Abstract: The accurate estimation of battery state of charge (SOC) is very important for the management of electric vehicle power battery, and electric vehicles often encounter the mutation of SOC data in actual operation. At the same time, there are some errors in the established battery models and noise models, which lead to the poor self adaptability and robustness of the traditional extended Kalman filter algorithm in the process of SOC estimation. To solve these problems, this paper proposes to use the adaptive fading extended Kalman filter algorithm (AFEKF) to estimate the SOC of lithium-ion battery. The fading factor is introduced to adaptively iterate the system noise covariance, so as to update the optimal Kalman gain in real time and reduce the influence of factors such as data burst and battery model error.It can be seen from the experimental comparison under complex working conditions, thatcompared with the standard EKF, the SOC estimation accuracy of AFEKF can be improved by about 0.78% under NEDC condition and 0.5% under variable current condition. At the same time, it can converge to the real value faster and more smoothly when the initial value of battery SOC is inaccurate, which shows that AFEKF algorithm has higher estimation accuracy and better robustness than EKF.

Key words: state of charge (SOC), parameter identification, adaptive fading extended Kalman filter (AFEKF), lithium ion battery, second order RC model