Physics informed machine learning for LiBs' aging behavior estimation and prediction

Abstract: The capacity and power capabilities of LiBs gradually decrease with dynamic operation, leading to reduced lifetime and even posing safety hazards, which represents a significant environmental threat. Accurate and robust methods for estimating the aging behavior of LiBs and their lifetime must be investigated to obtain a high-performance BMS. Black-box models that only detect external environments such as voltage, current, and temperature as inputs are strongly data-dependent and lack the physical views into aging mechanisms to guide generalized prediction. Therefore, this research aims to develop physics-informed algorithms considering internal aging mechanisms and external factors to obtain information on the LiBs’ state of health and failure lifetime.