电力系统实测过电压信号的特征量提取与验证  

Features Extraction and Validation from Measured Overvoltage Signals of Power System

作　　者：黄艳玲[1] 司马文霞[1] 杨庆[1] 袁涛[1] 杨鸣[1] 

机构地区：[1]重庆大学输配电装备及系统安全与新技术国家重点实验室,重庆400030

出　　处：《高电压技术》2013年第1期60-66,共7页High Voltage Engineering

基　　金：国家重点基础研究发展计划(973计划)(2009CB724504)~~

摘　　要：提高实测过电压信号识别率的关键是提取的特征量具有区别性、稳定性和明确的物理意义。为此,提出一种新的提取实测过电压信号特征量的研究思路:首先结合过电压信号产生的物理过程和影响因素对其时域和频域特征进行分析,然后确定其具有区别性、稳定性和明确物理意义的特征,最后选择合适的信号处理技术提取相关的特征量。以电容器组合闸、感应雷、弧光接地3类常见的实测过电压信号为对象,依照所提出的研究思路分别提取了相应的特征量,并构造了基于支持向量机的分类器对这3类过电压的实测样本(共98个)进行了识别,识别率分别为100%、100%、97.3%,总识别率达99%。识别结果证明了对以上3类过电压信号所提取特征量的有效性,同时表明所提出的思路可供更多类型的实测过电压信号的特征量提取参考。To increase the identification rate of measured overvoltage signals,the extracted features must be distinctive,stable,and explicitly and physically meaningful.Consequently,we put forward a new research method to extract features from measured overvoltage signals.Firstly,we analyzed overvoltage signals in their time-domains and frequency-domains by taking the physical process and influential factors of the signals generation into consideration;secondly,we determined the characteristics to meet the requirements mentioned above;finally,we selected appropriate signal processing techniques to extract corresponding features.Taking three types of common overvoltage signals（caused by capacitor bank energization,inductive lightning and single phase arc to ground,respectively） as examples,we extracted their features according to the proposed method.Then 98 samples of the mentioned three types of signals were identified by binary classifiers constructed based on support vector machines.It is found that the identification rates of the three types of signals are 100%,100%,and 97.3%,respectively,which is 99% in total for all the samples.The identification results validate the effectiveness of the extracted features for the three types of measured overvoltage signals;meanwhile,the proposed method can be referred by feature extractions of more types of measured overvoltage signals.

关 键 词：过电压信号 特征量提取 电容器组合闸 感应雷 弧光接地 识别 

分 类 号：TM863[电气工程—高电压与绝缘技术]