Co-estimation of model parameters and state-of-charge for lithium-ion batteries with recursive restricted total least squares and unscented Kalman filter

• Published in: Applied energy Vol. 277; p. 115494
• Main Authors: Zhu, Rui, Duan, Bin, Zhang, Junming, Zhang, Qi, Zhang, Chenghui
• Format: Journal Article
• Language: English
• Published: Elsevier Ltd 01.11.2020
• Subjects: Kalman filter; Lithium-ion battery; Parameter identification; Restricted total least squares; State-of-charge; State-of-charge; Lithium-ion battery; Parameter identification; Restricted total least squares; Kalman filter
• ISSN: 0306-2619; 1872-9118
• DOI: 10.1016/j.apenergy.2020.115494

[Display omitted] •The biases occur to RLS even only in noisy voltage or current measurement case.•The model parameters are estimated by RRTLS to attenuate the noise effects.•Noise statistics are online estimated by PKS based noise covariance estimator.•The accuracy of SOC estimation is improved by the accurate model identification.•The superiority of the proposed method is verified by comparing with two methods. Due to the flawed sensor and the harsh electromagnetic interference in the electric vehicle, the measured current and voltage data can be seriously corrupted by noises, which poses a great challenge to the model-based state-of-charge estimation method. Through theoretical analysis and simulation experiments, this paper indicates that the conventional recursive least squares method can suffer from the identification biases, no matter whether the current or voltage measurement is corrupted by noises. Further, the biased results will cause the accuracy of state-of-charge estimation to be deteriorated significantly. In order to enhance the accuracy of state-of-charge estimation, a co-estimation method is proposed that employs recursive restricted total least squares to identify model parameters and unscented Kalman filter to estimate the state-of-charge. The required noise covariance matrix is estimated by noise covariance estimator, which is based on polynomial Kalman smoother. Moreover, the superiority of the proposed method is verified by comparing with the two existing state-of-the-art methods in terms of the accuracy and convergence speed. By employing the proposed method, the mean absolute errors and the convergence time of state-of-charge estimation can be limited within 1.2% and 88 s under different driving cycles and ambient temperatures, respectively.