An Innovative Magnetic Field Sensor based Method Predicting Lithium-ion Battery State-of-health

Hui, Z; Mamlouk, M; Zhang, J; Bhowmik, D

doi:10.1109/SAS65169.2025.11105130

An Innovative Magnetic Field Sensor based Method Predicting Lithium-ion Battery State-of-health

Lookup NU author(s): Zachary Hui, Professor Mohamed Mamlouk ORCiD, Dr Jie Zhang ORCiD, Dr Deepayan Bhowmik ORCiD

Downloads

Accepted version [.pdf]

Licence

This work is licensed under a Creative Commons Attribution 4.0 International License (CC BY 4.0).

Abstract

As a portable energy storage device, lithium-ion batteries are widely used in various industries. However, with the increasing number of consumers, the number of accidents caused by the instability of aging lithium-ion batteries, such as spontaneous combustion due to thermal runaway, has also increased, raising significant safety concerns among consumers and manufacturers. Battery aging is fundamentally the result of structural changes within the cell, which are presented as phenomena such as swelling of the battery. These internal structural alterations lead to an inhomogeneous current distribution during operation, further exacerbating the instability of the battery. To address the above issue, we propose an innovative magnetic sensor-based battery state-of-health prediction method that differs from traditional state-of-health prediction approaches, such as data-driven models and equivalent circuit-based methods. Instead of relying on indirect data analysis, our approach employs a magnetic field sensing device to capture real-time variations in the battery magnetic field during charging and discharging. A traditional partial least squares regression(PLSR) model and a convolutional neural network (CNN) are trained to capture the evolving characteristics of the battery’s surface magnetic field, which grows increasingly inhomogeneous with aging. By directly utilizing the intrinsic properties of the battery instead of relying on secondary data analysis to infer lifespan, our method achieves a high R2 value of 0.92, providing a non-destructive, more precise, and reliable approach to predicting battery state-of-health.