Fault Detection in Oil-Immersed Power Transformers Based on Generalizable Graph Knowledge Distillation

doi:10.19678/j.issn.1000-3428.0070580

Computer Engineering ›› 2026, Vol. 52 ›› Issue (9): 449-456. doi: 10.19678/j.issn.1000-3428.0070580

• Interdisciplinary Integration and Engineering Applications • Previous Articles Next Articles

Fault Detection in Oil-Immersed Power Transformers Based on Generalizable Graph Knowledge Distillation

ZHANG Ruijia¹, MA Huifang¹^,*(), ZHANG Yingyue¹, PENG Shengjiang²

1. College of Computer Science and Engineering, Northwest Normal University, Lanzhou 730070, Gansu, China
2. State Grid Gansu Electric Power Company Wuwei Power Supply Company, Wuwei 733000, Gansu, China

Received:2024-11-05 Revised:2025-03-03 Online:2026-09-15 Published:2025-04-21
Contact: MA Huifang

基于可泛化图知识蒸馏的油浸式电力变压器故障检测

张瑞佳¹, 马慧芳¹^,*(), 张映月¹, 彭生江²

1. 西北师范大学计算机科学与工程学院, 甘肃兰州 730070
2. 国网甘肃省电力公司武威供电公司, 甘肃武威 733000

通讯作者: 马慧芳
作者简介:
张瑞佳, 女, 硕士研究生, 主研方向为电力变压器故障诊断
马慧芳(CCF会员、通信作者)，教授、博士
张映月, 硕士研究生
彭生江, 教授级高级工程师、博士
基金资助:
甘肃省重点基础研究项目(24JRRA123); 甘肃省产业支撑项目(2022CYZC11)

Abstract

Abstract:

Dissolved Gas Analysis (DGA) aims to identify potential fault types by monitoring the dissolved gases in insulating oil. However, existing DGA methods exhibit limited performance because of the constraints imposed by the scarcity of labeled data. To address this issue, a novel Graph Knowledge Distillation method (GKDG) is proposed to enhance the accuracy and efficiency of the DGA. It employs a dual-view graph construction strategy to obtain additional supervision from sample neighborhoods, aggregating information directly from other samples through propagation. Furthermore, knowledge from the teacher Graph Neural Network (GNN) is distilled into the student GNN model, ensuring that the student model can effectively capture and interpret the complex relationships among the dissolved gases. Additionally, to align the student and teacher graphs in the embedding space, multiple types of knowledge are introduced, thereby enhancing the learning capability of the student model and enabling it to learn better from the teacher model. The experimental results validate the effectiveness of the GKDG in improving the DGA performance, providing strong support for the maintenance and fault detection of power equipment.

Key words: oil-immersed power transformers, fault detection, Dissolved Gas Analysis (DGA), Graph Neural Network (GNN), knowledge distillation

摘要：

溶解气体分析(DGA)旨在通过监测绝缘油中的溶解气体来识别潜在的故障类型。然而, 现有DGA方法受到有限标记数据制约导致性能不佳。为此, 提出一种新的图知识蒸馏方法(GKDG), 旨在提高DGA的准确性和效率。采用双视角图构建策略从样本邻域中获得额外的监督, 通过传播直接从其他样本中聚合信息。进一步地, 将教师图神经网络(GNN)中的知识蒸馏到学生GNN模型中, 确保学生模型能够有效地捕捉并解释溶解气体之间的复杂关系。此外, 为了对齐嵌入空间中的学生图和教师图, 引入多种知识, 从而增强学生模型的学习能力, 使其更好地学习教师模型。实验结果验证了GKDG在提升DGA性能方面的显著效果, 其能为电力设备的维护和故障检测提供有力支持。

关键词: 油浸式电力变压器, 故障检测, 溶解气体分析, 图神经网络, 知识蒸馏

ZHANG Ruijia, MA Huifang, ZHANG Yingyue, PENG Shengjiang. Fault Detection in Oil-Immersed Power Transformers Based on Generalizable Graph Knowledge Distillation[J]. Computer Engineering, 2026, 52(9): 449-456.

张瑞佳, 马慧芳, 张映月, 彭生江. 基于可泛化图知识蒸馏的油浸式电力变压器故障检测[J]. 计算机工程, 2026, 52(9): 449-456.

/ Recommend / Download Citations

URL: https://www.ecice06.com/EN/10.19678/j.issn.1000-3428.0070580

https://www.ecice06.com/EN/Y2026/V52/I9/449

Figures/Tables 8

Fig.1 The overall framework of GKDG

Fig.2 Performance of GKDG and baseline methods on different datasets

Fig.3 Performance of GKDG under different values of k

Fig.4 Performance of GKDG under multiple knowledge distillation weights

References 25

1	LIAO W , YANG D , WANG Y , et al. Fault diagnosis of power transformers using graph convolutional network. CSEE Journal of Power and Energy Systems, 2020, 7 (2): 241- 249.
2	罗文萱. 基于胶囊神经网络的电力变压器故障诊断方法研究. 高压电器, 2024, 60 (5): 92- 98.
	LUO W X . Research on fault diagnosis method of power transformer based on capsule networks. High Voltage Apparatus, 2024, 60 (5): 92- 98.
3	CHEN H C , ZHANG Y . Rethinking shallow and deep learnings for transformer dissolved gas analysis: a review. IEEE Transactions on Dielectrics and Electrical Insulation, 2025, 32 (1): 3- 10. doi: 10.1109/TDEI.2025.3526080
4	PATIL M , PARAMANE A , DAS S , et al. Hybrid algorithm for dynamic fault prediction of HVDC converter transformer using DGA data. IEEE Transactions on Dielectrics and Electrical Insulation, 2024, 31 (4): 2128- 2135. doi: 10.1109/TDEI.2024.3379954
5	毛业栋, 张春辉, 陈杰. 融合特征分析及机器学习的可演进变压器故障诊断模型. 计算机工程, 2024, 50 (8): 379- 388. doi: 10.19678/j.issn.1000-3428.0068224
	MAO Y D , ZHANG C H , CHEN J . Evolvable transformer fault diagnosis model combining feature analysis and machine learning. Computer Engineering, 2024, 50 (8): 379- 388. doi: 10.19678/j.issn.1000-3428.0068224
6	ZHANG X , YANG K Y . Transformer fault diagnosis method based on MTF and GhostNet. Measurement, 2025, 249, 117056. doi: 10.1016/j.measurement.2025.117056
7	GOUDA O E , ELHOSHY S H , TAMALY H H E L . Condition assessment of power transformers based on dissolved gas analysis. IET Generation, Transmission & Distribution, 2019, 13 (12): 2299- 2310.
8	王宇, 祁琦, 王纯, 等. 储能变流器信号高精度故障诊断方法. 计算机工程, 2024, 50 (8): 389- 396. doi: 10.19678/j.issn.1000-3428.0068520
	WANG Y , QI Q , WANG C , et al. High-precision fault diagnosis method for energy storage inverter signals. Computer Engineering, 2024, 50 (8): 389- 396. doi: 10.19678/j.issn.1000-3428.0068520
9	CHEN H C , ZHANG Y . Dissolved gas analysis using knowledge-filtered oversampling-based diverse stack learning. IEEE Transactions on Instrumentation and Measurement, 2025, 74, 2505211.
10	THOTE P B , DAIGAVANE M B , DAIGAVANE P M , et al. An intelligent hybrid approach using KNN-GA to enhance the performance of digital protection transformer scheme. Canadian Journal of Electrical and Computer Engineering, 2017, 40 (3): 151- 161. doi: 10.1109/CJECE.2016.2631474
11	MENEZES A G C , ARAUJO M M , ALMEIDA O M , et al. Induction of decision trees to diagnose incipient faults in power transformers. IEEE Transactions on Dielectrics and Electrical Insulation, 2022, 29 (1): 279- 286. doi: 10.1109/TDEI.2022.3148453
12	LI K, YANG G Q, WANG K, et al. Fault diagnosis method of transformer based on WSO and SVM[C]//Proceedings of the 7th IEEE Conference on Energy Internet and Energy System Integration (EI2). Washington D.C., USA: IEEE Press, 2024: 3732-3736.
13	JIN Y S , WU H , ZHENG J F , et al. Power transformer fault diagnosis based on improved BP neural network. Electronics, 2023, 12 (16): 3526. doi: 10.3390/electronics12163526
14	SHU K , MA H F , YANG J P , et al. GraphSmin: imbalanced dissolved gas analysis with contrastive dual-channel graph filters. Advanced Engineering Informatics, 2024, 62, 102839. doi: 10.1016/j.aei.2024.102839
15	ZHANG Y Y, MA H F, ZHANG D, et al. Graph contrastive learning for dissolved gas analysis[EB/OL]. [2024-10-05]. https://link.springer.com/chapter/10.1007/978-981-96-0821-8_12.
16	JIN W Y , MA H F , ZHANG Y Y , et al. Multi-view discriminative edge heterophily contrastive learning network for attributed graph anomaly detection. Expert Systems with Applications, 2024, 255, 124460. doi: 10.1016/j.eswa.2024.124460
17	TIAN Y J , PEI S C , ZHANG X L , et al. Knowledge distillation on graphs: a survey. ACM Computing Surveys, 2025, 57 (8): 1- 16.
18	HINTON G, VINYALS O, DEAN J. Distilling the knowledge in a neural network[EB/OL]. [2024-10-05]. https://arxiv.org/abs/1503.02531.
19	LIU J J , KE W J , WANG P , et al. Towards continual knowledge graph embedding via incremental distillation. Proceedings of the AAAI Conference on Artificial Intelligence, 2024, 38 (8): 8759- 8768. doi: 10.1609/aaai.v38i8.28722
20	SUN W J, CHEN D F, LÜ S W, et al. Knowledge distillation with refined logits[EB/OL]. [2024-10-05]. https://arxiv.org/abs/2408.07703.
21	WANG Q, ZHOU J J. Multi-perspective contrastive logit distillation[EB/OL]. [2024-10-05]. https://arxiv.org/abs/2411.10693.
22	TIAN Y J , XU S K , LI M Y . Decoupled graph knowledge distillation: a general logits-based method for learning MLPs on graphs. Neural Networks, 2024, 179, 106567. doi: 10.1016/j.neunet.2024.106567
23	GUO Z , WANG D , HE Q , et al. Leveraging logit uncertainty for better knowledge distillation. Scientific Reports, 2024, 14, 31249. doi: 10.1038/s41598-024-82647-6
24	XU L X , WANG Z W , BAI L , et al. Multi-level knowledge distillation with positional encoding enhancement. Pattern Recognition, 2025, 163, 111458. doi: 10.1016/j.patcog.2025.111458
25	MA Y C, CHEN Y B, AKATA Z. Distilling knowledge from self-supervised teacher by embedding graph alignment[EB/OL]. [2024-10-05]. https://arxiv.org/abs/2211.13264.

[1]	HU Caifu, WEI Bo, REN Ruibin. Network Traffic Classification Based on Incremental Learning from the Perspective of Space Optimization [J]. Computer Engineering, 2026, 52(8): 138-148.
[2]	MENG Yun, WEI Fan, LIU Xinyi, WANG Wei, CHAI Jiahui. Graph-DRL-Based Spectrum Resource Allocation in Air-Ground Cooperative UDN [J]. Computer Engineering, 2026, 52(8): 283-292.
[3]	FAN Xinggang, SHI Xuegang, LIAO Siteng, ZHAO Yiyi, LIANG Yuzhu, WANG Tian. Towards Efficient AI: A Survey of Model Compression Techniques on the Edge [J]. Computer Engineering, 2026, 52(7): 1-21.
[4]	ZHOU Lijun, ZHANG Junran, WANG Kaiyuan, XIANG Junlian. Drug Recommendation Based on Graph-Enhanced Attention Integrating Clinical Features of Patients [J]. Computer Engineering, 2026, 52(6): 314-325.
[5]	XIANG Yilong, NI Nan, ZOU Yuechao, YUAN Tong, LI Bohan, WANG Tiexin. Multi-Indicator Sinking Attitude Prediction of Open Caisson Based on Graph Neural Network [J]. Computer Engineering, 2026, 52(6): 307-313.
[6]	LUO Heng, WAN Liang. Network Traffic Intrusion Detection Method Based on Dynamic Spatio-Temporal Graph Neural Network [J]. Computer Engineering, 2026, 52(6): 202-213.
[7]	CUI Shaoguo, XU Song, WANG Mingyang, ZHOU Yue. Research Progress on Deep Learning Knowledge Tracing for Intelligent Education [J]. Computer Engineering, 2026, 52(4): 39-61.
[8]	ZHANG Zhen, YOU Lan, PENG Qingxi, JIN Hong, ZENG Haoqiu, XIA Yuchun. XSGCL: A Lightweight Graph Contrastive Learning Framework for Recommendation [J]. Computer Engineering, 2026, 52(4): 163-175.
[9]	WEI Haodong, WAN Liang. Android Malware Detection Method Based on Static Feature Combination in Graph Neural Networks [J]. Computer Engineering, 2026, 52(3): 190-200.
[10]	XUE Yang, QIN Yao, ZHANG Shuxiang. Collaborative Recommendation Based on Graph Neural Network Clustering Subgraphs Using Dual Graph Attention Network [J]. Computer Engineering, 2026, 52(2): 89-100.
[11]	XU Xiaoyang, WEI Wei, GAO Chongyang. Infrared Ship Target Detection Based on Improved YOLOv7-tiny [J]. Computer Engineering, 2026, 52(2): 209-220.
[12]	RAN Tongyu, TANG Mengzi, XIE Qing, LIU Yongjian. Semi-Supervised Ordinal Classification Framework Based on Information Fusion [J]. Computer Engineering, 2026, 52(2): 287-298.
[13]	JIA Jianghao, ZHANG Ziwei, GAO Liting, WEN Juan, XUE Yiming. Text Steganalysis Method Based on Hierarchy-Aware Matching [J]. Computer Engineering, 2026, 52(2): 245-252.
[14]	GUO Tiansheng, XIE Jinkui. Adaptive Adjustment Graph Augmentation and Representation Structures for Recommendation Model [J]. Computer Engineering, 2026, 52(2): 69-78.
[15]	LI Zhiren, ZHENG Weiguo. Link Prediction Based on Simple Path Graphs [J]. Computer Engineering, 2026, 52(1): 95-104.

Please choose a citation manager

Content to export