State of charge evaluation of lithium-ion batteries under wide temperature range using multi-feature ultrasonic indicators

Abstract

The state of charge (SOC) of the battery, as a core parameter in the Battery Management System (BMS), directly affects the battery performance, lifespan, and safety. Traditional rudimentary methods such as interpolation and time-history integration of on-board collected data, utilized in the form of SOC-OCV lookup and coulomb counting, are generally not highly accurate because of changes in temperature and current, sensor measurement errors or difference between battery open circuit voltage and terminal voltage. Other strategies rely on models of battery dynamics, requiring physical and electrochemical models, but they disregard the internal structure and the mechanical evolution of the battery during charging and discharging. In response to the limitations of existing SOC estimation methods, this study proposes a lithium-ion battery SOC estimation method based on ultrasonic multi-feature indicators under different environmental temperature conditions. The method first acquires the acoustic response of the battery internal structure through non-destructive ultrasonic testing technology, then extracts key feature parameters from three dimensions: time domain, frequency domain, and time–frequency domain. Subsequently, the hyperparameters of the XGBoost model were optimized using the Whale Optimization Algorithm (WOA) to improve its predictive accuracy and robustness. Experimental validation has shown that the proposed WOA-XGBoost model exhibits superior performance in SOC estimation, with a root mean square error (RMSE) lower than other comparative models. Additionally, this study explores the impact of different feature parameter combinations on the estimation effect, further confirming the importance of multi-dimensional feature parameters in improving the accuracy of SOC estimation.