Early detection of Internal Short Circuits in series-connected battery packs based on nonlinear process monitoring. (April 2022)

• Record Type: Journal Article
• Authors: Schmid, Michael; Kleiner, Jan; Endisch, Christian
• Is Part Of: Journal of energy storage. Volume 48(2022)
• Publication Date: 2022-04
• Subjects: Lithium-ion battery -- Internal Short Circuit -- Fault diagnosis -- Kernel Principal Component Analysis -- Cell inconsistencies -- Battery safety; Energy storage -- Periodicals; Energy storage -- Research -- Periodicals; 621.3126
• DOI: 10.1016/j.est.2021.103732
• Languages: English
• ISSNs: 2352-152X

Abstract

Abstract: While the development of new materials in recent years has enabled an increase in energy density, power density and cycle life of batteries, safety remains a challenge. For electric vehicle applications, thermal runaway of a battery cell can lead to serious consequences. Thermal runaways are often caused by an Internal Short Circuit (ISC). This study aims to detect ISCs in the early latent phase, before extensive heat generation on the battery cell surface is measurable. To obtain high sensitivity, we apply a novel data-driven approach based on the cell voltage differences within the battery pack. Using the Kernel Principal Component Analysis (KPCA), a nonlinear data model is trained and applied for online detection of ISCs. By combining multiple kernel functions, fast detection and robust behavior is achieved for progressed ISCs while maintaining high sensitivity to soft ISCs. To demonstrate the applicability of the method in the presence of cell inconsistencies, the approach is experimentally validated on a calendar and cyclic aged module. A comparison with existing methods shows significant reduction in the detection time using the presented approach. Graphical abstract: Highlights: Novel approach for early detection of soft internal short circuits in battery packs. Training of a nonlinear data model based on the single cell voltage differences. Theoretical derivation and analysis to ensure a robust behavior for progressed faults. Experimental validation on module level shows significant reduction in detection time. Applicability independent of current profile, state of charge or cell inconsistencies.