Modified extended Kalman filtering algorithm for precise voltage and state-of-charge estimations of rechargeable batteries

Abstract

With the popularization of electric vehicles (EVs), the voltage and state-of-charge (SOC) estimations of rechargeable batteries are of great significance. The SOC parameter has been used as an indicator for delivering the electrical energy of rechargeable lithium-ion batteries (LIBs), while the voltage has been a critical parameter needed to monitor to prevent the cause of battery damage especially during the charging and discharging process. Thus, the research focus is to estimate the SOC and voltage accurately using algorithms. With the capability of avoiding major estimation errors, conventional extended Kalman Filtering (EKF) has been employed to estimate the optimal value of SOC using the parameters obtained by indirect measurements, such as voltage and current. However, the algorithm suffers from limited precision in SOC and voltage estimations, and there is still no in-depth investigation of the error reduction in voltage prediction. Although the SOC accuracy can be improved by a joint algorithm such as double Kalman filtering, optimization of EKF itself is still needed due to the superposition of nonlinear errors. This study reveals that the conventional EKF algorithm induces estimation errors especially when the current abruptly changes. In this study, the research on the modified extended Kalman filtering (MEKF) algorithm was conducted for estimating the voltage and SOC of the LIBs with great improvement in estimation accuracy. The YUASA LEV50 cell was subjected to a standard discharging rate of 0.2C at 298 K to acquire offline parameters, followed with the newly proposed dynamic estimation mathematical battery model (DBOFT) for the optimization of estimations. This is the first time to propose a method of combining gain matrix and noise to reduce the error of voltage estimation at the current turning points, which greatly improves the accuracy of voltage estimation. Specifically, the MEKF algorithm is capable of adjusting the parameters in real-time and reducing the SOC estimation error. The SOC error estimated by the MEKF algorithm was reduced to 0.0052 % under the current rate of 0.2C. Finally, the maximum estimated voltage error was reduced to 0.972 % while the SOC estimation error was reduced to 0.01016 % under the same working current. The precise voltage and SOC estimations obtained by the MEKF with higher accuracy than traditional methods were validated under experimental verification, which is crucial for the life extension of LIBs and the safety of the battery management system (BMS).