摘要
变电站在线监测技术在"遥听"方面存在缺陷,提出了一种基于声音信号识别的电气主设备在线监控方法,通过安装在设备上的定向拾音器将设备声音转换为电信号后经网络录音盒送入服务器,服务器提取声音信号的MFCC参数后输入小波神经网络进行设备运行状态识别。介绍了监测系统的关键技术即声音信号特征参数提取和声音识别过程,以及现场提取的声音信号仿真分析。该技术能对设备实现在线监测且能很好地克服人工监听可靠性不高的缺点。
In view of the defects of current substation online monitoring technology in "Remote Hearing", this paper proposes an online monitoring method of main electrical equipment based on sound signal recognition, in which the device sound is transferred into electrical signals through the directional pickup installed on the device, and then transmitted into the monitoring server, and through server extraction of sound signal, the MFCC parameters are input into the wavelet neural network for equipment operation state identification. The system will pick up the audio signals in the early stage of the operation as a benchmark to establish a basic database, and the identification result in the later stage is compared with the database in the early stage. When the identification result prove a fault, an alarm will be issued and the fault sound will be classified and stored for subsequent analysis. This paper introduces the key technology of the monitoring system, that is, the extraction of sound signal feature parameters and the process of sound recognition. This technology can realize on-line monitoring of equipment and overcome the disadvantages of low reliability of manual monitoring.
作者
王林
扈海泽
方梦鸽
WANG Lin;HU Haize;FANG Mengge(State Grid Chenzhou Electric Power Supply Company,Chenzhou 423000,Hunan Province,China;Jishou University,Jishou 416106,Hunan Province,China;Changsha University of Science and Tech no logy,Changsha 410114,Hunan Provin ce,China)
出处
《电力与能源》
2019年第6期660-663,673,共5页
Power & Energy
基金
湖南省教育厅一般项目(18C0565)
吉首大学科研项目(Jdx17030)
