| 引用本文: | 刘元,苏盛,刘正谊,夏云峰,刘贯科,李彬.基于时、频域自相似性的配电变压器环境噪声识别方法[J].电力自动化设备,2022,42(3): |
| LIU Yuan,SU Sheng,LIU Zhengyi,XIA Yunfeng,LIU Guanke,LI Bin.Environmental noise recognition method for distribution transformer based on time domain and frequency domain self-similarity[J].Electric Power Automation Equipment,2022,42(3): |
|
| 摘要: |
| 配电变压器位于开放的嘈杂环境中，环境噪声对基于声音的设备运行状态在线监测具有突出影响。如何识别和剔除含噪声干扰的录音片段，是推进基于声音的配电变压器状态在线监测亟待解决的瓶颈问题。分析指出配电设备正常和异常的运行声音与各种环境噪声在时域和频域上的差异性特征，提出基于时、频域自相似性的配电设备录音数据中含环境噪声片段的识别方法。对分钟级录音数据分帧后提炼每帧的时域和频域特征指标，通过聚类判断该录音片段是否具有自相似性，识别和剔除含环境噪声的录音数据。实验结果表明，所提方法能够有效识别和剔除含环境噪声的录音片段，从而得到环境噪声含量较低的录音数据，为后续开展基于声音信号的配电变压器状态监测提供有力支撑。 |
| 关键词: 声音信号 状态监测 配电变压器 噪声消除 |
| DOI：10.16081/j.epae.202111021 |
| 分类号:TM41 |
| 基金项目:国家自然科学基金资助项目(51777015)；湖南省自然科学基金资助项目(2020JJ4611)；湖南省教育厅重点项目(19A011) |
|
| Environmental noise recognition method for distribution transformer based on time domain and frequency domain self-similarity |
|
LIU Yuan1, SU Sheng1, LIU Zhengyi2, XIA Yunfeng3, LIU Guanke3, LI Bin1
|
|
1.Disaster Prevention and Reduction Center of Renewable Energy System, Changsha University of Science and Technology, Changsha 410004, China;2.State Grid Changde Power Supply Company, Changde 415000, China;3.Dongguan Power Supply Bureau of Guangdong Power Grid Co.,Ltd.,Dongguan 523000, China
|
| Abstract: |
| The distribution transformer is located in an open noisy environment, and environmental noise has a prominent impact on the online monitoring of sound-based equipment operating status. How to identify and eliminate recordings containing noise is a bottleneck problem to be solved urgently in the promotion of sound-based online monitoring of the status of distribution transformer. The analysis points out the difference between the normal and abnormal operating sound of power distribution equipment and various kinds of environmental noises in time and frequency domains. The method for identifying environmental noise fragments in the recording data of power distribution equipment based on self-similarity in time and frequency domains is proposed. After the minute-level recording data is divided into frames, the time domain and frequency domain characteristic indexes of each frame are extracted. Whether the recording fragments have self-similarity is judged by clustering, and the recording data containing environmental noise are identified and eliminated. The experimental results show that the proposed method can effectively identify and eliminate the recording segments containing environmental noise, thereby recording data with lower environmental noise content can be obtained, and providing strong support for the subsequent development of sound signal-based status monitoring of distribution transformer. |
| Key words: sound signal status monitoring distribution transformer noise rejection |
