|About this Abstract
||2022 TMS Annual Meeting & Exhibition
||Advanced Materials for Energy Conversion and Storage 2022
||Prevention of Thermal Runaway in Li-ion Batteries Using Machine Learning Model Prediction
||Meghana Sudarshan, Alexey Serov, Casey Jones, Vikas Tomar
|On-Site Speaker (Planned)
Thermal runaway failure of Lithium-ion battery (LIB) pack can be caused by a rise in temperature of a single battery affecting other batteries in the pack, leading to catastrophic results. Thus, it is vital to detect overheating and predict the temperature rise of batteries in the upcoming battery cycles to avoid excessive heat generation. Prediction of temperature can provide a warning for improved safety of batteries by avoiding dangerous situations. In this study, we train, test, and compare machine learning-based models using experimental data of 18650 LIB and publicly available datasets to predict the temperatures of cells for future cycles. A well-trained offline temperature prediction model makes a battery management system online monitoring and prediction of temperature feasible in LIBs. A data-driven model architecture is proposed considering major thermal runaway factors during battery cycling like over-charging, over-discharging, and battery aging during temperature prediction.