Early Detection of Rare Events: Predicting Battery Cell Deviations

Abstract

Despite rigorous quality controls in battery cell production, the production process is still subject to quality deviations. These quality deviations; known as "rare events", initially act as passive quality deviations which may not affect the battery’s perfor mance. However, a passive quality deviation can transition into an active deviation that may give rise to behavioral issues in the battery cell at some point during its lifetime. An active quality deviation can cause the entire battery to misbehave and eventually fail. This thesis investigates the possibility of predicting these cell devia tions in car batteries. Better predictions of these events would avoid expensive and troublesome car failures and enable preventive car maintenance to solve the problem. In this report, different models are created and evaluated with the aim of preventing these deviations. The dataset is supplied by Volvo Cars and contains a large amount of data collected from Battery Electric Vehicles (BEVs), where the arguably largest challenge comes from the imbalance of the dataset. In addition to the modelling, the thesis includes a thorough data analysis with the aim of improving both the dataset itself and the data collection process at Volvo Cars. These deviations occur extremely rarely, making a relatively large amount of false positives difficult to avoid. The results show that a simple time series model can catch these deviations relatively well but also brings along a large number of false positives. A neural network is able to improve this significantly, still being able to catch the majority of the deviations while producing a lot fewer false positives.