基于生成对抗网络小样本扩展与多模态特征深度融合的变压器绕组故障诊断方法

邓祥力; 沈宇卿; 曾平; 鲍伟; 周德生; 熊小伏

引用本文:	邓祥力,沈宇卿,曾平,等.基于生成对抗网络小样本扩展与多模态特征深度融合的变压器绕组故障诊断方法[J].电力系统保护与控制,2026,54(09):89-101.
	DENG Xiangli,SHEN Yuqing,ZENG Ping,et al.Transformer winding fault diagnosis method based on small sample extension of generative adversarial network and deep multimodal feature fusion[J].Power System Protection and Control,2026,54(09):89-101

【打印本页】【下载PDF全文】【查看/发表评论】【EndNote】【RefMan】【BibTex】

←前一篇|后一篇→

过刊浏览高级检索

本文已被：浏览 232次下载 37次	码上扫一扫！
分享到：微信更多字体:加大+\|默认\|缩小-
基于生成对抗网络小样本扩展与多模态特征深度融合的变压器绕组故障诊断方法
邓祥力¹，沈宇卿¹，曾平²，鲍伟²，周德生²，熊小伏³
1. 上海电力大学电气工程学部，上海 200090；2. 国网上海市电力公司，上海 200122；3. 重庆大学电气工程学院，重庆 400044

摘要:

针对变压器绕组故障诊断中故障样本稀缺与多模态特征分散的问题，提出一种基于梯度惩罚Wasserstein条件生成对抗网络(Wasserstein conditional generative adversarial network with gradient penalty, WCGAN-GP)与多模态特征深度融合的故障诊断方法。首先，利用WCGAN-GP生成振动与超声故障样本，以缓解小样本与类别不平衡问题，并引入基于动态时间规整(dynamic time warping, DTW)的1-最近邻(1-nearest neighbor, 1NN)方法，对生成样本质量进行定量评估。随后，对不同模态数据采用差异化特征提取策略：漏磁信号输入一维改进残差网络提取局部动态特征；振动与超声信号转化为二维特征图并行输入二维改进稠密连接网络，以挖掘全局序列关系和时频特征。最后，通过分层跨模态交互机制(layer-wise cross-modal interaction, LCMI)实现三模态特征的深度融合。实验结果表明，该方法在小样本条件下能够有效提升诊断性能，为变压器绕组故障诊断提供了一种有效的解决方案。

关键词: 变压器绕组故障诊断多模态融合 WCGAN‑GP 残差网络稠密连接网络分层跨模态交互机制

DOI：10.19783/j.cnki.pspc.251078

分类号:

基金项目:国家自然科学基金项目资助(52277079)；国家电网有限公司科技项目资助(520940240037)

Transformer winding fault diagnosis method based on small sample extension of generative adversarial network and deep multimodal feature fusion

DENG Xiangli¹, SHEN Yuqing¹, ZENG Ping², BAO Wei², ZHOU Desheng², XIONG Xiaofu³

1. Faculty of Electrical Engineering, Shanghai University of Electric Power, Shanghai 200090, China; 2. State Grid Shanghai Electric Power Company, Shanghai 200122, China; 3. School of Electrical Engineering, Chongqing University, Chongqing 400044, China

Abstract:

To address the scarcity of fault samples and dispersed multimodal features in transformer winding fault diagnosis, this paper proposes a diagnostic method combining a Wasserstein conditional generative adversarial network with gradient penalty (WCGAN‑GP) and deep multimodal feature fusion. First, WCGAN‑GP is employed to generate vibration and ultrasonic fault samples, alleviating the problems of small sample size and class imbalance. A dynamic time warping (DTW)‑based 1‑nearest neighbor (1NN) is introduced to quantitatively evaluate the quality of the generated samples. Then, differentiated feature extraction is applied for different modalities. Leakage flux signals are fed into a one‑dimensional improved residual network to extract local dynamic features. Vibration and ultrasonic signals are transformed into two‑dimensional feature maps and input in parallel into a two‑dimensional improved dense network to extract global sequence relationships and time‑frequency features. Finally, a layer‑wise cross‑modal interaction (LCMI) mechanism is designed to achieve deep fusion of the three modalities. Experiments show that the method effectively improves diagnostic performance under small‑sample conditions, providing a reliable solution for transformer winding fault diagnosis.

Key words: transformer winding fault diagnosis multimodal fusion WCGAN‑GP residual network dense convolutional networks layer‑wise cross‑modal interaction mechanism

X关闭