Kalman filtering techniques for the online model parameters and state of charge estimation of the Li-ion batteries: A comparative analysis

The state of charge (SoC) is the most commonly used performance indicator of battery used in various applications. A chronic erroneous estimation of battery SoC may result in constant over charging and discharging, which in turn causes permanent damage to the internal structure of the battery cells along with system disruptions. This paper presents a comprehensive review of different techniques for SoC estimation of batteries, followed by a review of Li-ion battery model parameter estimation methods. Then this paper classifies the Kalman filters (KFs) in a systematic manner and conducts a detailed literature review on the linear Kalman filter (LKF) and non-linear Kalman filters (NLKFs). In recent literature, the NLKFs such as extended Kalman filter (EKF), adaptive EKF (AEKF), unscented Kalman filter (UKF), and adaptive UKF (AUKF) are the most extensively established techniques for an accurate and reliable SoC estimation of batteries. However, the precise estimation of battery SoC using the Kalman filters largely relies on accurate battery modeling and its online model parameter estimation. According to the literature, the recursive least square (RLS) and the polynomial regression-based battery model (PRBM) are the most often used techniques for estimating real-time model parameters of Li-ion batteries. Therefore, this paper performs an experimental comparative performance evaluation of the most popularly used NLKFS and battery modeling techniques in terms of SoC estimation accuracy at constant and varying operating conditions. The EKF, AEKF, UKF, and AUKF techniques augmented with the popularly used RLS or PRBM are first developed and tested with offline measured data in the MATLAB platform. Then they are implemented on the LabVIEW based battery testing platform using the Math-Script feature of MATLAB for real-time parameters and SoC estimation. Rigorous experimental studies have been carried out for comparative performance evaluation of the PRBM-EKF, PRBM-AEKF, PRBM-UKF, PRBM-AUKF, RLS-EKF, RLS-AEKF, RLS-UKF, and RLS-AUKF techniques under the standard room temperature (25 °C) and a wide temperature range (−5 °C to 45 °C). Overall, the PRBM-AUKF and RLS-AUKF surpassed other approaches in terms of SoC estimation accuracy.