触电信号暂态特征提取及故障类型识别方法  

Fault Transient Feature Extraction and Fault Type Identification for Electrical Shock Signals

作　　者：韩晓慧[1] 杜松怀[1] 苏娟[1] 刘官耕 

机构地区：[1]中国农业大学信息与电气工程学院,北京市海淀区100083

出　　处：《电网技术》2016年第11期3591-3596,共6页Power System Technology

基　　金：国家自然科学基金项目(51177165);国家电网公司科技项目(PDB5120152336)~~

摘　　要：针对传统剩余电流保护装置只能监测到总泄漏电流信号大小,但不能根据监测到的总泄漏电流信号自动分类和识别触电类型,提出了一种基于统计特征参数与支持向量机的触电信号暂态特征提取及故障类型识别的新方法。该方法首先针对3种生物体触电总泄漏电流故障信号提取出表征图谱特征的29个时域和频域统计特征参数,然后采用主成分分析法对特征空间进行降维处理得到5个新的特征量,将降维后的特征量输入到支持向量机中进行分析,利用粒子群算法优化支持向量机模型参数,并与GA-SVM,CV-SVM进行了对比分析。结果表明,PSO-SVM的识别率最高,能够有效地诊断触电信号故障类型。To overcome defects of traditional residual current operated protective device（RCD）, only detecting size of whole leakage current signal, but not classifying and identifying electric shock fault type automatically based on whole leakage current signal, a novel method was proposed based on statistical parameters and support vector machine（SVM） for fault type recognition of electrical shock signals. Firstly, 29 time and frequency statistical characteristic parameters were extracted from 3 kinds of whole biological electric shock leakage current signals. Then, principal component analysis（PCA） was applied to reduce feature space dimension and 5 new characteristic parameters were obtained. The characteristic parameters after dimension reduction were input to SVM classification. Particle swarm optimization（PSO） was taken to optimize SVM model parameters. Finally, GA-SVM and CA-SVM were adopted to perform comparison with PSO-SVM. Results show that PSO-SVM method recognition rate is the highest for effectively diagnosing electric shock fault type.

关 键 词：触电信号 统计参数 主成分分析 支持向量机 模式分类 

分 类 号：TM721[电气工程—电力系统及自动化]