电气工程与自动化外文翻译中压补偿电网发生间歇性接地故障的检测.docx

翻译部分中文译文中压补偿电网发生间歇性接地故障的检测摘要从波兰中压配电网获得的经验表明，当地判断间歇性接地故障的标准是不可靠的，造成这种情况的原因有两个：一是电网缺乏稳定性，二是测量信号的功率总是会降到保护启动时的功率值以下。本文提出了一种新的基于小波分析理论的自适应算法，它可以检测到发生间歇性接地故障时测量信号的具体动态过程。并且通过EMTP程序包产生的信号对该算法进行了模拟分析。绪论一般而言，波兰的中压配电网中性点是经消弧线圈接地运行的，以此补偿短路电容电流。它主要用于农村地区，线路为架空线路。这种电网的特点是发生接地故障的概率很大，占了所有记录故障的90%多。由于接地点相对较高的过渡电阻，以及降水、风力、高低温的影响，会导致线路的连续性被破坏，从而导致接地故障的发生。这些故障的特点是使得故障检测与定位变得困难。在讨论故障类型时，下面的故障类型可能会遇到： 具有高过渡电阻的电阻性故障， 互感器发生短路， 被周期性和非周期性破坏的故障。实际故障中，可能会表现出上面所列的其中一个或者所有的类型。本文只针对间歇性故障期间自动保护单元的运行情况进行分析。为了评价保护的可操作性，故障时发出的测量信号的大小和类型是确定的。能够表明网络中发生间歇性接地故障的最重要的信号是零序电压，其值通常是通过增加相电压的瞬时值来找出。故障定位可以通过以下几种标准来找出： 零序电流， 零序电流和零序电压产生的功率， 零序电压和零序电流之间的相位差， 零序导纳，或者它的有功电导与无功电纳。然而，上述所列的标准在间歇性接地故障发生时通常是不可行的。网络模型为了模拟和研究间歇性接地故障中的故障现象，本文建立了一个典型的中压对称电网，如图1所示。图1 中压电网图应用EMTP/ATP程序包对故障进行模拟仿真，仿真参数选择如表1所示：表1 模拟15KV网络的参数电网电容电流101，3A故障线电容电流10，6A补偿程度+15%过渡电阻2模型中假设故障发生在一条线路上，其对地电容电流为10，6A，带负荷150KW。考虑以下故障类型：A类型永久性故障，B类型间歇性故障，持续时间=10ms，暂停时间=10ms，C类型间歇性故障，持续时间=10ms，暂停时间=100ms，D类型间歇性故障，持续时间=50ms，暂停时间=50ms，所有故障都是在线路1靠近总线处的始端被模拟的，并标注出了如下量：网络零序电压，故障线路零序电流和非故障线路零序电流。实验结果导纳判据在第部分所列出的用于保护的判据中，可用于判断间歇性接地故障这种“困难型”故障最有效的是零序导纳判据，或者根据它的组成部分：电导和电纳。为了实现基于零序导纳的接地故障保护，本文运用了幅值比较（图2中的CA1，CA2）。在这种最简单的情况下，被比较的输入信号是通过如下式子给出的： （1） （2） （3）系数，决定测量通道中处理信号的比值。图2基于零序导纳接地故障保护的等效模型这种接地故障保护取决于零序导纳模块，通过比较和信号来实现。这种比较发生在幅值比较模块CA1中，保护动作的条件是： （4）也就是说，安装在第i条线路上的保护装置的运行范围是由导纳矢量图的末端决定的，其值满足如下条件： （5）图3中给出了这种保护的启动响应，这是一个光滑的导纳弧线，覆盖了所有的象限。保护运行的补充条件是要有合适的电压零序分量，在这个解决方法里，启动单元是幅值比较模块CA2，它将信号和启动信号相比较，判据如下： （6）其中是预定的启动值，根据它就可以区分接地故障状态与正常运行状态。图3基于导纳的接地故障保护启动响应图4给出了线路1发生永久性故障（A类型）时故障线路的零序电流、非故障线路零序电流、零序电压以及故障线路的零序导纳和非故障线路零序导纳。图4线路1发生永久性故障故障线路零序导纳要比非故障线路的高出许多倍，故根据标准（4）来区分故障线路是可行的。对于中断频率相对较高的间歇性故障（B类型），也有类似的情况，相应的运行结果如图5所示。故障产生的暂态量衰减完之后，零序电压依然接近永久性故障的零序电压值，而且故障线路的零序导纳明显高于非故障线路，并一直持续这种状态。图5线路1发生B类型间歇性接地故障图6线路1发生C类型间歇性接地故障图7线路1发生D类型间歇性接地故障当故障间的暂停时间较长时（如C类型和D类型中的），区分短路线路就比较困难了。这种故障的运行结果如图6和图7所示。整个故障期间零序电压（与启动信号成比例）一直保持较高的幅值。由于故障消失瞬间产生的电压暂态量衰减较低，所以故障检测不会很难。但区分故障线路时就比较困难，因为故障点循环电弧在电网中产生的暂态量，会导致故障线路导纳周期性地降至非故障线路导纳值。C类型和D类型故障均涉及到此问题。在这种情况下，基于零序导纳的保护方法就变得不合适了。而采用小波分析，尤其是多分辨率分解测量信号会更加合适。多分辨率小波分析小波分析的主要工具是多分辨率分解测量信号，它是通过设置多级波形附加过滤器（高通波形和低通尺度函数）实现的。这种可分解波形的计算程序叫和被分析信号的低频成分，其中高频成分在迭代期间不再需要被过滤，所以信号的分解过程就是一种多层次的迭代过程，是通过低通滤波通道实现的，而且连续的近似与连续的分解有关。当选择母小波来分析图5、6、7中的测量信号时，“光滑型”小波（如Morlet小波）是分析信号频谱较好的解决方案，即“光滑型”小波沿频率轴有着较好的频率定位，而对于沿时间轴的，选择连续型小波（如Haar小波）会好一些。查阅了诸多小波类型的性能概要之后，笔者得出一个结论：Haar小波的性能最能满足应用的要求，可以把计量学视为运行在保护方面的实时计量速度。基本的Haar小波定义如下： （7）产生一系列小波，其成分为： （8）函数的优点是定位效果好，通过函数不连续性有利于任意精度的定位来获得时间上的无限精确定位，函数扩展被定义为Haar小波集。Haar缩放函数是通过如下关系式给出： （9）测量信号多级分解的实现过程的特点是计算效率高。Haar小波的应用结果里，每一级的分解被限定在平均值（低通滤波器）的计算和差值（高通滤波器）的计算。平均值比较粗略（近似）而差值比较精确（详细）。图8信号小波分解树测量信号的分解和间歇性接地故障时故障线路的辨别为了辨别故障时的故障线路，第部分应用了分解测量信号的迭代方法，并分析了故障线路零序电流和非故障线路零序电流。图9中给出了B、C、D类型间歇性接地故障产生的信号的6级分解过程，其结果是采用基本Haar小波分析得到的，采样频率为10KHz。图形中可以很明显的看出不同频率范围内信号的特点、幅值和在时间轴上的位置，同时，d1、d2、d3级别上相应的衰减现象也清晰可见。在较高的分解级d4、d5、d6上，衰减现象期间的定位变得要么不那么明显（非故障线路），要么很明显（故障线路），这是区分故障线路的基础，即从非故障线路中区分出故障线路。信号中断瞬间会产生很明显的“尖峰”，导致信号平均值远低于峰值。在d4、d5、d6级中含有高频成分，与绝对平均值成正比，在宽为N的窗口中计算出的第K个值由下式给出： （10）假设间歇性接地故障期间短路线路的辨别是通过比较C和实现的，前者是绝对瞬间值与绝对平均值的比，后者是绝对瞬间值与启动值的比。 （11）图10 B类型间歇性故障时非故障线路零序电流的5级分解图11 C类型间歇性故障时故障线路零序电流的5级分解图12 C类型间歇性故障时非故障线路零序电流的5级分解图13 D类型间歇性故障时故障线路零序电流的5级分解图14 D类型间歇性故障时非故障线路零序电流的5级分解图15间歇性接地故障线路信号的分解算法当保护符合条件（11）时，如果存在零序电压（OS1=1），即接地故障依然存在，且任何关掉导纳保护（OS2=0）的脉冲都没有发出，那么断路器就会跳开。图15给出了间歇性接地故障线路信号的分解算法。图16是一种自适应接地故障保护的模块图，它是传统导纳保护的扩展，基本补偿模块是WD模块和PAC比较器，前者用于测量信号的多级分解，后者用于执行线路分解算法。图16自适应接地故障保护模块图由于保护采用了这种逻辑结构，WD和PAC模块便可以增加故障切除的可靠性，而依靠传统的保护标准是无法实现可靠保护的。总结以上关于间歇性接地故障的和测量信号表明传统的故障保护装置在故障保护及故障线路确定方面存在一些问题，当故障暂停时间超过10毫秒时选线难度会增加。因为这种暂停有可能发生在对称中压配电网，会导致一些非中断型接地故障，所以应该研究可消除这种故障的新型保护方法。分析了间歇性故障的例子之后，可以很容易地看出多级小波分解能够明显地区分信号特点。所有的例子表明，在较低的分解级上（d1，d2，d3），来自故障线路和非故障线路的原始信号的扰动似乎很明显。在较高的分解级上（d4，d5，d6），故障信号和非故障信号的图像有着明显的不同，从而可以区分出故障线路并消除故障。英文原文Detection of the Intermittent Earth Faultsin Compensated MV NetworkAbstract The experience acquired from the Polish medium voltage power distribution networks shows the unreliability of the localization criterions applied to the intermittent earth faults. It results from the lack of stability and low power level of the measuring signals falling often down below the protections start-up level. In the paper, a new adaptive algorithm based on the wavelet analysis enabling detection of specific dynamics of the measuring signal during intermittent earth faults is presented.The algorithm was analyzed utilizing the signals generated in the EMTP program package.I. INTRODUCTIONIn general, the MV distribution networks in Poland operate with the neutral point grounded through the coil to compensate the capacitive short circuit current to the earth. It refers mainly to the rural area networks where the lines are the overhead ones. Such networks are characterized by large number of the earth faults exceeding 90% of all recorded faults. Due to the relatively high cross resistance at the defects location () as well as to the effects of the weather phenomena such as discharges, gusts of wind, high and low temperatures resulting in the rupture of the line conductors continuity, the earth faults occur.Characteristics of these faults makes impossible the detection and localization of such disturbance. The following fault types can be encountered to the discussed faults group: - resistance faults of high cross resistance, - break in the live wire short circuit on the receiver side, - faults being broken cyclically and non-cyclically.An actual fault can show either one or all of the listed features. In the paper, the analysis is limited to the automatic protective units operation during intermittent faults.To assess the protections operability,the levels and features of measuring signals which can occur during the fault are to be identified.The most important signal indicating occurrence of the intermittent earth fault in the network is a zero-voltage component the values of which is often found by adding the instant values of phase voltages.The criterion value of the fault localization can he: - zero current component,- power of the zero current component, and zero voltage component, - phase shift angle between the zero current and voltage components, - , or its components: active zero admittance component,or reactive,. However, the criterion values as listed above are often unreliable when the intermittent earth fault occurs. MODEL OF NETWORKFor modeling and studies of the earth fault phenomena accompanying the intermittent earth faults,a typical medium voltage balanced network has been chosen. The scheme of modeled network is shown in Fig.1.FLg.1. Medium voltage network schemeThe faults were modeled and simulated using the EMTP/ATP program package. Chosen parameters of network assumed for simulation purposes are shown in Table 1. TABLE 1MODEUED 15 KV NETWORK PARAMETERS电网电容电流101，3A故障线电容电流10，6A补偿程度+15%过渡电阻2In the model the assumption was made that the faults occur in a line with to-ground-capacitive current of 10,6 A and a moderated power load of 150 kW. The following fault types have been considered:-A-type - continuous fault, -B-type - intermittent fault of =l0ms duration time,=l0ms pause time, -C-type - intermittent fault of =l0ms duration time,=l00ms pause time, -D-type - intermittent fault of =50ms duration time,=50ms pause time. All faults have been modelled at the beginning of the line 1,adjacent to the bus bars.The following magnitudes have been registered: network voltage zero component, ,as well as the zero component current of the damaged line,and that of the undamaged line,. RESULTS OF EXPERMENTS - ADMTIANCE CRITERIONAmong the criterion values for protections as specified in section 1,for the difficult fault cases such as intermittent earth ones, thc most effective is either the admittance,or one of its components: conductance,or susceptance,. To implement the admittance-type eaah fault protections,the amplitude comparators (CAI,CA2 in Fig 2) are used.In the simplest case,the input signals created according to the following rules are being compared: (1) (2) (3)The,and,coefficients determine the proportionality of the input signal processing in the measuring paths.Fig.2. Admittance-type earth fault protection -equivalent schemeThe earth fault protection responding to the zero- admittance module compares the and signals.The comparison takes place within the amplitude comparator CA1. The protection in the i - th line acts when (4)It means that the operation area of protection located in the i-th line is determined by the ends of admittance vectors of values meeting the condition: (5)In Fig.3,the start-up response of such a protection is shown.It is a plain-admittance curve including equally all quarters of the complex admittance plane. Supplementary clause of the protection operation is a proper level of voltage zero component. In such a solution,the start-up unit is the amplitude comparator CA2 in which thesignal level is compared to that of thestart-up signal level, according the the clause: (6)where theis the preset start-up value due to which the earth fault can be differentiated from the normal network operation.Fig3 Start-up response of the admittance-type earth-faul protenionIn Fig.4,the current zero components in the damaged line,and in the undamaged line, voltage zero component as well as zero-admittances in the damaged line, and undamaged line, during continuous fault (A-type) in line 1 are presented.Fig.4 Continuous eanh fault (A-type) in line 1The zero admittance measured in the damaged line is some-fold higher than that in the undamaged line.Discrimination of damaged line according to criterion (4) should not pose the problems. Similar case is for an intermittent fault of relatively high interruption frequency (B-type), the corresponding runs of nalysed magnitudes are shown in Fig.5. After decay of the transient state resulting from the fault ocurrence,the zero voltage remains at the level adjacent to the voltage during the continuous fault,and the admittance measured in the damaged line is all the time evidently higher than the admittance measured in the undamaged line.Rys.5.Intermittent earth fault (B-type) in line1Fig.6.Intermittent earth fault (C-type) in line1Fig.7.Intermittent earth fault (D-type) in line1A discrimination of the short-circuited line in case when the pause between successive faults is relatively long (long time- C and D types) can he more difficult. The runs related to such faults are shown in figures 6 and 7. During entire duration of fault,the voltage zero component(to which the start-up signal is proportional) remains at the high level;it means that the fault detection should not be difficult due to the relatively low attenuation of the voltage transients after instantaneous disappearing of the fault.However, the problems could arise with damaged line discrimination.Due to the features of the transient process in the network resulting from the cyclic arc ignitions in the fault location, the damaged line admittance falls cyclically down to the undamaged line admittance level.It refers to both the C-type and D-type faults.In such a case,an improper operation of the admittance criterion-related protection can be expected. Better opportunities open when using the wavelet expansions,especially the multi-resolution decomposition of the measuring signals (WD). MULTI-RFSOLUTION WAVELET ANALYSISA main tool of the wavelet analysis in the proposed application is the multi-resolution decomposition of measuring signals realized by the multistage set of the wavelet complementary filters(high-pass wavelets and low-pass scaling functions).The calculating procedure leading to the decomposition is called the Mallut algorithm.The iteration process of creating the multi-resolution signal representation can be presented in the form of the wavelet signal decomposition tree as shown in Fig.8.At any iterative step,the analysed signal is filtered.The number of iterative steps is unlimited;in Fig.8,n steps have been assumed.Each iteration results in both:the high-frequency component called a detail (Di) which is no more filtered during successive iterative steps,and the low-frequency component(Ai) of analysed original signal S,called an app roximafion.Thus,the signal decomposition process has a form of the multilevel iterative process carried out on the low-pas filtration channel, and the successive approximations are subject to the successive decomposition.When choosing the mother wavelet for analysis of measuring signals shown in figures 5,6 and7,the known rule has been taken into account the smooth - shape wavelets (the Morlets wavelet, for example) are of better resolution when analysing the signal frequency spectrum,i.e. they have better localization of frequency components along the frequency axis,while the discontinuously-shaped wavelets (the Haars wavelet,for example) have better resolution along the time axis.Referring to the overview of properties of many wavelets types,the Authors drew a conclusion that the Haars wavelets properties meet in the hest way the requirements of the considered application,regarding both the metrological aspects as the speed of real-time calculations carried out in protections. The basic Haars wavelet is defined as follows:and generates a set of wavelets with elements as:Advantage of thefunction in proposed application is their good localization as for an infinitely precise localization in time is obtained enabling arbitrary accuracy of localization of the function discontinuity (especially that of the step - see Figures 5 and 6), the function expansion being defined regarding the Haars wavelet set. The Haars scaling function is given by relationship:Presented realization of measuring signals multi-level decomposition is characterized by high calculation efficiency.In r