Correlation analysis and feature extraction using impedance spectroscopy over aging of lithium ion batteries

• Published in: Journal of energy storage 2025-01, Vol.105, p.114715, Article 114715
• Main Authors: Pohlmann, Sebastian, Mashayekh, Ali, Buberger, Johannes, Estaller, Julian, Wiedenmann, Andreas, Kuder, Manuel, Neve, Antje, Weyh, Thomas
• Format: Article
• Language: English
• Subjects: Electrochemical impedance spectroscopy; Feature selection; Lithium-ion battery; Machine learning; State-of-health
• ISSN: 2352-152X
• DOI: 10.1016/j.est.2024.114715

To optimize the operation of batteries and to accelerate the transition to a renewable energy supply and sustainable transportation, it is crucial to determine the condition of Lithium-Ion Batteries at all time. A non-invasive tool for determining the State-of-Health of a battery cell is the electrochemical impedance spectroscopy, in which the impedance is calculated as a function of the excitation frequency. This paper presents a detailed correlation analysis between impedance and State-of-Health and, therefore, the dependency between capacity and impedance over the life cycle of a battery. After extensive testing series with 48 cells resulting in 25,344 impedance spectra, the highest correlations between the impedance and the State-of-Health could be obtained at a State-of-Charge of 10%. Further, features are extracted from the impedance based on their correlation to estimate the State-of-Health. To validate the results of the correlation analysis, a support vector regression and a multi-layer perceptron are trained and tested resulting in a mean absolute error of 0.86% and 0.84%. The estimation results confirm the correlation analysis and further substantiate the need for an appropriate feature extraction method. •Analysis of extensive EIS measurements over battery aging.•Impact analysis of the SoC on correlation between impedance and SoH.•Identification of the most suitable condition of a LIB to estimate the SoH.•Development of a feature extraction method for ML models based on EIS data.•Application of the correlation results in two ML models to estimate the SoH.