A novel charging framework for series-connected battery cells using desensitized Kalman filter-based model predictive control

In many high-power applications, lithium-ion cells must be connected in a series arrangement. Along with electrical states, the thermal states of these cells must be carefully managed. For this aim, model-based battery management systems (BMSs) shall be designed to ensure the safety and optimal performance of batteries. However, the performance and reliability of the model-based controllers can be greatly influenced by parameter uncertainties. As a result, a BMS must be designed such that uncertain parameters of cells can be effectively tolerated. This paper proposes a desensitized Kalman filter-based (DEKF) model predictive control (MPC) for the optimal charging of lithium-ion battery cells subject to parameter uncertainty. In this study, an experimentally derived electrothermal model is used to capture the cell dynamics. Further, the states of the model are estimated with EKF and DEKF when internal resistance is subject to +/- 10%\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\pm 10\%$$\end{document} uncertainty. The simulation results indicate that DEKF can reduce bias estimation while saving computational time. Also, the devised charging framework for series-connected cells demonstrates that the confidence bounds generated by DEKF can be utilized in MPC architecture to modestly tighten batteries' physical constraints.