State of Charge Estimation for Battery Cells: A Sliding–Mode Approach

Abstract

This thesis presents a comparative analysis of different state of charge (SOC) estimators in a battery management system. Different standard and sliding–mode algorithms to solve the robust SOC estimation problem are evaluated. An Equivalent Circuit Model (ECM) is presented for formulation purposes, with a parameterization based on a P2D model given by a high–fidelity simulator. The standard observers comprise the Luenberger and Extended Kalman Filter (EKF) algorithms, whereas the sliding–mode algorithms considered are first–order, second–order, and terminal–twisting algorithms. A tuning framework is proposed to find the optimal observer gains based on a global hyperparameter optimization algorithm. The results show, under an ideal scenario, that Kalman filtering provides a similar result to the best sliding–mode approaches. However, the sliding–mode algorithms present a higher robustness against nonlinear unmodeled dynamics, large parameter uncertainties, and measurement noise. Additionally, a computational effort analysis is performed, which demonstrates that the Luenberger observer is the least expensive algorithm followed by the sliding–mode approaches. Moreover, the results show that all the proposed algorithms are feasible; still, the best balance between robustness and computational effort is demonstrated by the first and second order sliding–mode observers followed by Luenberger observer. These statements are supported by the simulation results, confirming the feasibility of the proposed algorithms. To conclude, selecting an optimal state estimation technique requires a trade–off between accuracy, robustness, and computational efficiency.