Electrochemical-distributed thermal coupled model-based state of charge estimation for cylindrical lithium-ion batteries. (April 2021)

• Record Type: Journal Article
• Authors: Chen, Guangwei; Liu, Zhitao; Su, Hongye; Zhuang, Weichao
• Is Part Of: Control engineering practice. Volume 109(2021)
• Publication Date: 2021-04
• Subjects: Distributed one-dimensional thermal model -- Electrochemical model -- Robust H-infinity temperature observer -- Backstepping technique -- State of charge (SoC); Automatic control -- Periodicals; 629.89
• DOI: 10.1016/j.conengprac.2021.104734
• Languages: English
• ISSNs: 0967-0661

Abstract

Abstract: State of Charge (SoC) is essential in a smart Battery Management System (BMS). Traditional SoC estimation methods consider the lithium battery cell as an isothermal body simplistically, which is not accurate. Due to the heat conduction and convection, the temperature distribution along the radius is nonuniform, which means the battery model parameters subject to temperature are not identical radially. This paper proposed a modified electrochemical-distributed thermal coupled model (MEDTM) for improving the accuracy of SoC estimation. The distributed thermal model is employed to design a robust H-infinity temperature observer for the estimation of the radial temperature distribution of battery cell, where a radial discretized method is used for the realization of the observer design. Based on the observed temperature, a SoC observer is developed with the backstepping method. Finally, experiments and simulation are implemented. The two constant discharging experiments indicate that MEDTM is more accurate than the single particle model with surface temperature (SPMST), and it can generate reliable temperature prediction. A comparative simulation and a constant discharging experiment verify the H-infinity temperature observer can obtain a more accurate estimation of the battery cell's temperature than the luenberger observer, and the H-infinity temperature observer has an estimation error of 0.3 K below. Then, through a constant discharging experiment and a simulation with UDDS current, the proposed SoC observer is verified to accurately estimate the SoC by comparing with the true SoC, where the estimated error of SoC under the two cases is below 1.5% and below 0.4%, respectively. Therefore, the proposed SoC observer also has good performance.