【作者】 刘倩；

【导师】 米建伟；

【作者基本信息】 西安电子科技大学 ， 机械电子工程， 2020， 硕士

【摘要】 电缆在生产制造及投运中,会出现故障造成突发性断电,给人们的生活乃至社会的经济带来不便与损失。调查发现电缆故障80%的概率是由于绝缘缺陷造成的,而局部放电信号的检测能够反映电缆的绝缘状态。常用的电缆局部放电检测方法有在线检测和离线检测两种,其中离线检测中的振荡波局部放电检测方法因与工频电压具有等效性、电气干扰小,得到了广泛应用。现有的振荡波局部放电检测技术主要解决局部放电源定位的问题,对局部放电类型识别的研究相对较少,且识别的准确率不高,不能给后续电缆修复工作提供有利支持。因此,本文深入研究振荡波检测系统(Oscillating Wave Test System,OWTS),仿真分析了电缆产生局部放电的原因,研究振荡波电压下的局部放电识别算法,以准确判断出电缆缺陷类型,提高后续电缆修复的效率,预防电缆事故的发生。首先,本文研究了电缆绝缘检测技术的国内外发展现状,阐明电力电缆产生局部放电的原因以及深层机理,分析了振荡波局部放电检测系统的工作原理。针对传统气隙型经典三电容模型未包含感应电荷概念的问题,提出了一种利用系统电容变化量和电缆结构电阻电容建立的改进气隙型局部放电模型。在SIMULINK中对振荡波局部放电信号检测系统搭建模型进行仿真,证明了所搭建模型的正确。其次,研究多种模式识别方法,针对支持向量机(Support Vector Machine,SVM)模型中正则因子和核函数参数对算法分类精度有显著影响,而现有寻参算法不能快速准确地寻找到最优值的问题,提出了改进DE-PSO优化算法,其利用粒子群优化(Particle Swarm Optimization,PSO)算法前期搜索能力强、收敛速度快的优点,结合差分进化(Differential Evolution,DE)算法具有从群体中筛选出更优个体的特性寻找参数最优解,以提高振荡波电压下的局部放电识别准确率,提升参数寻优效率。在MATLAB中对所提算法进行仿真验证,结果表明该算法能够快速有效的寻找到支持向量机的正则因子和核函数参数值,且该参数构建的支持向量机分类器识别准确率有所提高。最后,分析电缆中间接头缺陷产生的局部放电类型,并根据产生原因制作了四种电缆接头缺陷模型,搭建了振荡波局部放电检测系统的实验平台。针对振荡波系统下的局部放电类型多,传统模式识别建模复杂、识别准确率低的问题,提出了一种基于改进DE-PSO的M-ary多分类最小二乘支持向量机(Least Squares Support Vector Machine,LS-SVM)算法用以识别振荡波下局部放电类型,从而判断电缆绝缘状态。该算法采集四种类型的局部放电信号,根据各种类型信号之间的差异提取特征向量作为模式识别分类器的输入,基于M-ary多分类思想构建由两个分类器组成的局部放电模式识别机器,每个分类器选择最小二乘支持向量机作为模型,应用改进DE-PSO算法求解支持向量机最优参数,使用训练集数据训练机器以获得最优性能,最终达到能够通过任一振荡波下电缆局部放电信号判断出电缆缺陷类型,进而判断电缆绝缘状态的目的。通过现场采集数据进行验证,结果表明了所提算法能够对振荡波下局部放电信号进行类型识别,判断评估电缆绝缘状态,为后续电缆修复工作提供理论依据。更多还原

【Abstract】 During the production and commissioning of cables,there may be some safety hazards that cause sudden power outages,which will bring inconvenience and losses to people’s lives and even the society’s economy.Studies have found that 80% of cable damage is caused by cable insulation defects,and partial discharge signals can reflect the insulation status of cable.Commonly used partial discharge detection methods of cable are online detection and offline detection.Among them,the oscillating wave partial discharge test system(OWTS)in offline detection has been widely used because it is equivalent to the power frequency effect and has small electrical interference.The existing oscillating wave partial discharge detection technology mainly solves the problem of localization of the local discharge source.There is relatively little research on the identification of partial discharge types,and the recognition accuracy is not high,which cannot provide favorable support for subsequent cable repair work.Therefore,in this thesis,the OWTS is deeply studied,the reason for the partial discharge of cables is simulated and analyzed.Aiming at the current lack of pattern recognition of partial discharge signals under oscillating waves,a support vector machine(SVM)based power cable partial discharge pattern recognition algorithm was proposed.The main tasks of the thesis include the following parts:First of all,the development of cable condition detection technology at home and abroad is studied.The reasons and the underlying mechanism for the partial discharge of power cables is clarified.And the working principle of the OWTS is analyzed.Aiming at the problem that the traditional classic air-gap three-capacitance model does not include the concept of induced charge,an improved air-gap partial discharge model based on system capacitance variation and the resistance and capacitance of the cable structure is proposed.The mode of oscillating wave partial discharge test system is established in SIMULINK.According to this model,the oscillating wave high voltage and partial discharge signals were obtained in accordance with the theoretical values,which proved that the model was correct.Then,for the SVM model,the regular factors and kernel function parameters have a significant impact on the classification accuracy of the algorithm,and the existing parameterseeking algorithm cannot quickly and accurately find the optimal value.An improved DEPSO optimization algorithm is proposed.The algorithm uses the advantages of particle swarm optimization(PSO)early search ability and fast convergence speed,combined with Differential Evolution(DE)algorithm has the characteristics of screening out better individuals from the group,looking for the most optimal parameters to improves the accuracy of partial discharge recognition under the oscillating wave voltage.The proposed algorithm is simulated and verificated in MATLAB.The result shows that the algorithm can quickly and efficiently find the regularization factors and kernel function parameter values,and the recognition accuracy of the SVM classifier constructed by this parameter has been improved.Finally,the types of partial discharges caused by defects in the middle joint of the cable are analyzed,and four defect models of cable joints are made according to the causes.An experimental platform for an oscillation wave partial discharge detection system is built.Aiming at the problems of complex modeling and low recognition accuracy of traditional pattern recognition,a M-ary multi-class least squares support vector machine(LS-SVM)algorithm based on improved DE-PSO is proposed to identify partial discharges under oscillating waves type to judge the cable insulation status.The algorithm collects four types of partial discharge signals,and extracts feature vectors as the input of the pattern recognition classifier according to the differences between the various types of signals.Then,based on the M-ary multi-classification idea,a partial discharge pattern recognition machine composed of a least squares support vector machine as a sub-classifier is constructed.The improved DE-PSO algorithm is applied to solve the optimal parameters of the support vector machine,and the training data is used to train the machine to obtain the optimal performance.The algorithm is verified by building a field experiment platform.The results confirm that the proposed algorithm can identify the type of partial discharge signal under the oscillating wave,judge and evaluate the cable insulation status,and provide a theoretical basis for the subsequent cable repair work.更多还原