SOC values are configured for comparative simulations and experiments. Compared with the traditional SRUKF, the estimation error is reduced by 10% (pulsed-current condition) and 49% (World Transient Vehicle Cycle, WTVC) by the proposed method. The experimental results demonstrate that the improved SRUKF can overcome the influence of colored noise on the SOC estimation.