Abstract External short circuit (ESC) of lithium-ion batteries is one of the common and severe electrical failures in electric vehicles. In this study, a novel thermal model is developed to capture the temperature behavior of batteries under ESC conditions. Experiments were systematically performed under different battery initial state of charge and ambient temperatures. Based on the experimental results, we employed an extreme learning machine (ELM)-based thermal (ELMT) model to depict battery temperature behavior under ESC, where a lumped-state thermal model was used to replace the activation function of conventional ELMs. To demonstrate the effectiveness of the proposed model, we compared the ELMT model with a multi-lumped-state thermal (MLT) model parameterized by the genetic algorithm using the experimental data from various sets of battery cells. It is shown that the ELMT model can achieve higher computational efficiency than the MLT model and better fitting and prediction accuracy, where the average root mean squared error (RMSE) of the fitting is 0.65 °C for the ELMT model and 3.95 °C for the MLT model and the RMES of the prediction under new data set is 3.97 °C for the ELMT model and 6.11 °C for the MLT model.

中文翻译：

---

基于极限学习机的电动汽车锂离子电池外部短路热模型

摘要 锂离子电池外部短路（ESC）是电动汽车常见且严重的电气故障之一。在这项研究中，开发了一种新的热模型来捕捉 ESC 条件下电池的温度行为。在不同的电池初始充电状态和环境温度下系统地进行了实验。基于实验结果，我们采用基于极限学习机（ELM）的热（ELMT）模型来描述ESC下的电池温度行为，其中使用集总态热模型代替传统ELM的激活函数。为了证明所提出模型的有效性，我们使用来自各种电池组的实验数据，将 ELMT 模型与通过遗传算法参数化的多集总态热 (MLT) 模型进行了比较。结果表明，ELMT模型可以实现比MLT模型更高的计算效率以及更好的拟合和预测精度，其中拟合的平均均方根误差（RMSE）对于ELMT模型为0.65°C，对于拟合模型为3.95°C。 MLT 模型和新数据集下预测的 RMES ELMT 模型为 3.97 °C，MLT 模型为 6.11 °C。