Detecting faulty lithium-ion cells in large-scale parallel battery packs using current distributions.

One of the main concerns affecting the uptake of battery packs is safety, particularly with respect to fires caused by cell faults. Mitigating possible risks from faults requires advances in battery management systems and an understanding of the dynamics of large packs. To address this, a machine learning classifier based upon a support vector machine was developed that detects cell faults within large packs using a limited number of current sensors. To train the classifier, a modelling framework for parallel-connected packs is introduced and shown to generalise to Doyle-Fuller-Newman electrochemical models. The fault classification performance was found to be satisfactory, with an accuracy of 83% using current information from only 27% of the cells. Validation on experimental pack data is also shown. These results highlight the potential to combine mathematical modelling and machine learning to improve battery management systems and deal with the complexities of large packs.