Research on Early Multi-fault Diagnosis of Lithium Battery Under Dynamic Working Condition

The frequent occurrence of battery pack failures brings a great threat to the development of electric vehicles. Battery pack faults are generally multiple and diverse and have similar fault characteristics, which are difficult to distinguish and detect, and are not conducive to fault diagnosis and classification. Therefore, this paper proposes a new sensor connection topology to distinguish the types of faults within a battery pack, such as connection faults, sensor faults, and cell short-circuit faults. In particular, for connection faults, the topology proposed in this paper can more accurately locate the specific point of the connection fault. In addition, due to the large inconsistency of the battery pack data, and the fault features inside the battery are very tiny in the pre-fault period. To address this problem, this paper proposes a battery feature extraction method based on the autocorrelation coefficient, which can effectively avoid the problem of battery inconsistency and amplify the fault features. Finally, for the automatic identification of faulty cells, this paper proposes a fault value calculation method based on Kullback-Leibler (KL) dispersion, which can effectively identify faulty cells. The method validates the performance of multiple fault detection under UDDS operating data, and the results show that the method proposed in this paper can effectively detect and differentiate multiple faults in battery packs.