Accelerated state of health estimation of second life lithium-ion batteries via electrochemical impedance spectroscopy tests and machine learning techniques. (February 2023)

• Record Type: Journal Article
• Authors: Faraji-Niri, Mona; Rashid, Muhammad; Sansom, Jonathan; Sheikh, Muhammad; Widanage, Dhammika; Marco, James
• Is Part Of: Journal of energy storage. Volume 58(2023)
• Publication Date: 2023-02
• Subjects: Electrochemical impedance spectroscopy -- Second life lithium-ion batteries -- Machine learning -- State of health -- Prediction; Energy storage -- Periodicals; Energy storage -- Research -- Periodicals; 621.3126
• DOI: 10.1016/j.est.2022.106295
• Languages: English
• ISSNs: 2352-152X

Abstract

Abstract: Estimating the state of health (SoH) of lithium-ion (Li-ion) batteries is a challenging task due to cross coupling and dependency between ageing mechanisms. An accurate estimation is particularly essential for second-life batteries to facilitate their successful implementation in the new application. By adopting the electrochemical impedance spectroscopy (EIS) test and a machine learning (ML) approach, the accelerated SoH estimation problem is studied here. For this purpose, 325 experiments for 30 Li-ion cells were conducted at various SoH, temperature, and state of charge. First an optimised Gaussian process regression model is developed and validated for SoH estimation. Then the sensitivity of the model is evaluated relative to measurement noise. Finally, the model's robustness is quantified through a case study involving cells that have been characterised with different physical test equipment. The results demonstrate that the model can predict the SoH of Li-ion cells with an error about 1.1 % and is reasonably robust to the various testing conditions of the battery. The methodology for handling the EIS data within a machine learning framework, the sensitivity analysis and the robustness quantification techniques are the main novelties of this study in the context of grading Li-ion batteries for second-life applications. Highlights: Accelerated SoH estimation of second life Li-ion batteries via machine learning and EIS tests Bayesian optimisation of Gaussian Process Regression models for hyperparameter tuning Quantifying the contribution of cell's SoC and temperature in their SoH prediction Evaluating the robustness of machine learning models to the test equipment and environment uncertainties