基于声音特征的变电站电力变压器故障检测

发布时间：2019-04-19 15:22

【摘要】：变电站电力变压器的故障检测技术,是指通过监控变压器运行状态来检测变电站电力变压器是否仍处于正常工作中,若发生故障,能够做到及时报警,方便工作人员对其进行检测维修,同时还可以预测变压器未来一段时间内的工作情况。传统的变压器故障检测方法都是依靠人工完成的,包括状态检修、定期检修等,这些方法均需要相关工作人员定期在现场进行操作,对工作人员的技术水平以及工作经验都有要求,但是传统检测方法不但缺乏时效性,而且具有一定的安全风险。随着工业发展、科技进步、人们生活水平的不断提高,随之而来的是用电需求大幅增长、电网规模变大大,传统的检测手段已无法满足日益变化的需求,此时,伴随着计算机与电子技术的进步,实时的在线故障检测技术已逐步发展起来。本文提出的变电站电力变压器故障检测方案是通过分析、提取变压器所发声音的幅频特征,并结合相应的检测算法来达到变压器故障检测检测目的。基于声音信号的故障检测技术至今已发展了几十年,六十年代主要应用在核电、航空航天等高尖行业;七十年代发展至船舶、石化、冶金等行业;八十年代起逐步向各行业迅速扩展。在对变压器进行检测维修时,相关的工作人员可以凭借所听到的变压器运行声,进而知晓变压器是否发生了故障,本论文就是依据这种维修人员依靠声音判断变压器运行状态的方式,提出了一种模拟人类听觉系统的变压器故障检测方案,本方案可作为变压器故障检测的有效辅助手段。在本方案中,首先对变压器工作时发出的声音进行采集,然后通过检测系统对这些声音进行判断,进而达到对变压器运行情况检测的目的。在实验过程中,本实验组成员通过实地采集的方式对变压器所发出的各种声音进行收集,并建立相应的声音样本库。通过对这些声音样本进行分析、研究、统计,设计相应的算法,进而构建出变电站电力变压器的故障检测系统。本论文介绍了一种基于变压器声音数据幅频特性的特征提取的方法,并将所提取的每个声音数据特征组成相应的一维向量或二维矩阵,随后应用主分量分析(PCA)和二维主分量分析(2DPCA)算法分别对声音数据的频谱特征进行降维处理,提取主要特征信息,然后应用支持向量机(SVM)算法对声音信号进行分类,从而达到对变电站电力变压器状态检测的目的。
[Abstract]:The fault detection technology of substation power transformer means that it can detect whether substation power transformer is still in normal work by monitoring the operation status of transformer, and in case of failure, it can make timely alarm. It is convenient for the staff to test and repair the transformer, and it can also predict the future work of the transformer in a period of time. The traditional methods of transformer fault detection are completed manually, including condition-based maintenance, regular maintenance and so on. These methods require the relevant staff to operate regularly on the spot. However, traditional detection methods are not only lack of timeliness, but also have certain safety risks. With the development of industry, the progress of science and technology, and the continuous improvement of people's living standards, the demand for electricity and the scale of the power grid have been greatly increased, and the traditional means of detection have been unable to meet the ever-changing needs, at this time, With the progress of computer and electronic technology, real-time on-line fault detection technology has been gradually developed. The fault detection scheme proposed in this paper is to extract the amplitude-frequency characteristics of the sound generated by the transformer and combine with the corresponding detection algorithm to achieve the purpose of transformer fault detection by analyzing and extracting the amplitude-frequency characteristics of the sound produced by the transformer. The fault detection technology based on sound signal has been developed for decades, mainly used in nuclear power, aerospace and other advanced industries in the 1960s, in the seventies to ship, petrochemical, metallurgy and other industries; Since the 1980s, it has gradually expanded rapidly to various industries. During the inspection and maintenance of the transformer, the relevant staff member can rely on the sound heard in the operation of the transformer, and then know whether the transformer has broken down or not. In this paper, a transformer fault detection scheme simulating human auditory system is proposed according to the way that the maintenance personnel rely on sound to judge the transformer operation state. This scheme can be used as an effective auxiliary method for transformer fault detection. In this scheme, firstly, the sound emitted by the transformer is collected, then these sounds are judged by the detection system, and then the purpose of detecting the transformer operation is achieved. During the experiment, all kinds of sound emitted by the transformer were collected by the members of the experimental group, and the corresponding sound sample database was set up. Through the analysis, research, statistics and design of the corresponding algorithm, the fault detection system of substation power transformer is constructed. In this paper, a method of feature extraction based on the amplitude-frequency characteristics of transformer sound data is introduced, and each feature of the extracted sound data is composed of the corresponding one-dimensional vector or two-dimensional matrix. Then, principal component analysis (PCA) and two-dimensional principal component analysis (2DPCA) algorithm should be used to reduce the dimension of the spectral features of the sound data, extract the main feature information, and then use the support vector machine (SVM) algorithm to classify the sound signals. In order to achieve the substation power transformer status detection purpose.
【学位授予单位】：山东大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：TM63;TM41

【参考文献】