Comparative Analysis of Non-Linear Kalman Filters for Li-ion Battery SoC Estimation with RLS-VDF Technique for Parameters Identification

• Published in: 2023 International Conference on Computer, Electronics & Electrical Engineering & their Applications (IC2E3) pp. 1 - 6
• Main Authors: Jaiswal, Himanshu, Ghule, Swapnil Sanjay, Shah, Varsha A.
• Format: Conference Proceeding
• Language: English
• Published: IEEE 08.06.2023
• Subjects: Adaptation models; adaptive improved unscented Kalman filter; Estimation; Filtering algorithms; Li-ion batteries; Lithium-ion batteries; Mathematical models; non-linear Kalman filters; Parameter estimation; recursive least square variable direction forgetting; state of charge estimation; Uncertainty
• DOI: 10.1109/IC2E357697.2023.10262704

Accurate state of charge (SoC) estimation is necessary for safe and efficient functioning of lithium-ion (Li-ion) batteries in a miscellany of applications, such as portable gadgets, grid storage systems, and electric cars. However, the accuracy of SoC estimation highly depend on the accuracy of the battery model and its parameters identification, which is often difficult to obtain due to the complicated behavior of Li-ion batteries. Recursive Least Square variable directional forgetting (RLS-VDF) technique is a parameter identification algorithm that can improve the parameters estimation of the battery model in real-time, leading to more accurate battery models for non-linear Kalman filters (NLKFs). NLKFs have been generally employed for SoC estimation due to their capability to handle non-linearities and uncertainties in the battery system. This paper assessed the effectiveness of various NLKFs, such as extended Kalman filter, adaptive extended Kalman filter (AEKF), unscented Kalman filter (UKF), and adaptive improved unscented Kalman filter (AIUKF) with the RLS-VDF technique for parameter identification. The accuracy and robustness of the various NLKFs are evaluated using experimental data from a commercial Li-ion battery. The results indicate that the AIUKF with RLS-VDF provides the most accurate and robust SoC estimation performance among the four filters.