Key points are not available for this paper at this time.
State-of-charge (SOC) estimation for lithium–iron phosphate (LFP) batteries is a challenging task due to their path-dependent behavior, flat open circuit voltage (OCV) characteristics, and hysteresis effects. This work proposes a machine-learning-based SOC estimation method designed for onboard applications, addressing the challenges of SOC initialization when using the Coulomb counting method. The proposed approach relies on low sampling frequency measurements during short-term rest periods. Experiments were conducted on LFP 26650 cells across more than 430 working conditions, involving four temperatures, three current rates, four cycling scenarios, with various resting periods at different SOC levels. A comprehensive analysis of SOC estimation errors, including initial value errors, sensor noise, and sampling frequency, is provided. Using relaxation voltage data recorded at intervals as short as 1 min, the SOC resetting estimation solution proposed in this paper achieves mean absolute errors lower than 3.25%, demonstrating its potential for real-world applications. This solution can be readily integrated into existing battery management systems.
Che et al. (Mon,) studied this question.