Online diagnosis of soft internal short circuits in series-connected battery packs using modified kernel principal component analysis

Abstract

Safe operation of large battery storage systems requires advanced fault diagnosis that is able to detect faults and provide an early warning in the event of a fault. Since Internal Short Circuits (ISC) are the most common abuse condition leading to thermal runaway, this study addresses the early detection of incipient soft ISCs at the stage when the fault is still uncritical and does not lead to significant heat generation. The differences in cell voltages as measured by conventional battery management systems prove to be indicative features for ISC diagnosis. However, due to poor balancing and parameter variations, the cell voltage differences exhibit nonlinear variations. This work addresses this challenge with a nonlinear data model based on Kernel Principal Component Analysis (KPCA). To enable an online application in a vehicle, the present work reduces the computational complexity of the method by an optimal choice of training data. An analysis of the contribution of each cell to the fault statistics enables identification of the faulty cell. Since early-stage ISCs can exhibit a wide range of short-circuit resistances, experimental validation is performed with resistances from 10Ω to 10kΩ, which are correctly detected and isolated by the optimized cross-cell monitoring in all cases.