A Simple Method for State of Health Estimation of Lithium-ion Batteries Based on the Constant Voltage Charging Curves

Abstract

The extraction of health factors (HFs) is a crucial step in estimating the state of health (SOH) of lithium-ion batteries using data-driven methods. In this study, five feature indices were extracted from the constant voltage stage of the constant current-constant voltage (CC-CV) charging curve. The correlation between these feature indices and SOH exceeds 80 %, reaching up to 98 %. Unlike traditional lithium-ion battery data processing methods such as incremental capacity analysis (ICA) or probability density function (PDF), this strategy avoids complex data preprocessing. Through variance inflation factor (VIF) analysis, the slope index was identified as a key health factor. When combined with data-driven algorithms, it successfully estimated the SOH of lithium-ion batteries composed of three different material types under different proportions of training sets. The results show that with a 20 % training set proportion, the LSTM algorithm achieved an RMSE of less than 1.3 % and a mean absolute error (MAE) of no more than 1.1 %. Additionally, the health factor extraction strategy proved to be robust across different time and current parameter ranges, with RMSE fluctuations within 0.35 %.

Keywords

• lithium-ion battery
• health factor
• long short-term memory
• random forests