Abstract
Lithium-ion battery failure is the main cause of electric vehicle fire accidents. In this paper, we propose a fault analysis framework for big data-driven fault trace extraction based on the whole-life-cycle charging data of onboard lithium-ion batteries. Firstly, battery voltage features strongly correlated with faults are mined and automatically selected by a random forest algorithm from the last-one-cycle operation data before sample accidents. Secondly, by usage of the vector sample points composed of selected features, density clustering is applied to identify faulty cells, and their fault traces in the whole life cycle are tracked through utilizing the Gaussian mixture model. This work uses more than ten real vehicle data for verification. The results show that the proposed method can detect the abnormality of one fault cell at least dozens of cycles in advance, or even in the earliest stage.
Original language | English |
---|---|
Pages (from-to) | 1-13 |
Number of pages | 13 |
Journal | IEEE Transactions on Power Electronics |
DOIs | |
Publication status | Accepted/In press - 2024 |
Keywords
- Accidents
- Batteries
- Circuit faults
- Data mining
- Fault diagnosis
- Feature extraction
- Gaussian mixture model
- Lithium-ion batteries
- Lithium-ion batteries
- density cluster
- fault diagnosis
- fault tracing