심층 연산자 네트워크(DeepONet)를 이용한 리튬이온 배터리 열폭주 예측

초록

The development of lithium-ion batteries (LIBs) has rapidly increased owing to their significant advantages. However, LIBs have several concerns related to their safety, including fires and explosions, and they suffer from the thermal runaway phenomenon, which limit their applications. This study proposes a deep operator network (DeepONet) for predicting the thermal runaway phenomenon of LIBs under a variety of thermal operational and abuse conditions. In particular, the DeepONet aims to use the functional mapping derived from a heating curve to predict the evolution of surface temperature and the dimensionless concentrations of the dominant components of LIBs, such as the anode, cathode, electrolyte, and solid-electrolyte interphase. Temperature evolution under various thermal operational and stress conditions was simulated using the high-fidelity finite element analysis (FEA) method, because thermal runaway measurements in batteries are complicated. A comparison of the DeepONet, high-fidelity FEA model, and experimental results revealed that the DeepONet has high accuracy and robustness under various thermal operational and abuse conditions. Moreover, the proposed surrogate model was considerably faster than the FEA model. Thus, the proposed surrogate model is thought to be effective for the thermal, power, and energy management of battery management systems during application.

키워드