FURI | Spring 2026

Machine Learning Enabled Prediction of Lithium-Ion Battery Degradation

Forecasting lithium-ion battery degradation remains a key challenge for modern energy storage. Prior studies using Gaussian Process Regression (GPR) identified low frequency Electrochemical Impedance Spectroscopy (EIS) signals as indicators of aging in lab-scale batteries. This research evaluates an alternative machine learning approach, eXtreme Gradient Boosting (XGB), and compares it to these prior methods. The models are applied to commercial batteries and benchmarked based on how clearly they identify which frequency ranges the EIS signals are more important for determining degradation. The objective is to determine the most effective approach for improving battery health predictions and enabling more reliable lifespan estimation.