基于多级二阶注意力孪生网络的小样本GIS局部放电诊断方法

doi:10.19595/j.cnki.1000-6753.tces.220104

Abstract
Figure/Table
References
Related Citation (4)

Download: PDF (3526 KB) HTML (1 KB)
Export: BibTeX | EndNote (RIS)

Abstract Monitoring and diagnosing partial discharge in gas-insulated switchgear (GIS) is an effective means to evaluate its insulation state. Deep learning methods have shown excellent performance in GIS partial discharge diagnosis in recent years because they can automatically learn the nonlinear relationship between GIS partial discharge signals and fault types. However, the existing deep learning methods are all trained on the premise of massive data, and the model's performance decreases severely when the amount of data decreases. Therefore, this paper proposes a novel Siamese network for diagnosing the few-shot partial discharge of GIS. The meta-training approach can achieve on-site small-sample GIS partial discharge (PD) diagnosis accurately and robustly.
Firstly, a multi-level second-order attention convolutional network is constructed to mine comprehensive feature information in partial discharge data and extract discriminative and representative features. Secondly, the metric learner is designed to learn the similarity of metric features. Then, the similarity metric compares or matches samples under the target task to achieve classification. Finally, a multi-level second-order attention Siamese network is designed for partial discharge diagnosis in GIS. Through the A-way K-shot training, the ability of model learning is endowed, and the learned meta-knowledge can be transferred to new tasks through fine-tuning, which solves the problem of small samples.
The experimental results show that the GIS partial discharge diagnosis accuracy of the method proposed in this paper reaches 93.76 % when each type of support sample is 5, which is 57.34 %, 45.53 %, 25.97 %, 65.91 %, and 33.97 % higher than direct training, FT, DATL, TCNN, and SVM, respectively. Simultaneously, box plots and scatter plots show that the method proposed in this paper has the most negligible dispersion, indicating that it is robust. Moreover, through the confusion matrix and t-distribution stochastic neighbor embedding, compared with traditional methods, this method proposed in this paper effectively solves the problem of small sample diagnosis and improves diagnostic performance under small samples. In addition, the proposed method has strong fault tolerance for unbalanced samples.
The following conclusions can be drawn: (1) The constructed Siamese network can realize high-precision and high-robust diagnosis of partial discharge in small-sample GIS on-site, especially for scenarios where UHV GIS samples are scarce. (2) The introduction of multi-scale convolution and second-order attention modules effectively improves the diagnostic accuracy of GIS partial discharge. (3) The training and testing of the method proposed in this paper are carried out under small samples, and the training is strictly in accordance with the A-way K-shot. K samples are randomly selected for each task from many training sets as the support set for model training, which has a strong tolerance for unbalanced samples.

Key words： Siamese network multi-scale convolution second-order attention module few-shot GIS partial discharge

Received: 19 January 2022

PACS:

TM835

	Service

	E-mail this article
	Add to my bookshelf
	Add to citation manager
	E-mail Alert
	RSS
	Articles by authors
	Wang Yanxin
	Yan Jing
	Wang Jianhua
	Geng Yingsan

Cite this article:

Wang Yanxin,Yan Jing,Wang Jianhua等. Few-Shot Partial Discharge Diagnosis for Gas-Insulated Switchgear Using a Novel Multi-Level Second-Order Attention Siamese Network[J]. Transactions of China Electrotechnical Society, 2023, 38(8): 2255-2264.

URL:

https://dgjsxb.ces-transaction.com/EN/10.19595/j.cnki.1000-6753.tces.220104 OR https://dgjsxb.ces-transaction.com/EN/Y2023/V38/I8/2255

[1] 刘青, 常丁戈, 邓军波. 用于变电站站域局部放电特高频测向的空间谱估计算法优化选择[J]. 电工技术学报, 2020, 35(16): 3551-3560.
Liu Qing, Chang Dingge, Deng Junbo.Optimal sele- ction on spatial spectrum estimation algorithms for UHF direction finding of partial discharge in sub- station[J]. Transactions of China Electrotechnical Society, 2020, 35(16): 3551-3560.
[2] 王艳新, 闫静, 王建华, 等. 基于域对抗迁移卷积神经网络的小样本GIS绝缘缺陷智能诊断方法[J]. 电工技术学报, 2022, 37(9): 2150-2160.
Wang Yanxin, Yan Jing, Wang Jianhua, et al.Intelligent diagnosis for GIS with small samples using a novel adversarial transfer learning in convolutional neural network[J]. Transactions of China Electro- technical Society, 2022, 37(9): 2150-2160.
[3] 李泽, 王辉, 钱勇, 等. 基于加速鲁棒特征的含噪局部放电模式识别[J]. 电工技术学报, 2022, 37(3): 775-785.
Li Ze, Wang Hui, Qian Yong, et al.Pattern recog- nition of partial discharge in the presence of noise based on speeded up robust features[J]. Transactions of China Electrotechnical Society, 2022, 37(3): 775-785.
[4] Karimi M, Majidi M, MirSaeedi H, et al. A novel application of deep belief networks in learning partial discharge patterns for classifying corona, surface, and internal discharges[J]. IEEE Transactions on Indu- strial Electronics, 2020, 67(4): 3277-3287.
[5] 王卓, 王玉静, 王庆岩, 等. 基于协同深度学习的二阶段绝缘子故障检测方法[J]. 电工技术学报, 2021, 36(17): 3594-3604.
Wang Zhuo, Wang Yujing, Wang Qingyan, et al.Two stage insulator fault detection method based on collaborative deep learning[J]. Transactions of China Electrotechnical Society, 2021, 36(17): 3594-3604.
[6] 臧海祥, 郭镜玮, 黄蔓云, 等. 基于深度迁移学习的时变拓扑下电力系统状态估计[J]. 电力系统自动化, 2021, 45(24): 49-56.
Zang Haixiang, Guo Jingwei, Huang Manyun, et al.State estimation for power systems with time-varying topology based on deep transfer learning[J]. Auto- mation of Electric Power Systems, 2021, 45(24): 49-56.
[7] Wang Yanxin, Yan Jing, Yang Zhou, et al.A domain adaptive deep transfer learning method for gas- insulated switchgear partial discharge diagnosis[J]. IEEE Transactions on Power Delivery, 2022, 37(4): 2514-2523.
[8] 李凡长, 刘洋, 吴鹏翔, 等. 元学习研究综述[J]. 计算机学报, 2021, 44(2): 422-446.
Li Fanzhang, Liu Yang, Wu Pengxiang, et al.A survey on recent advances in meta-learning[J]. Chinese Journal of Computers, 2021, 44(2): 422-446.
[9] 赵凯琳, 靳小龙, 王元卓. 小样本学习研究综述[J]. 软件学报, 2021, 32(2): 349-369.
Zhao Kailin, Jin Xiaolong, Wang Yuanzhuo.Survey on few-shot learning[J]. Journal of Software, 2021, 32(2): 349-369.
[10] Wang Duo, Zhang Ming, Xu Yuchun, et al.Metric- based meta-learning model for few-shot fault diagnosis under multiple limited data conditions[J]. Mechanical Systems and Signal Processing, 2021, 155: 107510.
[11] Zhang Jinglin, Wang Yanbo, Zhu Kai, et al.Diagnosis of interturn short-circuit faults in permanent magnet synchronous motors based on few-shot learning under a federated learning framework[J]. IEEE Transactions on Industrial Informatics, 2021, 17(12): 8495-8504.
[12] 朱瑞金, 郝东光, 胡石峰. 小样本条件下基于卷积孪生网络的变压器故障诊断[J]. 电力系统及其自动化学报, 2021, 33(1): 64-69, 84.
Zhu Ruijin, Hao Dongguang, Hu Shifeng.Trans- former fault diagnosis based on convolutional Siamese network with small samples[J]. Proceedings of the CSU-EPSA, 2021, 33(1): 64-69, 84.
[13] 高昂, 郑建勇, 梅飞, 等. 基于三元组孪生网络的窃电检测算法[J]. 中国电机工程学报, 2022, 42(11): 3975-3985.
Gao Ang, Zheng Jianyong, Mei Fei, et al.Electricity theft detection algorithm based on triplet network[J]. Proceedings of the CSEE, 2022, 42(11): 3975-3985.
[14] 高浩寒, 潮群, 徐孜, 等. 小样本下基于孪生神经网络的柱塞泵故障诊断[J]. 北京航空航天大学学报, 2023, 49(1): 155-164.
Gao Haohan, Chao Qun, Xu Zi, et al.Piston pump fault di-agnosis based on siamese neural network with small samples[J]. Journal of Beijing University of Aeronautics and Astronautic, 2023, 49(1): 155-164.
[15] 刘鑫, 周凯锐, 何玉琳, 等. 基于度量的小样本分类方法研究综述[J]. 模式识别与人工智能, 2021, 34(10): 909-923.
Liu Xin, Zhou Kairui, He Yulin, et al.Survey of metric-based few-shot classification[J]. Pattern Recognition and Artificial Intelligence, 2021, 34(10): 909-923.
[16] Bhukya A, Koley C.Bi-long short-term memory networks for radio frequency based arrival time detection of partial discharge signals[J]. IEEE Transactions on Power Delivery, 2022, 37(3): 2024-2031.
[17] Li Wenjing, Ren Tingting, Li Fang, et al.Contextual similarity-based multi- level second-order attention network for semi- supervised few-shot learning[J]. Neurocomputing, 2021, 461: 336-349.
[18] Li Zhuo, Luo Shaojuan, Chen Meiyun, et al.Infrared thermal imaging denoising method based on second- order channel attention mechanism[J]. Infrared Physics & Technology, 2021, 116: 103789.
[19] Zhu Yuanheng, Zhao Dongbin, He Haibo.Optimal feedback control of pedestrian flow in heterogeneous corridors[J]. IEEE Transactions on Automation Science and Engineering, 2021, 18(3): 1097-1108.
[20] Wang Yanxin, Yan Jing, Sun Qifeng, et al.A MobileNets convolutional neural network for GIS partial discharge pattern recognition in the ubiquitous power internet of things context: optimization, com- parison, and application[J]. IEEE Access, 7: 150226-150236.
[21] 杨为, 朱太云, 张国宝, 等. 电力物联网下基于卷积神经网络和迁移学习的GIS局部放电模式识别分类方法研究[J]. 高压电器, 2020, 56(9): 20-25, 32.
Yang Wei, Zhu Taiyun, Zhang Guobao, et al.Research on partial discharge pattern recognition and classification in GIS based on convolutional neural network and transfer learning in power internet of things[J]. High Voltage Apparatus, 2020, 56(9): 20-25, 32.
[22] 冯双, 陈佳宁, 汤奕, 等. 基于SPWVD图像和深度迁移学习的强迫振荡源定位方法[J]. 电力系统自动化, 2020, 44(17): 78-91.
Feng Shuang, Chen Jianing, Tang Yi, et al.Location method of forced oscillation source based on SPWVD image and deep transfer learning[J]. Automation of Electric Power Systems, 2020, 44(17): 78-91.
[23] 孙曙光, 李勤, 杜太行, 等. 基于一维卷积神经网络的低压万能式断路器附件故障诊断[J]. 电工技术学报, 2020, 35(12): 2562-2573.
Sun Shuguang, Li Qin, Du Taihang, et al.Fault diagnosis of accessories for the low voltage con- ventional circuit breaker based on one-dimensional convolutional neural network[J]. Transactions of China Electrotechnical Society, 2020, 35(12): 2562-2573.
[24] 范贤浩, 刘捷丰, 张镱议, 等. 融合频域介电谱及支持向量机的变压器油浸纸绝缘老化状态评估[J]. 电工技术学报, 2021, 36(10): 2161-2168.
Fan Xianhao, Liu Jiefeng, Zhang Yiyi, et al.Aging evaluation of transformer oil-immersed insulation combining frequency domain spectroscopy and support vector machine[J]. Transactions of China Electrotechnical Society, 2021, 36(10): 2161-2168.