FURI | Spring 2025

Battery Health Prediction of Lithium-ion Batteries From Impedance Spectroscopy Using Machine Learning

Predicting the health of lithium-ion batteries is important for their reliable and safe use in applications like electric vehicles and consumer electronics. This project uses Gaussian process regression, a machine learning approach, to uncover patterns between various measured parameters of a battery and battery state of health. The model is trained on a diverse dataset of commercial lithium-ion batteries spanning different charging rates and temperatures. The model identifies resistance measurements as the dominant predictor of battery capacity. The ultimate goal is to develop a data-driven framework for real-time battery health diagnosis to predict capacity and degradation that can be easily integrated into battery management systems.