Enhancing multi-type fault diagnosis in lithium-ion battery systems: Vision transformer-based transfer learning approach

Ensuring the accurate diagnosis of internal short circuit (ISC) faults and consistency anomalies is crucial for maintaining the high safety and longevity of battery systems. The challenge lies in the high similarity between the features of ISC faults and consistency anomalies, which can complicate accurate fault diagnosis. This study introduces a multi-fault diagnostic model that leverages the Vision Transformer (ViT) and employs simulated data to accurately pinpoint ISC faults and intricate consistency anomalies, such as capacity and state of charge anomalies, achieving an impressive accuracy rate of up to 100 %. The process begins with an in-depth analysis of fault mechanisms to extract and differentiate multi-fault features using a mean difference model. Subsequently, the ViT's exceptional ability to capture temporal information, which aids in the accurate identification of fault features. The integration of transfer learning with experimental data further refines the real-world diagnosis accuracy. The results confirm the robustness of the proposed method, highlighting its potential to substantially mitigate safety hazards in battery systems and enhance operational reliability.