Online State-of-Charge Estimation of Lithium-Ion Battery Using a Fault Tolerant and Noise Immune Threefold Modified Adaptive Extended Kalman Filter

In electric vehicle, a reliable battery management system (BMS) enhances the performance by accurately estimating the state of charge (SOC) of battery. However, the SOC estimation accuracy degrades with change in operating temperature, improper initialization of process noise covariance matrix ( $Q$ ) and measurement noise covariance matrix ( $R$ ), and error in sensor measurement. In view of these aforesaid issues, a threefold modified adaptive extended Kalman filter (TMAEKF) is proposed in this work. The design of the proposed TMAEKF is of threefold. First, a three-segment design-based adaptive factor is implemented to update the posteriori state error covariance matrix to improve the robustness of SOC estimation under parametric uncertainty due to variation in operating temperature. Second, an amnestic factor and sliding window-based innovation covariance matrix (ICM) are incorporated in the design of $Q$ and $R$ to handle the effect of initialization error in SOC estimation. Third, an adaptive correction factor is designed by comparing the actual and theoretical ICM to update the priori state error covariance matrix (SECM) to eliminate the adverse effect of sensor measurement error in SOC estimation. The proposed TMAEKF is developed in MATLAB, and the performance is evaluated for LA92 and US06 drive cycles at 0 °C, 25 °C, and 40 °C temperature. To verify the efficacy of the proposed TMAEKF in SOC estimation, the root-mean-square error and maximum absolute error of TMAEKF are compared with other adaptive extended Kalman filter available in the literature. Moreover, robustness of the proposed TMAEKF is examined by applying offset current, bias voltage, and transient load current disturbance. To enhance the persuasiveness of the argument results, an OPAL-RT-based real-time experimental setup is developed and the aforesaid simulation results are validated. The test results confirm that the proposed TMAEKF estimates the SOC accurately and shows strong robustness under different dynamic operating conditions.