Domain Adaptation With Contrastive Learning for Lithium-Ion Battery Packs Fault Diagnosis

Battery pack multi-fault diagnosis is essential for the safe operation of electric vehicles. However, the diverse operating conditions of electric vehicles contribute to a shift in the fault characteristics of battery packs, which further renders the battery fault diagnosis model defective in practical applications. Moreover, the scarcity of fault samples limits the advances in data-driven fault diagnosis methods. Addressing these problem, this paper proposes an unsupervised domain adaptation with contrastive learning multi-fault diagnosis method for electric vehicle battery packs. In the proposed framework, it employs momentum contrast to learn embedding representations that distinguishable fault types, and further align embedding representations between source and target domains through customized nearest neighbor contrastive learning. The effectiveness of the proposed method is experimentally demonstrated through different operating conditions and simulation to real-world transfer tasks. To the best of our knowledge, it is the first attempt of simulation to real-world knowledge transfer in electric vehicle battery pack fault diagnosis. The average fault diagnosis accuracy of the proposed method is 96.17% and 95.34% on different operating conditions and simulation to real-world transfer tasks, respectively, indicating that the proposed method outperforms the state-of-the-art fault diagnosis methods and can effectively detect and diagnose multiple faults in battery packs.