As discussed above, the faults diagnosis and abnormality of battery pack can be detected in real time. In addition, timely detection and positioning of faults and defects of cells can improve the health and safety of the whole battery pack.
By applying the designed coefficient, the systematic faults of battery pack and possible abnormal state can be timely diagnosed. 2) The t-SNE technique, The K-means clustering and Z-score methods are exploited to detect and accurately locate the abnormal cell voltage.
Common electrical faults of battery packs can be divided into three categories: abuse , sensor faults and connection faults . Battery abuse faults mainly refer to external short circuit (ESC), internal short circuit (ISC), overcharge and over-discharge.
However, the proposed methods in these works [, , , ] are mainly based on the voltage data of a single cell in battery packs, and they cannot accurately diagnose faults and anomalies incurred by variation of other parameters, such as current, temperature and even power demand.
To verify the fault diagnosis ability of proposed method for battery packs, the partial random historical operation data of one E-scooter is considered for validation. Note that the E-scooter selected for validation is excluded from the data of 86 E-scooters for training.
Then, using the three driving fragments, the inconsistency fault factors on a battery of EV is studied. A fault diagnosis method based on spatial clustering for EV inconsistency fault detection and LS-SVR method for fault prediction under real-world driving conditions is proposed.