A novel few-shot learning approach for cross-domain fault diagnosis in high-voltage circuit breakers

High-voltage circuit breakers (HVCBs) play a pivotal role in ensuring the reliability and safety of power systems. However, cross-domain fault diagnosis remains a challenging task due to domain shift and limited labeled data. This paper introduces a novel few-shot learning framework, the integrated transfer fine-grained metric network (ITFGMN), specifically designed to tackle these issues. The proposed framework integrates three key innovations: (1) a channel-focused convolutional neural network module is introduced for effective feature extraction, enabling the capture of domain-invariant patterns; (2) a domain alignment engine is incorporated to bridge the domain gap, facilitating improved feature alignment across domains; and (3) a fine-grained metrics module employs a weighted prototype-based strategy to dynamically optimize the contribution of support samples and mitigate negative transfer. Comprehensive experiments on real-world HVCB datasets demonstrate that ITFGMN achieves superior performance compared to state-of-the-art methods in cross-domain fault diagnosis, showcasing its potential for practical deployment. • A novel approach for cross-domain fault diagnosis in high-voltage circuit breakers. • Tackles data scarcity and domain shifts using few-shot learning techniques. • Enhances model generalization with domain-invariant feature extraction. • Mitigates negative transfer by dynamically adjusting sample influence. • Outperforms existing methods, demonstrating robust performance across domains.