A Rapid Diagnosis Procedure of Ultra Micro-Short Circuit for Lithium-Ion Batteries Considering Computation-Constrained Condition

Early identification of ultra micro-short circuit (UMSC) in batteries can capture abnormal characteristics at the earliest opportunity, which is of great significance for extending battery lifespan and preventing thermal runaway (TR). However, existing studies primarily address microshort circuit (MSC), which manifest after a certain period of battery damage, due to the inherent challenge of detecting UMSC signatures. In this work, to cater to the cloud-edge collaborative battery management system (BMS) framework, we design a multilevel UMSC fault intelligent diagnosis procedure (FIDP) that enables edge deployment. Based on proposed UMSC time-variant model with the premise of active detection means, FIDP employs Markov Chain Monte Carlo (MCMC)-based pseudo-experimental data generation technique and multidimensional feature screening engineering to achieve a fault diagnosis accuracy of 96.91% on the test set covering continuous fault conditions using Bayesian optimized support vector machine (SVM) model. Finally, the UMSC fault diagnosis model produced from FIDP framework has been successfully embedded into a computation-constrained BMS board, significantly enhancing its applicability in electric vehicles (EVs).