Fault Diagnosis of High- Temperature Superconducting Cables Based on TL-AGWO-BiLSTM

As high-temperature superconducting (HTS) cables become increasingly integral to modern power systems, the need for precise fault diagnosis has grown more pressing. Yet, traditional diagnostic approaches often fall short due to the lack of labeled data and the intricate, nonlinear nature of fault patterns in real-world scenarios. To tackle these issues, this study introduces an innovative fault diagnosis framework for HTS cables, leveraging transfer learning (TL), adaptive grey wolf optimization (AGWO), and bidirectional long short-term memory (BiLSTM) networks. Within this framework, AGWO is utilized to fine-tune the hyperparameters of BiLSTM through a comprehensive global search, significantly boosting the model's capacity to capture the transient behaviors of HTS cable quenching. Additionally, transfer learning is employed to transfer fault-related insights from a source domain to the target domain, effectively addressing the challenge of limited data and enhancing the model's adaptability across diverse operating conditions. Experimental findings reveal that the proposed method achieves an impressive 98.2 % accuracy in diagnosing short-circuit faults. Beyond validating the model's efficacy, this research enhances the identification of complex fault characteristics in HTS ca-bles, minimizes reliance on extensive datasets, lowers training expenses, and offers a robust solution for efficient and reliable fault monitoring in practical engineering applications.