Intelligent fault diagnosis methods, especially convolutional neural network (CNN), have made significant progress in gas-insulated switchgear (GIS) partial discharge (PD) diagnosis, which are attributable to two reasons: 1) the training and testing samples come from the identical distribution; 2) there are massive labeled data with PD information. However, owing to the specific operating conditions of GIS, collecting massive samples from the same distribution is difficult in field conditions. With the purpose of resolving the data dilemma of conventional diagnosis methods in the field, we propose a domain adaptive deep transfer learning (DADTL) CNN for small samples GIS PD diagnosis. First, we adopted a CNN to automatically extract transferable features from PD samples. Then, a multilayer DADTL is developed to reduce the marginal and conditional distribution of learned transferable features, and Sliced Wasserstein distance (SWD) is employed as a penalty to reduce the negative migration. The experimental verification was performed on experimental and on-site GIS PD datasets. The results show that DADTL CNN can effectively achieve a high-accuracy diagnosis of GIS PD for small samples. Compared with other methods, the DADTL CNN can achieve more accurate and robust GIS PD diagnosis for on-site small samples.

Data-driven artificial intelligence methods, especially convolutional neural networks (CNNs), have achieved excellent performance in high-voltage circuit breaker (HVCB) fault diagnosis. However, CNN relies heavily on massive data. When the amount of data decreases, the fault diagnosis performance drops severely. To settle these problems, a few-shot transfer learning (FSTL) with attention mechanism (AM) to realize the mechanical fault diagnosis of HVCBs is proposed. First, a one-dimensional CNN with AM is used to extract the fault features of HVCBs. The introduction of the AM makes CNN pay more attention to the interesting part of the fault signal to extract discriminative features. Then, domain adaptive transfer learning is used to realize a reliable diagnosis of HVCBs in small samples.The subdomain adaptation is adopted to adjust the distribution of related subdomains under the same category. The proposed subdomain adaptation can not only align the global distribution well but also effectively align the distribution of the same category of subdomains. Experimental results show that the FSTL proposed can achieve highly accurate and robust fault diagnosis of HVCBs with few-shot on-site. Compared with the traditional methods, the FSTL is obvious and provides a reliable reference for the diagnosis of HVCBs.

Convolutional neural networks (CNNs) have been widely used for gas‐insulated switchgear (GIS) partial discharge (PD) pattern recognition due to their powerful feature extraction ability. However, there is commonly a scarcity of fault samples due to low insulation failure rate of GIS equipment, which degrades the diagnostic performance of these CNN networks when directly applied to small and unbalanced datasets. Therefore, we propose a novel auxiliary classifier generative adversarial network for GIS PD pattern recognition for small and unbalanced samples. First, we propose using synchrosqueezed wavelet transform to extract time‐frequency characteristics of PD pulses and obtain a time‐frequency image with high energy aggregation and time‐frequency distribution rate. Then, we propose an improved generative adversarial network with an auxiliary classier and self‐attention mechanism, which can generate high‐quality PD samples for situations with few classes. Experiments show that our proposed method can reach 95.75% recognition accuracy for small datasets, which is the highest among several comparable methods. Furthermore, the proposed method has excellent and stable recognition performance for various unbalanced datasets.

A porcelain insulator is an important part to ensure that the insulation requirements of power equipment can be met. Under the influence of their structure, porcelain insulators are prone to mechanical damage and cracks, which will reduce their insulation performance. After a long-term operation, crack expansion will eventually lead to breakdown and safety hazards. Therefore, it is of great significance to detect insulator cracks to ensure the safe and reliable operation of a power grid. However, most traditional methods of insulator crack detection involve offline detection or contact measurement, which is not conducive to the online monitoring of equipment. Hyperspectral imaging technology is a noncontact detection technology containing three-dimensional (3D) spatial spectral information, whereby the data provide more information and the measuring method has a higher safety than electric detection methods. Therefore, a model of positioning and state classification of porcelain insulators based on hyperspectral technology is proposed. In this model, image data were used to extract edges to locate cracks, and spectral information was used to classify the surface states of porcelain insulators with EfficientNet. Lastly, crack extraction was realized, and the recognition accuracy of cracks and normal states was 96.9%. Through an analysis of the results, it is proven that the crack detection method of a porcelain insulator based on hyperspectral technology is an effective non-contact online monitoring approach, which has broad application prospects in the era of the Internet of Things with the rapid development of electric power.

Data-driven deep learning methods excel in gas insulated switchgear (GIS) partial discharge (PD) diagnosis, but their performance falters in data-scarce scenarios. Constrained by the intricate operating environment of GIS and the transient and stochastic nature of PD, acquiring sufficient PD samples in practical applications remains challenging. Therefore, this paper proposes an improved capsule generative adversarial network (GAN) for small-sample GIS PD diagnosis. Firstly, capsule network is introduced in both the generator and the discriminator to extract more comprehensive feature information from PD signals, thus avoiding the potential loss of crucial features due to pooling operations. Then, a dual-mode optimization method is used to update the model parameters of the generator, and sample quality and diversity are used as evaluation indicators to guide the training process of the generator, realizing automatic optimization of generator parameter-level model construction. Meanwhile, the discriminator stabilizes the GAN training by adopting the Wasserstein distance and incorporating a gradient penalty term. Experimental results show that the proposed method achieves diagnostic accuracy of 97.47% in small datasets, significantly outperforming other methods and exhibiting robust performance, provides a viable solution for high-precision and robust GIS PD diagnosis with limited samples.