基于声纹识别系统的局部放电超声信号识别研究  

Study on Ultrasonic Signal Recognition of Partial Discharge Based on Voiceprint Recognition System

作　　者：周梦茜 唐志国[1] 王泽瑞 曹智 何宁辉 刘博 ZHOU Mengxi;TANG Zhiguo;WANG Zerui;CAO Zhi;HE Ninghui;LIU Bo(Beijing Key Laboratory of High Voltage&EMC,North China Electric Power University,Beijing 102206,China;Electric Power Research Institute of State Grid Ningxia Electric Power Co.,Ltd.,Yinchan 750001,China;State Grid Ningxia Electric Power Co.,Ltd.,Overhaul Company,Yinchan 750011,China)

机构地区：[1]华北电力大学高电压与电磁兼容北京市重点实验室,北京102206 [2]国网宁夏电力有限公司电力科学研究院,银川750001 [3]国网宁夏电力有限公司检修公司,银川750011

出　　处：《高压电器》2022年第9期127-133,共7页High Voltage Apparatus

摘　　要：为准确构建局部放电的超声信号和对应缺陷的关系,文中提出了一种基于声纹识别系统的对超声信号进行模式识别的超声识别系统。该系统首先对局部放电超声信号进行分帧和加窗处理,然后提取MFCC和GFCC特征向量,分别根据MFCC或GFCC建立GMM模型,最后利用极大似然估计对待识别样本进行识别。为验证该系统的有效性,文中设计了自由金属颗粒放电模型、悬浮放电模型和尖刺放电模型,并充入不同的绝缘气体,对不同缺陷的超声信号应用超声识别系统进行计算分析。研究结果表明,利用MFCC或GFCC特征向量建立的GMM模型具有代表性,基于改进声纹识别系统的局部放电超声识别系统对缺陷类型的识别结果符合预期,该系统可为超声信号的模式识别提供一种新方法,为电力设备故障检测和工况判断提供依据。In order to accurately establish the relationship between the ultrasonic signals of partial discharge and cor-responding defects,a kind of ultrasonic recognition system for pattern recognition of ultrasonic signals based on voice-print recognition system is proposed in this paper.The system first performs framing and windowing on the partial dis-charge ultrasonic signals,then extracts MFCC and GFCC feature vectors,sets up GMM models in accordance with MFCC or GFCC,and finally uses Maximum Likelihood Estimation to identify samples to be identified.In order to ver-ify the effectiveness of the system,a free metal particle discharge model,a floating discharge model and a pin-plate discharge model are designed,the different insulating gases are filled and the ultrasonic signals of different defects are calculated and analyzed by using the ultrasonic recognition system.The research results show that GMM model which is set up by the GFCC or MFCC feature vector are representative,and the identification result of defect type which is performed by the partial discharge ultrasonic recognition system based on voiceprint recognition system com-plies with the expectation.The system can provide a new method for pattern recognition of ultrasonic signals and pro-vide the basis for fault detection and condition judgment for electrical equipment.

关 键 词：局部放电 超声识别 MFCC GMM 

分 类 号：TM855[电气工程—高电压与绝缘技术]