Page 123 - 《应用声学》2024年第1期
P. 123
第 43 卷 第 1 期 Vol. 43, No. 1
2024 年 1 月 Journal of Applied Acoustics January, 2024
⋄ 研究报告 ⋄
声纹信号-图形差分场增强和多头自注意力机制的
变压器工作状态辨识方法 ∗
张 寒 1,2 熊 云 1,2 唐 信 1,2 王 枭 3†
(1 国网湖南超高压变电公司 长沙 410000)
(2 变电智能运检国网湖南省电力有限公司实验室 长沙 410000)
(3 上海睿深电子科技有限公司 上海 200237)
摘要:为提升电力变压器工作状态的智能监测水平,提出声纹信号 -图形差分场增强和多头自注意力机制的变
压器工作状态辨识方法。基于图形差分场技术将声纹信号映射为二维图像,再借助多头注意力机制的视觉转
换器实现图像信息的深层挖掘与状态辨识,采用梯度加权类激活映射实现分类结果的可解释性分析。搭建了
包含变压器 4 种典型工作状态下的实验模拟测试系统平台,实验结果表明:所提方法不仅能够有效表征变压器
声纹信号的状态特征,且分类辨识精度相较于 “时频图 + 引入多头注意力机制的变换网络” 与 “图形差分场 +
引入残差模块的卷积神经网络” 的常规方法有显著提升,提升约 6%,同时也具备较好的鲁棒性,可为电气设备
的故障检测研究提供一定参考。
关键词:图形差分场;多头自注意力机制;变压器;状态辨识
中图法分类号: TM595 文献标识码: A 文章编号: 1000-310X(2024)01-0119-12
DOI: 10.11684/j.issn.1000-310X.2024.01.015
Transformer working state identification method based on voiceprint signal-motif
difference field enhancement and multi-head self-attention mechanism
ZHANG Han 1,2 XIONG Yun 1,2 TANG Xin 1,2 WANG Xiao 3
(1 State Grid Hunan Extra High Voltage Substation Company, Changsha 410000, China)
(2 Substation Intelligent Operation and Inspection Laboratory of State Grid Hunan Electric Power Co., Ltd.,
Changsha 410000, China)
(3 Shanghai Rhythm Electronic Technology Co., Ltd., Shanghai 200237, China)
Abstract: In order to improve the intelligent monitoring level of power transformer working state, a method of
transformer working state identification based on motif difference field (MDF) voiceprint signal enhancement
and multi-head self-attention mechanism is proposed. Based on the MDF technology, the sound signal is
mapped into a two-dimensional image, and then the depth mining and state recognition of image information
are realized with the help of the visual converter of multi-head attention mechanism, and the explanatory
power analysis of classification results is realized by using gradient weighted class activation mapping. The
experimental simulation test system platform containing four typical operating states of the transformer is
constructed, and the experimental results show that the proposed method not only can effectively characterize
the state characteristics of the transformer acoustic signal, but also has a higher classification accuracy compared
2023-07-24 收稿; 2023-10-14 定稿
国网湖南省电力有限公司科技项目 (5216A3210014)
∗
作者简介: 张寒 (1975– ), 男, 湖南长沙人, 博士, 高级工程师, 研究方向: 高压试验及带电检测技术。
† 通信作者 E-mail: wangxiao_2356@163.com