【作者】 董莉娜；

【导师】 卢继平；

【作者基本信息】 重庆大学 ， 电气工程， 2009， 硕士

【摘要】 由于10kV配网发生单相接地故障时的特征信息不明显,目前常采用稳态电流特征进行选线会产生误选和拒选,而应用故障线路暂态信息进行选线是目前研究的热点。本文在总结国内外10kV配网单相接地故障选线技术的基础上,分析了10kV配网发生单相接地故障时系统暂态零序电流和电压的变化规律,利用Matlab仿真平台并根据典型实小波和复小波变换的特征,分析和比较了实小波和复小波变换及其在单相接地故障选线中的应用,提出可根据各馈线零序电流暂态变化规律及小波多尺度分解的尺度系数的特征值进行选线。针对馈线数目少的10kV系统选线困难的问题,利用Matlab平台对4回馈线及其以下10kV系统的单相接地故障进行了仿真,采用Db N系列、Sym N系列、Bior Nr.Nd系列和Coif N系列共50种实小波对故障系统暂态零序电流进行小波变换,分析了零序电流小波变换后的特征,首次提出了暂态零序电流低频能量选线方法,并与目前采用的模极大值法和高频能量法进行了比较,结果表明:对于四回线及以下10kV系统,采用模极大值和高频能量法选线时,只有Sym 3、Coif 1、Bior 2.4、Bior 2.6和Bior 4.4等5个小波可实现正确选线;而采用本文提出的低频能量法进行选线时,不仅不受小波基选择的限制,且不受系统回路数的限制,对不接地系统和经消弧线圈接地系统均可实现正确选线。因此,本文提出了10kV系统发生单相接地故障时实小波选线可采用低频能量法。针对实小波选线存在的问题,采用Morlet 1-0.1、Cgaussian 1和Shannon 1-1等3种复小波对5回馈线10kV系统单相接地故障的暂态零序电压和各线路的暂态零序电流进行了变换,根据复小波变换具有的幅值和相位特征信息,提出了复小波变换的零序暂态电流功率、零序暂态电流能量和零序暂态电流幅值范数的故障选线方法,比较了不同选线判据和不同复小波选线的差异,提出了表征选线方法优劣的显著性指标,结果表明:采用零序暂态电流功率、零序暂态电流能量和零序暂态电流幅值范数进行选线,反映的故障特征突出和明显,具有综合特征;比较而言,采用Morelet复小波对10kV系统进行故障选线具有准确率高、选线可靠且易实现、特征值的显著性指标高等特点,且不受采样频率(3~100kHz)、过渡电阻、故障点位置和线路长度的限制,特别是对二回馈线谐振接地系统单相接地故障选线的显著性系数远高于实小波;而采用Cgaussian复小波和Shannon复小波虽也能实现正确选线,但因其边界效应明显,使信号边界部分严重失真,使用存在局限性。因此,复小波变换不仅可应用于10kV单相接地故障选线,而且对于馈线数少的系统选线具有优势。更多还原

【Abstract】 Because the characteristic signal of single-phase-grounded fault of 10kV distribution network system is not obvious, the fault detection based on the steady current of the faulted line might be not correct. Therefore, the fault detection based on the transient signal of the fault distribution network has become one of the key techiniques.Based on the art-of-the-state of the fault detection techniques for the single-phase-grounded fault of 10kV distribution network system, the performance of the transient zero sequence current and voltage in the the single-phase-grounded 10kV distribution network was analyzed in this thesis. Based on the simulating results of the transient zero sequence current and voltage with Matlab Simulink platform and according to the characteristics of both typical real and complex wavelet transform, the author analyzed and compared the applicaions of both the real and the complex wavelet transform in the fault detection of the single-phase-grounded fault of 10kV distribution network system, and proposed a fault detection method according to the performances of both the transient zero currents of all feeders and the scale-coefficients index-values of the multi-scale decomposition in the wavelet transform.Because of the difficulty for the fault detection of the two to four feeder 10 kV distribution network, based on the simulating results of transient zero sequence current for the single-phase-grounded fault for the four feeder 10kV distribution network and below on Matlab platform, the wavelet transform results of the transient zero sequence current with fifty real wavelets including Db N series, Sym N series, Bior Nr.Nd series and Coif N series, etc. were analyzed. And it put forward a fault detection method based on the transient low-frequency zero-sequence-current(ZSC) energy for the first time which was compared with the fault detection methods based on both the transient high-frequency ZSC energy or the maximun module value . For the four feeder 10kV distribution network and below, the results shows that only five real wavelets such as Sym 3、Coif 1、Bior 2.4、Bior 2.6 and Bior 4.4 can detect the faulted line correctly with the transient high-frequency ZSC energy and the maximun module value correctly, and with the low-frequency high-scale-coefficient of the multi-scale decomposition of wavelet transform, the fault line for both NUS and NES network could be located correctly which is not effected with the real wavelets and the feeder number of the network. Therefore, this paper concluded that the fault detection based on the low frequency energy can used to fault detection of 10kV single-phase-grounded system with the real wavelet transform.In order to avoiding the disadvantages of the fault detection with real wavelets, both the transient ZSCs of all feeders and the transient zero-sequence-voltage(ZSV) for the single-phase-grounded fault of five feeder 10kV distribution network were transformed with three kinds of complex wavelets such as Morlet 1-0.1, Cgaussian 1 and Shannon 1-1. According to the phase and mulitude performances of the complex wavelet, three kinds of fault detection methods, such as the transient power, the transient energy and transient module eigenvalue methods of ZSC are put forward. With the analyses and conparions of simulated results, the marked indexes of three fault detection methods, which show the stand or fall of a fault detection method, are indicated.The results show that the three kinds of fault detection methods above based on complex wavelet show a distincting character and a synthetical information. For the three complex wavelets analyszed in this thesis, Morlet wavelet is the best one, and is of a high reliability. Although Cgaussian and Shannon wavelets can select fault line, they are obviously influenced by boundary effection so as to the boundary signal’s distortion. And Morlet wavelet can not be effected by sampling frequency(3~100kHz), transition resistance, fault position, and line’s length. Therefore, the fault detection with complex wavelet has an advantage in the system with two to four feeders, for example, in a single-phase-grounded two feeder NES, its marked factor with Morlet 1-0.1 complex wavelet is greatly higher than that with the real wavelets.更多还原