Implementation of advanced machine learning on synthetic data for estimation of SOH and degradation of lithium-ion batteries

Lithium ion batteries have become one of the most important energy storage technologies due to their high energy density, adequate cycle life, and broad industrial applications. Given the critical need for precise monitoring of their state of health (SOH) to optimize both performance and safety, accurate estimation of battery health and the underlying causes of capacity fade is of dominant importance. In this study, using a synthetic database, we propose a comprehensive framework for estimating SOH and its respective degradation modes based on differential voltage and incremental capacity curves. Statistical analysis of the extracted features was conducted to implement support vector machine (SVM) and gradient boosting models for the estimation of SOH and the LLI, LAMPE, and LAMNE degradation modes. The results demonstrated that the SVM model outperformed the gradient boosting model, achieving R values of ۰.۹۷, ۰.۹۶, ۰.۹۴ and ۰.۹۹ for the LLI, LAMPE, and LAMNE degradation modes and SOH estimation, respectively.

Keywords:

SOH estimation , Lithium loss , degradation modes , SVR , machine learning