Estimation of State of Charge and State of Health for Cylindrical Lithium-ion Battery with C-Si Anode

Abstract

Accurate measurement of the state of charge (SOC) and state of health (SOH), including capacity fade (SOHQ) and power fade (SOHP) that predicts the energy and power of the remaining lifespan of the battery is the core task in the battery management system (BMS). Since SOC and SOH are not directly measured, their estimation is predominantly executed based on the battery models. The models can be either based on an electrochemical model that considers chemical reactions in the battery or an equivalent circuit model (ECM). The latter is used more because of the low computational time and less complexity in its parameter identification. On the other hand, the anode materials are enhanced from carbon only to the carbon-silicon to increase energy density and theoretical capacity. However, silicon induces an extremely high volume change compared to carbon, which leads to the open circuit voltage (OCV) gap between charge and discharge at specific SOC. In fact, OCV is the most crucial parameter for estimating SOC since the OCV gaps and associated errors lead to inaccurate estimations of SOC and SOH. Therefore, a second-order ECM with hysteresis (ECMwH) representing the mechanical stress model is proposed. Based on the model, the capacity and internal resistance are estimated using two extended Kalman filters (EKF). The proposed algorithms are validated with an NCA/C-Si 21700 cylindrical cell with a nominal capacity of 5.3Ah using multiple charge-discharge test profiles under various temperatures and varying aging cycles. With this proposed method, the estimated results indicate less than 2.5% SOC estimation root mean square error (RMSE) and 2.1mV terminal voltage RMSE, which demonstrates a 3% SOC RMSE improvement compared to average OCV based method. SOHQ is estimated to have 0.43% RMSE. The estimation of SOHP is also validated. The results have shown that overall estimation tends to follow the measured values; however, errors increase as the SOC decreases.