Research: Enhancing Lithium-Ion Battery State-of-Health Estimation via Electrochemical ...

Research question: Can machine learning models trained on electrochemical impedance spectroscopy data improve the accuracy of lithium-ion battery state-of-health estimation compared to traditional methods?

Lithium-ion batteries underpin a wide range of technologies, from electric vehicles to grid-scale energy storage. Accurate state-of-health (SOH) estimation is critical for optimizing battery lifespan, safety, and performance. Conventional SOH methods often rely on simplistic empirical models that may not capture the complex degradation mechanisms inherent in modern batteries.

Electrochemical impedance spectroscopy (EIS) offers a rich dataset reflecting internal battery dynamics, yet its integration with advanced machine learning (ML) techniques for SOH estimation remains underexplored. Existing literature primarily focuses on either EIS interpretation via physical models or ML approaches using limited sensor data, leaving a gap in combined methodologies.

This project proposes to systematically evaluate ML models—including random forest, support vector machines, and neural networks—trained on EIS data for SOH prediction. The methodology will involve experimental EIS measurement on commercial lithium-ion cells, preprocessing and feature extraction, and comparative analysis against baseline estimation methods.

The expected contribution is a rigorous assessment of the potential for ML-enhanced EIS-based SOH estimation, informing future battery management systems. The findings could significantly improve predictive maintenance, reduce operational risks, and facilitate more reliable integration of batteries into critical energy infrastructure.