Cloud Platform-Oriented Electrical Vehicle Abnormal Battery Cell Detection and Pack Consistency Evaluation with Big Data: Devising an Early-Warning System for Latent Risks

A battery is grouped into many cells, and inconsistency is unavoidable in the battery life cycle. If the battery is frequently charged or discharged without a balancer, the battery cells with the lowest capacity may be overcharged or overdischarged, which is one of the major reasons for battery thermal runaway, which can cause a fire. This article proposes a cloud data-based electric vehicle (EV) battery-voltage consistency evaluation technique for vehicles in service. The density-based spatial clustering of applications with noise (DBSCAN) method is employed to improve the computational efficiency of the variance of angle (VOA), which is a widely used outlier-detection scheme. The DBSCAN-VOA results are compared with VOA results, showing that the distinction capability of VOA is kept while the computational complexity is significantly reduced. To assess the pack's consistency under real operation, a benchmark open-circuit voltage (OCV)-based consistency approach is developed in a simulated lab environment. A big data-based online battery pack consistency-state evaluation technique is established using the deviation value statistical method, and the efficiency of the process is discussed.