《计算机应用》唯一官方网站 ›› 2026, Vol. 46 ›› Issue (10): 3400-3406.DOI: 10.11772/j.issn.1001-9081.2025091114

• 前沿与综合应用 • 上一篇    

融合双阈值初筛与边级动态图卷积的故障诊断方法

闵万雄1, 严意成1, 王久银1, 梁禹宽1, 梁文杨1, 潘作舟2, 廖勇3()   

  1. 1.贵州乌江水电开发有限责任公司 构皮滩发电厂,贵州 遵义 564408
    2.中国计量大学 计量测试与仪器学院,杭州 310018
    3.重庆大学 微电子与通信工程学院,重庆 401331
  • 收稿日期:2025-09-25 修回日期:2025-10-17 接受日期:2025-10-20 发布日期:2025-11-12 出版日期:2026-10-10
  • 通讯作者: 廖勇
  • 作者简介:闵万雄(1988—),男,贵州遵义人,高级工程师,主要研究方向:电气一次、安全生产管理
    严意成(1994—),男,贵州印江人,工程师,主要研究方向:电气一次、生产管理
    王久银(1986—),男,贵州遵义人,工程师,主要研究方向:生产管理及其数字化
    梁禹宽(1998—),男,贵州贵阳人,助理工程师,主要研究方向:电气一次
    梁文杨(2001—),男,贵州遵义人,助理工程师,主要研究方向:电气一次
    潘作舟(1994—),男,安徽巢湖人,副教授,博士,主要研究方向:智能诊断
  • 基金资助:
    浙江省自然科学基金青年科学基金资助项目(LQN25E050004);中国华电集团科技项目(GPTDC/2024-1209)

Fault diagnosis method integrating dual-threshold preliminary screening and edge-level dynamic graph convolution

Wanxiong MIN1, Yicheng YAN1, Jiuyin WANG1, Yukuan LIANG1, Wenyang LIANG1, Zuozhou PAN2, Yong LIAO3()   

  1. 1.Goupitan Power Plant,Guizhou Wujiang Hydropower Development Company Limited,Zunyi Guizhou 564408,China
    2.College of Metrology Measurement and Instrument,China Jiliang University,Hangzhou Zhejiang 310018,China
    3.School of Microelectronics and Communication Engineering,Chongqing University,Chongqing 401331,China
  • Received:2025-09-25 Revised:2025-10-17 Accepted:2025-10-20 Online:2025-11-12 Published:2026-10-10
  • Contact: Yong LIAO
  • About author:MIN Wanxiong, born in 1988, senior engineer. His research interests include primary electrical equipment, safety production management.
    YAN Yicheng, born in 1994, engineer. His research interests include primary electrical equipment, production management.
    WANG Jiuyin, born in 1986, engineer. His research interests include production management and digitalization.
    LIANG Yukuan, born in 1998, assistant engineer. His research interests include primary electrical equipment.
    LIANG Wenyang, born in 2001, assistant engineer. His research interests include primary electrical equipment.
    PAN Zuozhou, born in 1994, Ph. D., associate professor. His research interests include intelligent diagnosis.
  • Supported by:
    Youth Program of Zhejiang Provincial Natural Science Foundation(LQN25E050004);China Huadian Group Technology Project(GPTDC/2024-1209)

摘要:

电力变压器在运行过程中会持续发出振动信号,其中包含大量由机械故障引起的脉冲和波动,这些信号是评估电力变压器运行状况的主要数据来源。为解决传统电力变压器故障诊断技术诊断精度低、实时性差且无法有效应对多故障场景的问题,通过分析电力变压器内部振动声信号的声纹特征,并基于声学特征构建一个融合双阈值初筛与边级动态图卷积的故障诊断方法。首先,提出一种利用声学短时能量和过零率双阈值评估电力变压器实时运行状况的正常/故障信号分类模块;然后,设计一种边级动态图卷积网络(ELDGCN),通过优化边的权重自适应地学习声纹信号之间的空间相关性。在真实的电力变压器场景下的实验结果表明,采用本文方法显著提升了故障的识别效率,数据集含混合故障的识别率达98.17%。

关键词: 电力变压器, 声学短时能量, 过零率, 故障检测, 声纹信号, 图卷积网络

Abstract:

During operation, power transformers emit vibration signals continuously, which contain many pulses and fluctuations caused by mechanical faults. These signals serve as the primary data source for assessing the operating condition of power transformers. To address the issues of low diagnostic accuracy, poor real-time performance, and inability to handle multi-fault scenarios effectively in the traditional power transformer fault diagnosis techniques, the voiceprint characteristics of internal vibration acoustic signals in power transformers were analyzed, and a fault diagnosis method integrating dual-threshold preliminary screening and edge-level dynamic graph convolution was constructed on the basis of acoustic features. First, a normal/fault signal classification module was proposed to assess the real-time operating condition of power transformers using acoustic short-time energy and zero-crossing rate. Second, an Edge-Level Dynamic Graph Convolutional Network (ELDGCN) was designed to learn spatial correlations among voiceprint signals adaptively by optimizing edge weights. Finally, experiments were conducted under real-world power transformer scenarios. The results show that the proposed method improves the fault identification efficiency significantly, with the identification rate for datasets containing mixed faults reaching 98.17%.

Key words: power transformer, acoustic short-time energy, zero-crossing rate, fault detection, voiceprint signal, Graph Convolutional Network (GCN)

中图分类号: