State of charge estimation for lithium-ion battery based on an Intelligent Adaptive Extended Kalman Filter with improved noise estimator

Adaptive extended Kalman filter (AEKF) is widely used for lithium-ion battery (LIBs) state of charge (SOC) estimation. Innovation covariance matrix (ICM) of AEKF is estimated by fixed-length error innovation sequence (EIS) (the difference between measured and estimated voltages), which doesn't consider the distribution change of EIS. However, the distribution of EIS will change due to load current dynamics or error of battery model. Failing to consider the distribution change of EIS will lead to SOC estimation inaccuracy. To address this problem, this paper proposed an intelligent adaptive extended Kalman filter (IAEKF) method that can detect the moment of distribution change of EIS by the maximum likelihood function. Then, the ICM is updated based on the EIS after that moment to improve the SOC estimation accuracy. Results show that the proposed IAEKF method improves SOC estimation accuracy. Compared to that of the AEKF, the Root Mean Squared Error (RMSE) and the Mean Absolute Error (MAE) of SOC based on IAEKF decrease significantly by 43.34% and 55.80%, respectively, while the computation time only increases by 4.59%. In the end, the effect of initial parameters on the SOC estimation accuracy was analysed. It is found that the proposed IAEKF method is robust against parameter uncertainties. (C) 2020 Published by Elsevier Ltd.

Automated summary

This paper proposed an intelligent adaptive extended Kalman filter (IAEKF) method that can detect the moment of distribution change of EIS by the maximum likelihood function. The ICM is then updated based on the EIS after that moment to improve the SOC estimation accuracy, showing significant improvement compared to traditional AEKF.