Online model-based SOC and capacity co-estimation using battery internal characteristics

Abstract

Lithium-ion batteries, renowned for their enduring lifespan and low weight, are increasingly gaining importance across industries as a feasible alternative to conven tional fossil fuels. Consequently, the growing significance of battery management sys tem (BMS) for evaluating battery health has become apparent. However, the changing capacity of several factors hinders accurate state of charge (SOC) estimation. Al though several methods have been proposed to co-estimate SOC and capacity, they typically assume that capacity exhibits slower variations compared to SOC. To this end, we introduce an online SOC and capacity co-estimation method using the inter nal characteristics of lithium-ion batteries. This method is adept at capturing trends in changing capacity. Our method identifies battery internal parameters using the forgetting factor recursive least sqaure (FFRLS) algorithm. Next, the terminal volt age, internal resistance, and open circuit voltage (OCV) enter into the measurement values of the adaptive extended Kalman filter (AEKF). The measurement covariance matrix changes depending on the current because of the relationship among measure ment values. At each time step, the AEKF concurrently co-estimates SOC and capac ity. Through experiments involving aging and time-varying temperature conditions, the proposed method demonstrates SOC estimation errors consistently below 1% and maximum errors under 2% for aging experiments, while for time-varying temperature experiments, SOC errors remain under 1% with maximum errors below 7%.