Short circuit detection in lithium-ion battery packs

Abusive lithium-ion battery operations can induce micro-short circuits, which can develop into severe short circuits and eventually thermal runaway events, a significant safety concern in lithium-ion battery packs. This paper aims to detect and quantify micro-short circuits before they become a safety issue. We develop offline batch least square-based and real-time gradient-based state of health (SoH) estimation approaches, coupled with a state of charge (SoC) observer, to estimate the leakage current of individual cells from measured cell voltages and currents. Even in the presence of current sensor bias and cell heterogeneities, cell-to-cell comparison of leakage currents allows the determination of outlier cells that may have soft shorts. The proposed method is tested using field data from a battery electric locomotive under nominal operation and external short circuits (ESC). With sufficiently excited current inputs, the experimental results show that a leakage current of more than 27 mA ( C / 4000 ) can be accurately detected. Using field test data from a battery electric locomotive, an experimental 15 Ω ESC that produces a leakage current of C / 464 in a 3P-22S pack is detected within 2 h. • Micro short detection framework in lithium-ion battery pack is presented. • Offline least square-based and real-time gradient-based SoH estimators are proposed. • SoH estimators accurately estimate cell capacity, resistances, and current mismatch. • Micro short circuits are identified by cell-to-cell comparison of current mismatch. • Statistical testing results show fast and accurate fault detection capabilities.