FURI | Fall 2025

Machine Learning-enabled Prediction of Lithium-ion Battery Degradation

Forecasting lithium-ion battery degradation remains a key challenge for modern energy storage. Building on Gaussian Process Regression studies that identified low-frequency Electrochemical Impedance Spectroscopy (EIS) features as aging indicators, this research evaluates an alternative eXtreme Gradient Boosting regression along with frequency weighting for improved predictions. This research aims to leverage EIS as a rapid state-of-health assessment tool which can be used to predict capacity fade and degradation processes in batteries. This will contribute to the predictions of battery health/lifespan and identify protocols to longer-lasting batteries guided by machine learning design.