19 August 2024 in IEEE Access
Effective monitoring and diagnosis of partial discharge (PD) in power equipment is crucial for maintenance, particularly given the expectations of significant increases in energy generation and consumption. Although deep neural networks have been widely applied in PD fault detection and classification, their performance suffers from insufficient labeled data in power equipment. This study proposes semi-supervised learning (SSL) to address the scarcity of labeled training data for PD classification in gas-insulated switchgear (GIS). The proposed SSL was validated based on phase-resolved PD and on-site noise using an ultra-high frequency (UHF) PD measurement system. Experimental results show that the proposed SSL achieves a high classification accuracy of 94.59% by effectively utilizing unlabeled data to enhance classification performance in GIS.
31 March 2025 in IEEE Access
Deep neural networks have proven to be highly efficient in fault detection and classification using partial discharges (PDs) in gas-insulated switchgear (GIS). However, previous studies have not fully addressed the issue of limited labeled training data, which significantly impacts the performance of these models. Existing semi-supervised learning (SSL) approaches typically discard low-confidence pseudo-labels from unlabeled PD samples, because these unreliable labels can mislead the model. This gap in current research overlooks the potential value of low-confidence samples, which, despite their uncertain class, are unlikely to belong to the classes with the lowest probabilities. In this study, we aim to overcome this limitation by proposing a novel semi-supervised contrastive complementary learning (SCCL) method. Our SCCL approach generates a larger set of reliable negative pairs using complementary labels, allowing us to utilize the entire set of unlabeled PD samples effectively. We validate the feasibility of SCCL using phase-resolved PDs (PRPDs) and onsite noise data collected through an ultrahigh-frequency (UHF) PD measurement system. Experimental results indicate that the SCCL method achieves an impressive accuracy of 96.23% by effectively leveraging low-confidence unlabeled PD samples to improve classification performance in GIS under restricted labeling conditions.
17 November 2025 in IEEE Access
Partial discharge (PD) detection and classification are essential to ensure the reliability of gas-insulated switchgear (GIS). However, conventional deep learning approaches require extensive labeled data, which are expensive and time-consuming to obtain. To address this challenge, we propose a novel semi-supervised learning (SSL) framework called CDMAD-SSL, which integrates class distribution mismatch-aware debiasing (CDMAD) into the SSL pipeline. By refining the pseudo-labels through classifier bias correction during both training and testing, the proposed method mitigates the class imbalance and distribution mismatch—two critical challenges that hinder practical PD monitoring. The experimental results demonstrate that CDMAD-SSL achieves an overall classification accuracy of 96.64%, outperforming benchmark SSL methods by up to 2.52%, while maintaining robust precision, recall, and F1-scores across all fault types. Furthermore, the framework consistently improved the recognition of minority classes under skewed data distributions, thereby validating its effectiveness under realistic on-site conditions.