电气火花电磁波特征及识别的研究

发布时间：2018-03-29 06:00

  本文选题：电气火花　切入点：小波分析　出处：《中国矿业大学》2014年硕士论文

【摘要】：随着现代工业及科学技术的迅猛发展，变配电系统的运行压力正逐渐体现出来，而在日益复杂化的庞大电力系统中，异常放电引起的电气火花是造成系统故障的关键原因。而电气火花产生的过程中伴随着高频脉冲电磁、光、热等现象，所以其破坏性很大，不但可以造成电路、电气设备和设施的损坏，也会引发火灾等对人员生命财产造成损失。据统计，我国的电气火花引起的火灾所占总火灾事故的比例逐年上升，一些重大的火灾事故都是由于电气火花导致。 本文首先分析了电气火花的发生机理、分类及特征，最终确定高压放电火花、低压放电火花及辉光放电火花为本文的研究对象。其次，搭建了三种类型放电火花的试验系统，并通过试验采集了这三种类型的火花波形，进一步确定提取5个频率上对应的幅值作为放电火花电磁波的特征参数，并用作后期的识别。考虑到采集的火花波形信号存在噪声信号的干扰影响，运用小波分析对其进行去噪、重构，滤除噪声干扰，并进行频谱分析，最终获取较为精确的特征参数信息。最后，，比较分析了BP神经网络及支持向量机SVM这两种识别方法的原理、分类规则；建立了火花识别模型，通过训练学习，验证了对电气火花电磁波的识别工作。 实验结果表明，在对25组待测火花波形进行的识别实验中，BP神经网络识别正确21组，错误4组，准确率84%，而SVM识别正确23组，错误2组，准确率为92%。并且SVM模型识别的时间明显小于BP神经网络，所以，本文最终确定SVM为电气放电火花电磁波特征识别的工具。
[Abstract]:With the rapid development of modern industry and science and technology, the operating pressure of substation and distribution system is gradually reflected, and in the increasingly complicated power system, The electrical spark caused by abnormal discharge is the key cause of system failure, and the electrical spark is accompanied by high frequency pulse electromagnetic, light, heat and other phenomena, so it is very destructive, not only can cause circuit, The damage of electrical equipment and facilities will also cause fire and other losses to people's lives and property. According to statistics, the proportion of fires caused by electrical sparks in China has increased year by year. Some major fire accidents are caused by electrical sparks. In this paper, the mechanism, classification and characteristics of electrical discharge are analyzed, and the high voltage discharge discharge, low voltage discharge discharge and glow discharge discharge are determined as the research objects. Secondly, three kinds of discharge test systems are built. The three types of sparking waveforms are collected through experiments, and the corresponding amplitudes on five frequencies are further determined as the characteristic parameters of EDM electromagnetic waves. Considering the influence of noise on the collected spark waveform signal, wavelet analysis is used to de-noise, reconstruct, filter the noise interference, and analyze the spectrum. Finally, the principle and classification rules of BP neural network and support vector machine SVM are compared and analyzed, and the spark recognition model is established. The recognition of electric spark electromagnetic wave is verified. The experimental results show that the BP neural network can recognize the correct 21 groups, the 4 wrong groups, the accuracy 84 in 25 groups of spark waveforms, while the SVM recognizes 23 groups correctly and 2 groups of errors. The accuracy is 92 and the time of SVM model recognition is obviously shorter than that of BP neural network. Therefore, this paper finally determines that SVM is the tool to identify the electromagnetic wave characteristics of EDM.
【学位授予单位】：中国矿业大学
【学位级别】：硕士
【学位授予年份】：2014
【分类号】：TM73

【参考文献】

中国期刊全文数据库 前10条

1 梁健宁;;某型号电缆故障定位系统的原理及应用[J];电工文摘;2010年02期

2 周一民;方敏;刘薛萍;;神经网络技术与小波分析相结合的故障诊断方法[J];矿业工程;2006年06期

3 薛雷;宋阳;郭丹丹;沈重;;基于LS-SVM和数学形态谱的局部放电模式识别[J];黑龙江电力;2011年01期

4 沈蓉;刘永红;蔡宝平;;电火花加工间隙的检测与识别方法[J];机床与液压;2009年02期

5 ;B-SPLINE-BASED SVM MODEL AND ITS APPLICATIONS TO OIL WATER-FLOODED STATUS IDENTIFICATION[J];Journal of Electronics(China);2007年02期

6 陈韬伟;朱明;陈振兴;;雷达辐射源信号小波变换特征提取方法[J];计算机工程与应用;2010年06期

7 胡蕾;马凤苹;林广旭;;连续采煤机使用过程中的电气故障分析及控制[J];煤矿机械;2011年02期

8 刘晓波,邢韧,兴自中;煤矿井下静电火花点燃的危险及预防[J];煤矿安全;2001年10期

9 许婕;童敏明;邓世健;唐守峰;郭秀兰;;基于小波分析的电气火花监测[J];矿山机械;2012年08期

10 梁伟华;;浅谈工业机械电气设备的电气故障分析方法[J];黑龙江科技信息;2013年05期

中国博士学位论文全文数据库 前2条

1 杨涛;基于小波分时尺度矩特征的多状态信息融合振动故障诊断和预测方法研究[D];华中科技大学;2009年

2 陈世海;冲击地压电磁辐射前兆信息识别技术研究[D];中国矿业大学;2012年

本文编号：1679839