Integrating Thermal Mechanisms with Machine Learning for Accurate State of Health Estimation in Lithium-Ion Batteries

Lithium-ion batteries have been a crucial contributor to the development of e-mobility, due to their high energy density, extended cycle life and commercial needs. As efforts towards creating an eco-friendlier transportation system continue, lithium-ion batteries are expected to maintain their significance as a crucial component in achieving a sustainable future. However, lithium-ion battery performance is significantly impacted by numerous factors including factors like thermal stress. Accurate estimation of state of health provides quantification of degradation caused to the battery. The ability of machine learning (ML) techniques to analyse the behaviour of non-linear systems has received increasing attention. With the rise of big data and cloud computing, there is significant potential for ML technology to be utilized in the calculation of battery state of health. Purely data-driven models cannot accommodate in-depth analysis of ageing mechanism. It is therefore essential to integrate machine learning techniques with physical mechanisms that affect battery health. This paper focuses on this existing challenge by proposing a novel method for improving machine learning model accuracy by inducting thermal stress component. In this paper, comparative analysis of nine different machine learning algorithms has been performed using two methodologies, results show significant improvement in the performance metrics when thermal stress inducted.