基于聚类分析的超高压交流输电线路单永故障辨识研究

发布时间：2018-02-12 00:00

本文关键词： 单永故障辨识数据挖掘模式识别　出处：《昆明理工大学》2017年硕士论文　论文类型：学位论文

【摘要】：随着输电线路单永故障辨识研究的不断深入,国内外学者相继提出了一系列针对输电线路单永故障辨识的解决方案,输电线路单永故障辨识的理论体系日趋完善。单永故障辨识方法的研究以及相关技术的实用化,对提高输电线路传输效率,以至于整个电力系统的安全、可靠、稳定运行有重要意义。单永故障辨识方法的研究已持续很长一段时间,至今形成了单永故障辨识的两大主流方法,即根据故障断开相电弧过程电压、频率特性辨识单永故障和根据故障断开相恢复/残余电压过程电压特性差异辨识单永故障。但现有方法仍存在测量误差大、判别时间长、缺乏实测数据验证、缺乏普适性等欠缺,亟待进一步研究与验证。本文从电弧电压动态特性、故障断开相电压动态特性入手,从物理原理层面分析了自故障发生到自动重合闸装置动作这段时间内故障断开相电压特性,计算分析了并联电抗器接入及其参数的选择对故障断开相电压特性的影响。对电弧过程电压、恢复/残余电压过程断开相电压、电流波形特征进行分析,对比了单相瞬时性故障和单相永久性故障下不同故障过程电压波形的差异。以不同的故障过渡电阻、不同的故障位置发生单相瞬时性故障和单相永久性故障为条件进行仿真遍历,构建基于仿真数据的历史故障数据样本库;以幅值归一化后的恢复/残余电压波形作为特征量,提出基于层次聚类算法的输电线路单永故障辨识方法,方法以样本库为依托进行监督聚类,形成合并型聚类树,实现对单永故障的辨识。由于层次聚类算法能够突显聚类过程中数据间的相似度关系,该方法能够预估故障发生位置,确定故障发生区域。通过仿真测试样本验证,所提方法能够正确辨识单永故障,同时能够完成对故障发生区域的预估。以幅值归一化后的恢复/残余电压波形作为特征量,提出一种基于主成分分析-支持向量机(PCA-SVM)的单永故障辨识方法。利用主成分分析对测试样本进行降维,使其投影到基于样本库建立的特征空间,实现对测试样本的故障性质辨识;使用支持向量机构建最优超平面对特征空间进行划分,优化基于主成分分析方法的单永故障辨识结果。经仿真数据和实测数据验证,本方法不受故障条件影响,适应于带或不带并联电抗器的输电线路,以及不同电压等级的输电线路,是一种普适性的单永故障辨识方法。
[Abstract]:With the deepening of the research on single and permanent fault identification of transmission lines, scholars at home and abroad have put forward a series of solutions for single and permanent fault identification of transmission lines. The theoretical system of single and permanent fault identification of transmission lines is becoming more and more perfect. The research of single and permanent fault identification methods and the practical application of related technologies are helpful to improve transmission efficiency of transmission lines and even to ensure the safety and reliability of the whole power system. The study of single and permanent fault identification method has been going on for a long time, so far, two main methods of single forever fault identification have been formed, that is, according to the voltage of fault phase arc process, Single and permanent faults are identified by frequency characteristics and voltage characteristics of fault disconnected phase recovery / residual voltage process. However, the existing methods still have large measurement errors, long discriminant time and lack of verification of measured data. Due to the lack of universality and so on, it is urgent to further study and verify. This paper starts with the dynamic characteristics of arc voltage, the dynamic characteristics of fault disconnected phase voltage, and so on. From the aspect of physical principle, the characteristics of fault disconnected phase voltage in the period of time from the occurrence of fault to the action of automatic reclosing device are analyzed. The influence of the connection of shunt reactor and the selection of its parameters on the characteristics of the fault disconnected phase voltage is calculated and analyzed. The characteristics of the arc voltage, the recovery / residual voltage disconnect phase voltage and the current waveform are analyzed. The differences of voltage waveforms in different fault processes under single-phase transient fault and single-phase permanent fault are compared. Under the condition of single phase transient fault and single phase permanent fault occurring in different fault locations, the historical fault data sample database based on simulation data is constructed, and the amplitude normalized recovery / residual voltage waveform is taken as the characteristic value. This paper presents a single permanent fault identification method for transmission lines based on hierarchical clustering algorithm. The method is based on the sample base for supervised clustering to form a combined clustering tree. Because the hierarchical clustering algorithm can highlight the similarity relationship between the data in the clustering process, the method can predict the location of the fault and determine the fault occurrence area. The proposed method can correctly identify single and permanent faults, at the same time, it can complete the prediction of fault occurrence area. The amplitude normalized recovery / residual voltage waveform is used as the characteristic quantity. A single and permanent fault identification method based on principal component analysis-support vector machine (PCA-SVM) is proposed. The dimension of test samples is reduced by principal component analysis (PCA), which is projected to the feature space based on the sample base. The fault property identification of test samples is realized, the support vector mechanism is used to partition the feature space of optimal superplane face, and the results of single and permanent fault identification based on principal component analysis are optimized. The results are verified by simulation data and measured data. The method is suitable for transmission lines with or without parallel reactors and transmission lines with different voltage levels. It is a universal and single permanent fault identification method.
【学位授予单位】：昆明理工大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：TM755

【参考文献】