Model-Constrained Deep Learning for Online Fault Diagnosis in Li-ion Batteries over Stochastic Conditions

Here are the datasets for the publication named "Model-Constrained Deep Learning for Online Fault Diagnosis in Li-ion Batteries over Stochastic Conditions". We utilize data uploaded from vehicle onboard BMS for network design. We have released the real vehicle dataset of 18.2 million valid entries from 515 vehicles collected by the BMS data center on the cloud. The dataset primarily includes data from three battery manufacturers, which due to confidentiality restrictions, are referred to as DTI, QAS, and GIS in this paper. In addition to normal data samples, the dataset also contains four types of hard-to-collect safety failure samples: thermal runaway, electrolyte leakage, internal short circuit, and excessive aging.