Global Sensitivity Analysis on Temperature-Dependent Parameters of A Reduced-Order Electrochemical Model And Robust State-of-Charge Estimation at Different Temperatures

In this paper, a robust estimator is designed based on a reduced-order electrochemical model to achieve accurate state-of-charge estimation of lithium ion batteries at different ambient temperatures. The reduced-order electrochemical model, containing only two estimation states, describes the first principles within the battery and incorporates temperature-dependent properties, making it an ideal candidate for accurate estimation applications. A global sensitivity analysis based on Sobol indices is then conducted, investigating the impact of different temperatures on the model performance by changing the electrochemical properties of lithium ion batteries. The first-order and total Sobol indices obtained by Monte Carlo method show that variations of temperature-dependent parameters make different influence on the overall model uncertainties at different temperatures. Then, a robust adaptive extended Kalman filter based on the electrochemical model is designed for state-of-charge estimation to actively compensate such potential model uncertainties. Results against multiple experimental datasets at 45, 25 and 0 degrees C indicate that the proposed state-of-charge estimation algorithm is robust to imposed initial estimation errors, different driving cycle inputs and ambient temperatures. An extensive comparative analysis between the adaptive extended Kalman filter and a conventional extended Kalman filter is also presented. (C) 2021 Elsevier Ltd. All rights reserved.

Automated summary

This paper presents a robust estimator based on a reduced-order electrochemical model for accurate state-of-charge estimation of lithium ion batteries at different ambient temperatures. A global sensitivity analysis investigates the impact of temperatures on model performance, and a robust adaptive extended Kalman filter is designed to actively compensate potential model uncertainties. Results show the proposed algorithm's robustness against initial estimation errors, driving cycle inputs, and ambient temperatures, with extensive comparative analysis with a conventional extended Kalman filter.