A Noise Covariance Regulated Robust Modified Adaptive Extended Kalman Filter for State of Charge Estimation of Lithium-Ion Battery

A battery management system needs a robust algorithm for online state-of-charge estimation of batteries in different dynamic systems. Due to the ease of implementation, model-based state-of-charge estimation using the extended Kalman filter is popularly used in battery management systems for online state-of-charge estimation. However, the accuracy of the extended Kalman filter depends on the appropriate initialization of noise covariance. In this paper, a robust modified adaptive extended Kalman filter (RMAEKF) is proposed that enhances the state-of-charge estimation accuracy by incorporating recursive adaptive correction rules for process and measurement noise covariance matrices. The adaptive rule considers the predicted terminal voltage error and the state prediction error in each time step to provide the necessary correction of measurement noise covariance and process noise covariance respectively. Further, to validate the state-of-charge estimation accuracy of the proposed RMAEKF, its performance indices for LA92, US06, and mixed drive cycles are obtained at different operating temperatures and compared with the performance indices of the extended Kalman filter and forgetting factor-based adaptive extended Kalman filter. Moreover, the robustness of the proposed RMAEKF is examined with different initial values of state-of-charge, noise covariance matrices, offset current and bias voltage. In addition to that, an experiment using the OPAL-RT real-time simulator is also performed to validate the proposed methodology for online SOC estimation. Concurrently, mean execution time and computational complexity analyses of the proposed RMAEKF are performed to check its applicability in real-time battery management system applications. From the result analysis, it is observed that the proposed RMAEKF shows better robustness and higher state-of-charge estimation accuracy than other compared algorithms under dynamic operating conditions.