Extended and Unscented Kalman Filter Based Estimation of State-of-Charge of Li-Ion Batteries Using 3-RC Equivalent Circuit Model

The growing presence of electric vehicles (EVs) in the market has spurred investigations into various algorithms for estimating the state-of-charge (SoC) of the battery. Accurately estimating the SoC is of paramount importance for the battery management systems. This paper conducts an exhaustive literature review on diverse methods for SoC estimation, recognizing the critical importance of an accurate battery model in enhancing prediction efficiency. The study introduces and evaluates three methods, a) Coulomb Counting, b) Extended Kalman Filter and c) Unscented Kalman Filter for SoC estimation. Through simulation, the paper presents and analyzes the results derived from these methods. To simulate a Li-ion battery model, a 3-RC battery model was developed utilizing the Equivalent Circuit Model method. SoC estimation is conducted on a Li-ion battery with a capacity of 31A and voltage of 10.8V. In the simulations, a constant current charging method at 1C rate was employed for charging, while discharging was executed at various rates depending on the SoC of the battery, utilizing a random pulse generator and noise to replicate real-world scenarios. The UKF outperforms the EKF due to its ability to effectively handle highly nonlinear battery dynamics, thus yielding more accurate SoC estimations over time.