CN104655991A

CN104655991A - Power system fault matching method based on mutant point dejection combinational algorithm

Info

Publication number: CN104655991A
Application number: CN201510121511.3A
Authority: CN
Inventors: 龚庆武; 占劲松
Original assignee: Wuhan University WHU
Current assignee: Wuhan University WHU; State Grid Eastern Inner Mongolia Power Co Ltd
Priority date: 2015-03-19
Filing date: 2015-03-19
Publication date: 2015-05-27
Anticipated expiration: 2035-03-19
Also published as: CN104655991B

Abstract

The invention provides a power system fault data matching method based on a mutation point detection combination algorithm, which includes randomly extracting a group of fault data at both ends of a certain line from the fault recording database in the credit guarantee system, and using the mutation point detection combination algorithm to respectively Calculate the corresponding fault starting points at both ends, obtain the relevant criterion parameters according to the fault starting points at both ends calculated in step 2, and judge whether the fault data at both ends match according to the triple criterion of data matching, including applying mutation The fault phase selection method is based on the current measurement phase selection method to determine whether the fault types corresponding to the data at both ends match; if they match, judge whether the load currents at both ends match according to the load current criterion, and if they match, use the closing angle criterion to judge the data at both ends Whether it matches or not, the match means that the judgment results of the three criteria all match, and it is finally determined that the data at both ends match, otherwise the data at both ends do not match.

Description

Power System Fault Data Matching Method Based on Catastrophic Point Detection Combination Algorithm

技术领域technical field

本发明涉及到电力系统故障定位领域，特别涉及一种基于突变点检测组合算法的电力系统故障数据匹配方法。The invention relates to the field of power system fault location, in particular to a power system fault data matching method based on a mutation point detection combination algorithm.

背景技术Background technique

电力系统故障数据蕴含着丰富的信息资源，为使其得到充分利用，目前人们以原有的故障信息处理系统为基础，运用数据挖掘技术研制出保护设备故障信息管理与分析系统(简称“保信系统”)。保信系统读取由GPS精确授时系统同步的故障信息子站保护装置、录波装置等重要信息，完成计算、分析、绘图、信息输出等功能，为运行与继电保护人员提供决策信息。Power system fault data contains rich information resources. In order to make full use of it, people are currently developing a protection equipment fault information management and analysis system (referred to as "guarantee system") based on the original fault information processing system and using data mining technology. "). The Baoxin system reads important information such as fault information substation protection devices and wave recording devices synchronized by the GPS precise timing system, completes functions such as calculation, analysis, drawing, and information output, and provides decision-making information for operation and relay protection personnel.

故障定位是保信系统的一项重要功能。目前，在输电线路故障定位技术中，双端测距由于充分利用故障信息，具有测距精度高、操作简单、实用性强等优点，在电力系统中得到广泛应用。保信系统中的双端测距模块所利用的故障数据为故障线路两端同次故障的电量即所谓的匹配数据，而故障录波数据库中储存着海量的数据，因此在双端测距前必须对故障数据进行筛选以使其相互匹配。一般情况下，由于有GPS精确授时系统保证数据同步性，时标相同的故障数据即为匹配数据。然而实际中，外界干扰可能使得GPS精确授时系统产生误差，虽然有学者提出采用高精度晶振对GPS时钟进行在线监测与校正，但进一步考虑到互感器相移、硬件延时和采样率差别等因素引入的误差，保信系统中的匹配数据时标也并非严格相同。因此，当线路出现连续性故障，各次故障数据时标接近时，仅仅根据时标进行数据匹配可能使得测距模块读取数据出错，继而导致故障定位失败。因此，研究一种新的故障数据匹配方法使得双端测距模块在进行数据筛选时不再单一地依靠GPS精确授时系统，具有十分重要的意义。Fault location is an important function of Baoxin system. At present, in the transmission line fault location technology, double-terminal ranging has been widely used in power systems due to its advantages of high ranging accuracy, simple operation, and strong practicability due to the full use of fault information. The fault data used by the double-terminal ranging module in the Baoxin system is the power of the same fault at both ends of the fault line, which is the so-called matching data, and the fault recording database stores a large amount of data, so it must be used before the double-terminal ranging. Fault data is filtered to match each other. In general, due to the GPS precise timing system to ensure data synchronization, the fault data with the same time scale is the matching data. However, in practice, external interference may cause errors in the GPS precise timing system. Although some scholars have proposed to use high-precision crystal oscillators to monitor and correct the GPS clock online, further considerations such as transformer phase shifts, hardware delays, and sampling rate differences Due to the error introduced, the time scale of matching data in the guarantee system is not strictly the same. Therefore, when there is a continuous fault on the line and the time scales of each fault data are close to each other, only performing data matching based on the time scale may cause errors in reading data by the distance measuring module, which in turn leads to failure in fault location. Therefore, it is of great significance to study a new fault data matching method so that the double-terminal ranging module no longer relies solely on the GPS precise timing system when performing data screening.

发明内容Contents of the invention

本发明主要是解决现有方法存在的技术问题，设计出一种基于突变点检测组合算法的电力系统故障数据匹配方法。The invention mainly solves the technical problem existing in the existing method, and designs a power system fault data matching method based on a mutation point detection combined algorithm.

本发明提供的技术方案为一种基于突变点检测组合算法的电力系统故障数据匹配方法，包括以下步骤，The technical solution provided by the present invention is a power system fault data matching method based on a combination algorithm of mutation point detection, comprising the following steps,

步骤一、获取故障数据，包括在保信系统内的故障录波数据库中随机提取某条线路两端的故障数据各一组，记线路一端为m端，另一端为n端；Step 1. Acquiring fault data, including randomly extracting a group of fault data at both ends of a line from the fault recording database in the Baoxin system, and recording one end of the line as m-terminal and the other end as n-terminal;

步骤二、利用突变点检测组合算法分别计算出两端相应的故障起点，包括对线路的两端分别执行以下子步骤，Step 2, using the mutation point detection combination algorithm to calculate the corresponding fault starting points at both ends, including performing the following sub-steps on the two ends of the line respectively,

步骤A1、基于步骤一提取的某条线路一端的故障数据组，生成基本信号序列，实现方式为运用K变换对三相电流进行相模变换：Step A1, based on the fault data group at one end of a certain line extracted in step 1, a basic signal sequence is generated, and the implementation method is to use K transformation to perform phase-to-mode transformation on the three-phase current:

$[\begin{matrix} {i i}_{00} \\ {i i}_{11} \\ {i i}_{22} \end{matrix}] = = [\begin{matrix} 11 & 11 & 11 \\ 11 & 22 & - - 33 \\ 11 & - - 33 & 22 \end{matrix}] [\begin{matrix} {i i}_{a a} \\ {i i}_{b b} \\ {i i}_{c c} \end{matrix}]$

其中，i_a、i_b、i_c为三相电流，i₀、i₁、i₂为相应的三种模分量，选取一种模分量作为基本信号，生成基本信号序列；Among them, i _a , i _b , i _c are the three-phase currents, i ₀ , i ₁ , i ₂ are the corresponding three kinds of modulus components, and one kind of modulus component is selected as the basic signal to generate the basic signal sequence;

步骤A2、运用贝叶斯分类算法对步骤A1所得基本信号序列进行分类，找出可疑突变点；Step A2, using the Bayesian classification algorithm to classify the basic signal sequence obtained in step A1, and find out suspicious mutation points;

步骤A3、求出步骤A1所得基本信号序列的突变量电流，运用突变量电流算法计算出电流突变点n_s0；Step A3, obtain the mutation current of the basic signal sequence obtained in step A1, and calculate the current mutation point n _s0 using the mutation current algorithm;

步骤A4、运用突变点检测组合算法确定故障起点，包括以步骤A3所得n_s0为基准构造时间窗[n_s0-Δn,n_s0]，其中Δn为预设的窗宽，利用时间窗从由步骤A2所得可疑突变点中筛选出故障起点，筛选方式如下，Step A4, using the combination algorithm of mutation point detection to determine the starting point of the fault, including constructing a time window [n _s0 -Δn,n _s0 ] based on n _s0 obtained in step A3, where Δn is the preset window width, using the time window from the step The starting point of the fault is screened out from the suspicious mutation points obtained in A2, and the screening method is as follows,

1)当时间窗中只有一个可疑突变点时，故障起点n_s为此可疑突变点；1) When there is only one suspicious mutation point in the time window, the fault starting point n _s is the suspicious mutation point;

2)当时间窗中有两个及以上的可疑突变点时，故障起点n_s为这些可疑突变点平均后取整；2) When there are two or more suspicious mutation points in the time window, the fault starting point n _s is rounded after the average of these suspicious mutation points;

3)当时间窗中没有可疑突变点时，故障起点n_s为步骤A3所得电流突变点n_s0；3) When there is no suspicious sudden change point in the time window, the fault starting point n _s is the current sudden change point n _s0 obtained in step A3;

步骤三、根据步骤二中计算出的两端故障起点求取相关判据参数，并根据数据匹配的三重判据判断两端故障数据是否匹配，包括以下子步骤：Step 3, obtain the relevant criterion parameters according to the fault starting points at both ends calculated in step 2, and judge whether the fault data at both ends match according to the triple criterion of data matching, including the following sub-steps:

步骤B1、以两端相应的故障起点n_s为基准，运用傅氏算法分别计算出两端数据各相的正常与故障电流及电压相量；Step B1, based on the corresponding fault starting points n _s at both ends, use the Fourier algorithm to calculate the normal and fault current and voltage phasors of each phase of the data at both ends;

步骤B2、对两端分别运用突变量电流选相法进行故障选相，判断两端数据对应的故障类型是否匹配；Step B2, using the mutation current phase selection method for the two ends to select the fault phase, and judging whether the fault types corresponding to the data at the two ends match;

步骤B3、若步骤B2判断结果为两端数据的故障类型不匹配，则直接判定两端数据不匹配，结束流程；若匹配则计算m端电流和如下，Step B3, if the judgment result of step B2 is that the fault types of the data at both ends do not match, then directly determine that the data at both ends do not match, and end the process; if they match, calculate the current at the m terminal and as follows,

两端均以A相为基准相，运用傅氏算法计算由步骤二所得m端的故障起点前推一个周波的电流相量，得到m端电流运用傅氏算法计算由步骤二所得n端的故障起点前推一个周波的电流相量和电压相量，分别得到n端电流和n端电压 Both ends take phase A as the reference phase, and use the Fourier algorithm to calculate the current phasor of one cycle forward from the fault starting point of terminal m obtained in step 2, and obtain the current at terminal m Use the Fourier algorithm to calculate the current phasor and voltage phasor of one cycle forward from the fault starting point at the n-terminal obtained in step 2, and obtain the current at the n-terminal respectively and n-terminal voltage

再运用线路分布参数模型由n端电流和n端电压推算对应的m端电流 Then use the distribution parameter model of the line from the n-terminal current and n-terminal voltage Calculate the corresponding m-terminal current

${\overset{\cdot \cdot}{I I}}_{m m}^{' '} = = \frac{sh sh ((γ γ \times \times L L))}{{Z Z}_{c c}} \times \times {\overset{\cdot \cdot}{U u}}_{n no} - - ch ch ((γ γ \times \times L L)) \times \times {\overset{\cdot \cdot}{I I}}_{n no}$

其中，γ和z_c分别为线路的传播常数与特性阻抗，L为线路长度；Among them, γ and z _c are the propagation constant and characteristic impedance of the line respectively, and L is the line length;

步骤B4、根据负荷电流判据判断两端负荷电流是否匹配，包括若|ρ-1|≤λ，则负荷电流匹配；否则，负荷电流不匹配，其中，参数λ为预设的阀值；Step B4. Determine whether the load currents at both ends match according to the load current criterion, including if |ρ-1|≤λ, then the load current matches; otherwise, the load current does not match, wherein, the parameter λ is the preset threshold;

步骤B5、若步骤B4判断结果为两端负荷电流不匹配，则判定两端数据不匹配，结束流程；若匹配，则利用合闸角判据判断两端数据是否匹配，包括若参数则合闸角匹配，否则合闸角不匹配，其中，为预设的阀值；Step B5. If the judgment result of step B4 is that the load currents at both ends do not match, it is determined that the data at both ends do not match, and the process ends; if they match, use the closing angle criterion to judge whether the data at both ends match, including if the parameters Then the closing angle matches, otherwise the closing angle does not match, where, is the preset threshold;

步骤B6、若步骤B5判断结果为合闸角匹配，说明三重判据判定结果均匹配，则最终判定两端数据匹配，否则两端数据不匹配。Step B6. If the judgment result of step B5 is that the closing angle matches, it means that the judgment results of the three criteria all match, and it is finally judged that the data at both ends match, otherwise the data at both ends do not match.

而且，步骤A3运用突变量电流算法寻找电流突变点的实现如下：Moreover, the implementation of step A3 using the sudden change current algorithm to find the current sudden change point is as follows:

a)根据公式Δi(k)＝i(k)-i(k-N)计算突变量电流Δi(k)，其中N为故障录波装置一个周期的采样点数，i(k)为基本信号i的采样点k处电流；a) Calculate the sudden change current Δi(k) according to the formula Δi(k)=i(k)-i(k-N), where N is the number of sampling points in one cycle of the fault recording device, and i(k) is the sampling of the basic signal i current at point k;

b)根据检测判据确定电流突变点，判据如下，b) Determine the current mutation point according to the detection criterion, the criterion is as follows,

$\{\begin{matrix} | | ψ ψ ((k k)) | | > > ξ ξ \\ card card ((A A)) &GreaterEqual; &Greater Equal; β β \end{matrix}$

其中，参数A＝{ψ(n)||ψ(n)|＞ξ，k≤n＜k+α}，α与β、ξ为预设参数，满足上述判据的第一个k值为被检测信号对应的突变点。Among them, the parameter A={ψ(n)||ψ(n)|>ξ, k≤n<k+α}, α, β, and ξ are preset parameters, and the first k value that meets the above criteria is the detected signal the corresponding mutation point.

本发明针对电力系统故障数据设计出一种突变点检测组合算法，在此基础上提出了在故障录波数据库中进行数据匹配的三重判据，利用三重判据有效判定线路两端故障数据是否属于同次故障，即数据是否匹配，从而为输电线路双端故障测距的有效进行提供基本的数据保证。因此，本发明具有如下优点：运用基于电流突变量的故障起点检测法进行故障数据匹配，判定结果可靠性高，实时性强，简单实用，能够为保信系统中的双端测距模块提供一种比简单依靠GPS精确授时进行时标匹配更为可靠的电力系统故障数据匹配方法。The present invention designs a mutation point detection combination algorithm for power system fault data, and on this basis proposes a triple criterion for data matching in the fault record database, using the triple criterion to effectively determine whether the fault data at both ends of the line belongs to The same fault, that is, whether the data match, provides the basic data guarantee for the effective conduct of fault location at both ends of the transmission line. Therefore, the present invention has the following advantages: using the fault starting point detection method based on the sudden change of current to carry out fault data matching, the judgment result has high reliability, strong real-time performance, simple and practical, and can provide a double-terminal ranging module in the guarantee system It is a more reliable power system fault data matching method than simply relying on GPS accurate time service for time scale matching.

附图说明Description of drawings

图1为本发明实施例用于仿真的双端供电系统电路图。FIG. 1 is a circuit diagram of a double-terminal power supply system used for simulation according to an embodiment of the present invention.

图2为本发明实施例的电力系统故障数据匹配判定流程图。Fig. 2 is a flow chart of power system fault data matching determination according to an embodiment of the present invention.

图3为本发明实施例的故障选相流程图。Fig. 3 is a flow chart of fault phase selection in the embodiment of the present invention.

具体实施方式Detailed ways

下面通过实施例并结合附图，对本发明的技术方案作进一步具体的说明。The technical solutions of the present invention will be further specifically described below through the embodiments and in conjunction with the accompanying drawings.

参见图2，本发明的实施例包括步骤如下：Referring to Fig. 2, the embodiment of the present invention comprises steps as follows:

步骤一、获取电力系统故障数据的步骤：在保信系统内的故障录波数据库中随机提取某条线路两端的故障数据各一组。Step 1. The step of obtaining the fault data of the power system: Randomly extract a group of fault data at both ends of a certain line from the fault recording database in the Baoxin system.

参见图1，线路一端(左端)为m端，另一端(右端)为n端，E_m、E_n和Z_m、Z_n分别表示m端和n端电源的电压与阻抗，L为线路长度，d为m端的故障距离。实施例在故障录波数据库中分别随机选取故障线路m侧和n侧的故障数据各一组。Referring to Figure 1, one end (left end) of the line is m-terminal, the other end (right end) is n-terminal, E _m , E _n and Z _m , Z _n respectively represent the voltage and impedance of the m-terminal and n-terminal power supplies, and L is the length of the line , d is the fault distance at the m end. In the embodiment, a group of fault data on the m side and the n side of the fault line are respectively randomly selected in the fault recording database.

步骤二、检测故障起点的步骤：利用突变点检测组合算法分别计算出两端数据相应的故障起点，为之后运用三重判据进行故障数据匹配时求取相关判据参数作相应准备。Step 2. The step of detecting the starting point of the fault: use the mutation point detection combination algorithm to calculate the corresponding fault starting point of the data at both ends, and make corresponding preparations for obtaining the relevant criterion parameters when using the triple criterion for fault data matching.

实施例的步骤二具体包括对线路两端分别基于相应故障数据执行以下子步骤：Step 2 of the embodiment specifically includes performing the following sub-steps on the two ends of the line based on the corresponding fault data:

步骤A1、基于步骤一提取的某条线路一端的故障数据组，生成基本信号序列，作为待检测序列，实现方式为运用K变换对三相电流进行相模变换：Step A1, based on the fault data group at one end of a certain line extracted in step 1, generate a basic signal sequence as a sequence to be detected, and the implementation method is to use K transformation to perform phase-mode transformation on the three-phase current:

其中，i_a、i_b、i_c为三相电流，i₀、i₁、i₂为相应的三种模分量，具体实施时，可以选择其中任一种模分量进行后续分析。实施例选取模分量i₁＝i_a+2i_b-3i_c作为基本信号，生成基本信号序列，进行后文的分析与计算。Among them, ia, _ib , and _ic are the three-phase currents, and i ₀ , i ₁ , and i ₂ are the corresponding three _mode components. During specific implementation, any one of the mode components can be selected for subsequent analysis. The embodiment selects the modulus component i ₁ =i _a +2i _b -3i _c as the basic signal, generates a basic signal sequence, and performs analysis and calculation in the following.

步骤A2、运用贝叶斯分类算法对步骤A1所得基本信号序列进行分类，找出“可疑突变点”。Step A2, using the Bayesian classification algorithm to classify the basic signal sequence obtained in step A1, and find out "suspicious mutation points".

步骤A3、求出步骤A1所得基本信号序列的突变量电流，运用突变量电流算法计算出电流突变点n_s0。Step A3, calculating the sudden change current of the basic signal sequence obtained in step A1, and calculating the current sudden change point n _s0 by using the sudden change current algorithm.

步骤A4、运用突变点检测组合算法确定故障起点。Step A4, using the mutation point detection combination algorithm to determine the fault starting point.

实施例的步骤A4以步骤A3所得n_s0为基准构造时间窗[n_s0-Δn,n_s0]，其中Δn为预设的窗宽，具体实施时本领域技术人员可自行预设取值。利用时间窗从由步骤A2得出的“可疑突变点”中筛选出故障起点。筛选方案为：Step A4 of the embodiment constructs a time window [n _s0 -Δn,n _s0 ] based on n _s0 obtained in step A3, where Δn is a preset window width, and those skilled in the art can preset the value during specific implementation. Use the time window to screen out the fault starting point from the "suspect mutation point" obtained in step A2. The screening scheme is:

1)当时间窗中只有一个“可疑突变点”时，故障起点n_s即为此“可疑突变点”；1) When there is only one "suspicious mutation point" in the time window, the fault starting point n _s is this "suspicious mutation point";

2)当时间窗中有两个及以上的“可疑突变点”时，故障起点n_s为这些“可疑突变点”平均后取整；2) When there are two or more "suspicious mutation points" in the time window, the fault starting point n _s is rounded after the average of these "suspicious mutation points";

3)当时间窗中没有“可疑突变点”时，故障起点n_s为步骤A3所得电流突变点n_s0，n_s＝n_s0。3) When there is no "suspicious sudden change point" in the time window, the fault starting point n _s is the current sudden change point n _s0 obtained in step A3, n _s =n _s0 .

具体实施时，步骤A2采用的贝叶斯分类算法可参考：刘密歌，李小斌.基于Bayes决策的奇异点检测[J].计算机应用，2013，01:230-232.为便于实施参考起见，提供实施例中步骤A2运用贝叶斯分类算法寻找基本信号序列中的“可疑突变点”的具体实现如下：For specific implementation, the Bayesian classification algorithm used in step A2 can refer to: Liu Mige, Li Xiaobin. Singularity detection based on Bayesian decision [J]. Computer Applications, 2013, 01:230-232. For the convenience of implementation reference, The specific implementation of step A2 in the provided embodiment using the Bayesian classification algorithm to find "suspicious mutation points" in the basic signal sequence is as follows:

a)将基本信号序列中元素分为两类：定义突变点集合ω₁，先验概率为P(ω₁)；非突变点集合ω₂，先验概率为P(ω₂)；a) Divide the elements in the basic signal sequence into two categories: define the mutation point set ω ₁ , the prior probability is P(ω ₁ ); the non-mutation point set ω ₂ , the prior probability is P(ω ₂ );

b)对基本信号序列进行预处理得到序列{z(k)}。b) Preprocessing the basic signal sequence to obtain the sequence {z(k)}.

首先求出基本信号序列的各采样点k处的电流导数x(k)：First, calculate the current derivative x(k) at each sampling point k of the basic signal sequence:

$x x ((k k)) = = \frac{d d (({i i}_{11} ((k k))))}{dt dt} = = \frac{{i i}_{11} ((k k + + 11)) - - {i i}_{11} ((k k - - 11))}{22 T T}$

其中，i₁(k)为基本信号i₁的采样点k处电流。Wherein, i ₁ (k) is the current at the sampling point k of the basic signal i ₁ .

再求出x(k)在各采样点k处的突变量y(k)：Then calculate the mutation amount y(k) of x(k) at each sampling point k:

y(k)＝[x(k)-x(k+1)]² y(k)=[x(k)-x(k+1)] ²

最后对y(k)进行归一化处理得到z(k)：Finally, normalize y(k) to get z(k):

$z z ((k k)) = = \frac{y the y ((k k)) - - ((11 - - δ δ)) \cdot &Center Dot; min min ((y the y))}{((11 + + δ δ)) \cdot &Center Dot; max max ((y the y)) - - ((11 - - δ δ)) \cdot &Center Dot; min min ((y the y))}$

其中，maxy(k)、miny(k)分别为序列{y(k)}中的最大与最小元素，T为相邻采样点之间的采样间隔，参数δ为数值很小的正数，具体实施时本领域技术人员可自行预设参数δ的取值。Among them, maxy(k) and miny(k) are the largest and smallest elements in the sequence {y(k)} respectively, T is the sampling interval between adjacent sampling points, and the parameter δ is a positive number with a very small value, specifically During implementation, those skilled in the art can preset the value of parameter δ by themselves.

c)对类条件概率P(z(k)|ω₁)与P(z(k)|ω₂)作如下估计：c) Estimate the class conditional probability P(z(k)|ω ₁ ) and P(z(k)|ω ₂ ) as follows:

$\overset{^^}{P P} ((z z ((k k)) | | {ω ω}_{11})) = = \frac{{z z ((k k))}^{χ χ}}{{Σ Σ}_{k k = = 11}^{Γ Γ} z z {((k k))}^{χ χ}}$

$\overset{^^}{P P} ((z z ((k k)) | | {ω ω}_{22})) = = \frac{11 - - z z {((k k))}^{χ χ}}{Γ Γ - - {Σ Σ}_{k k = = 11}^{Γ Γ} z z {((k k))}^{χ χ}}$

其中，Γ为序列{z(k)}的元素个数，即采样点的总数，1≤k≤Γ；参数χ为正，具体实施时本领域技术人员可自行预设参数χ的取值；为类条件概率P(z(k)|ω₁)的估计值，为类条件概率P(z(k)|ω₂)的估计值。Wherein, Γ is the number of elements of the sequence {z(k)}, that is, the total number of sampling points, 1≤k≤Γ; the parameter χ is positive, and those skilled in the art can preset the value of the parameter χ during specific implementation; is the estimated value of class conditional probability P(z(k)|ω ₁ ), is the estimated value of class conditional probability P(z(k)|ω ₂ ).

d)分类原则d) Classification principles

计算分类错误率P(e)Calculate the classification error rate P(e)

$\begin{matrix} P P ((e e)) = = P P ((x x &Element; &Element; {R R}_{11},, x x &Element; &Element; {ω ω}_{22})) + + P P ((x x &Element; &Element; {R R}_{22},, x x &Element; &Element; {ω ω}_{11})) \\ = = \underset{{R R}_{11}}{&Integral; &Integral;} P P ((x x | | {ω ω}_{22})) P P (({ω ω}_{22})) dx dx + + \underset{{R R}_{22}}{&Integral; &Integral;} P P ((x x | | {ω ω}_{11})) P P (({ω ω}_{11})) dx dx \\ = = P P (({ω ω}_{22})) \underset{{R R}_{11}}{&Integral; &Integral;} P P ((x x | | {ω ω}_{22})) dx dx + + P P (({ω ω}_{11})) \underset{{R R}_{22}}{&Integral; &Integral;} P P ((x x | | {ω ω}_{11})) dx dx \\ = = P P (({ω ω}_{22})) {P P}_{22} ((e e)) + + P P (({ω ω}_{11})) {P P}_{11} ((e e)) \end{matrix}$

其中，R₁为序列{z(k)}中突变点对应的数据区域，R₂为序列{z(k)}中非突变点对应的数据区域，P(x∈R₁,x∈ω₂)表示将区域R₁中的数据判定为非突变点的概率，P(x∈R₂,x∈ω₁)表示将区域R₂中的数据判定为突变点的概率，P(ω₂)P₂(e)为将突变点检测为非突变点(漏检)的概率，P(ω₁)P₁(e)为将非突变点检测为突变点(误检)的概率，先验概率P(ω₁)与P(ω₂)为常数(均未知)。计算推论过程可参见贝叶斯分类算法相关文献。Among them, R ₁ is the data region corresponding to the mutation point in the sequence {z(k)}, R ₂ is the data region corresponding to the non-mutation point in the sequence {z(k)}, P(x∈R ₁ ,x∈ω ₂ ) represents the probability of judging the data in region R ₁ as a non-mutation point, P(x∈R ₂ , x∈ω ₁ ) represents the probability of judging the data in region R ₂ as a mutation point, P(ω ₂ )P ₂ (e) is the probability of detecting a mutation point as a non-mutation point (missed detection), P(ω ₁ )P ₁ (e) is the probability of detecting a non-mutation point as a mutation point (false detection), and the prior probability P (ω ₁ ) and P(ω ₂ ) are constants (both unknown). For the calculation and inference process, please refer to the relevant literature of Bayesian classification algorithm.

为了在P₂(e)＝ε时使得P₁(e)最小，可运用贝叶斯决策理论中的Neyman-Pearson准则进行分类。参数ε为一很小的正数，具体实施时本领域技术人员可自行预设参数ε的取值。In order to minimize P ₁ (e) when P ₂ (e)=ε, the Neyman-Pearson criterion in Bayesian decision theory can be used for classification. The parameter ε is a very small positive number, and those skilled in the art can preset the value of the parameter ε during specific implementation.

e)确定“可疑突变点”e) Determination of "suspect mutation points"

根据Neyman-Pearson准则，有：According to the Neyman-Pearson criterion, there are:

${&Integral; &Integral;}_{{R R}_{11}^{* *}} P P ((z z ((k k)) | | {ω ω}_{22})) dx dx = = ϵ ϵ$

$μ μ = = \frac{\overset{^^}{P P} ((z z (({k k}^{* *})) | | {ω ω}_{11}))}{\overset{^^}{P P} ((z z (({k k}^{* *})) | | {ω ω}_{22}))} = = \frac{z z {(({k k}^{* *}))}^{χ χ} \cdot \cdot [[Γ Γ - - {Σ Σ}_{k k = = 11}^{Γ Γ} z z {((k k))}^{χ χ}]]}{{Σ Σ}_{k k = = 11}^{Γ Γ} z z {((k k))}^{χ χ} \cdot \cdot [[11 - - z z {(({k k}^{* *}))}^{χ χ}]]}$

其中为区域R₁的某一子区域。in It is a certain sub-region of the region _R1 .

将元素(1≤k≤N，N为一个周期的采样点数)按照从小到大的顺序排列得到序列然后构造新序列{S(k)}，令其中元素根据不等式S(k^*)≤ε＜S(k^*+1)求出满足不等式的元素相应序号k^*，继而求出参数μ的值。will element (1≤k≤N, N is the number of sampling points in one period) Arranged in ascending order to obtain the sequence Then construct a new sequence {S(k)}, let the elements According to the inequality S(k ^* )≤ε<S(k ^* +1), the corresponding serial number k ^* of the element satisfying the inequality is obtained, and then the value of the parameter μ is obtained.

最后根据判据则 $z (k) = \{\begin{matrix} ω_{1} \\ ω_{2} \end{matrix}$ 进行信号采样点分类，其中集合ω₁中的元素统称为“可疑突变点”。Finally according to the criterion but $z (k) = \{\begin{matrix} ω_{1} \\ ω_{2} \end{matrix}$ The signal sampling points are classified, and the elements in the set ω ₁ are collectively referred to as "suspect mutation points".

为便于实施参考起见，提供实施例中步骤A3运用突变量电流算法寻找电流突变点的具体实现如下：For the convenience of implementation and reference, the specific implementation of step A3 in the embodiment to find the current sudden change point using the sudden change current algorithm is as follows:

a)根据公式Δi(k)＝i(k)-i(k-N)计算突变量电流Δi(k)：a) Calculate the sudden change current Δi(k) according to the formula Δi(k)=i(k)-i(k-N):

其中，i(k)为基本信号i的采样点k处电流。Wherein, i(k) is the current at the sampling point k of the basic signal i.

实施例根据公式Δi₁(k)＝i₁(k)-i₁(k-N)计算突变量电流Δi₁(k)，其中N为故障录波装置一个周期的采样点数；Embodiment According to the formula Δi ₁ (k)=i ₁ (k)-i ₁ (kN), the sudden change current Δi ₁ (k) is calculated, wherein N is the number of sampling points in one cycle of the fault recorder;

b)根据检测判据确定电流突变点，判据如下：b) Determine the current mutation point according to the detection criteria, the criteria are as follows:

其中，参数A＝{ψ(n)||ψ(n)|＞ξ，k≤n＜k+α}，card(A)表示集合A中的元素个数。参数α与β可预先设定，具体设置可与一个周期的采样点数N有关，一般可取(N较大时取上限，较小时取下限)。满足上述判据的第一个k值即为被检测信号对应的突变点。具体实施时本领域技术人员可自行预设参数ξ的取值。Among them, the parameter A={ψ(n)||ψ(n)|>ξ, k≤n<k+α}, card(A) indicates the number of elements in the set A. The parameters α and β can be set in advance, and the specific setting can be related to the number of sampling points N in a cycle, which is generally desirable (The upper limit is taken when N is larger, and the lower limit is taken when N is small). The first k value that satisfies the above criteria is the mutation point corresponding to the detected signal. During specific implementation, those skilled in the art can preset the value of the parameter ξ by themselves.

对线路的m端和n端分别执行步骤A1～A4后，得到相应的故障起点。After performing steps A1-A4 on the m-end and n-end of the line respectively, the corresponding fault starting point is obtained.

步骤三、故障数据匹配判定：根据上一步骤中计算出的两端相应故障起点求取相关判据参数并根据数据匹配的三重判据判断两端故障数据是否匹配。Step 3. Fault data matching judgment: Obtain relevant criterion parameters according to the corresponding fault starting points at both ends calculated in the previous step, and judge whether the fault data at both ends match according to the triple criterion of data matching.

本发明三重判据设计如下：The triple criterion design of the present invention is as follows:

实施例的步骤三具体包括以下子步骤：Step three of the embodiment specifically includes the following sub-steps:

步骤B1、以两端相应的故障起点n_s为基准，运用傅氏算法分别计算出两端数据各相的正常与故障电流及电压相量。Step B1. Based on the corresponding fault starting points n _s at both ends, the Fourier algorithm is used to calculate the normal and fault current and voltage phasors of each phase of the data at both ends.

步骤B2、对两端分别运用突变量电流选相法进行故障选相，判断两端数据对应的故障类型是否匹配，此即为故障类型判据，即后续步骤中，若两端故障选相结果不同，则直接判定两端数据不匹配，结束流程；若两端故障选相结果相同，即两端故障类型匹配，则须运用负荷电流判据与合闸角判据作进一步判定。具体实施时，故障选相实现可参考：杨奇逊.微型机继电保护基础[M].北京：中国电力出版社，2004.Step B2. Use the sudden change current phase selection method to select the fault phase at both ends, and judge whether the fault types corresponding to the data at both ends match. This is the fault type criterion, that is, in the subsequent steps, if the fault phase selection results at both ends If they are different, it is directly determined that the data at both ends do not match, and the process ends; if the fault phase selection results at both ends are the same, that is, the fault types at both ends match, then the load current criterion and closing angle criterion must be used for further judgment. For specific implementation, the implementation of fault phase selection can refer to: Yang Qixun. Microcomputer relay protection foundation [M]. Beijing: China Electric Power Press, 2004.

实施例对线路任一端进行的选相流程如附图3所示。图中，分别为三相的突变量电流。The phase selection process for any end of the line in the embodiment is shown in FIG. 3 . In the figure, are the abrupt currents of the three phases respectively.

首先根据步骤B1所得这一端数据各相的正常与故障电流及电压相量，计算三种相电流差的突变量然后进行相关比较与判定：First, according to the normal and fault current and voltage phasors of each phase of the data at this end obtained in step B1, calculate the sudden change of the three phase current differences Then make relevant comparisons and judgments:

当 $| {\overset{\cdot}{I}}_{B} - {\overset{\cdot}{I}}_{C} | = \min {| {\overset{\cdot}{I}}_{A} - {\overset{\cdot}{I}}_{B} |, | {\overset{\cdot}{I}}_{B} - {\overset{\cdot}{I}}_{C} |, | {\overset{\cdot}{I}}_{C} - {\overset{\cdot}{I}}_{A} |},$ 则判断 $| {\overset{\cdot}{I}}_{B} - {\overset{\cdot}{I}}_{C} | < < | {\overset{\cdot}{I}}_{C} - {\overset{\cdot}{I}}_{A} |$ 是否成立，when $| {\overset{&Center Dot;}{I}}_{B} - {\overset{\cdot}{I}}_{C} | = \min {| {\overset{\cdot}{I}}_{A} - {\overset{\cdot}{I}}_{B} |, | {\overset{&Center Dot;}{I}}_{B} - {\overset{&Center Dot;}{I}}_{C} |, | {\overset{&Center Dot;}{I}}_{C} - {\overset{&Center Dot;}{I}}_{A} |},$ then judge $| {\overset{&Center Dot;}{I}}_{B} - {\overset{&Center Dot;}{I}}_{C} | < < | {\overset{&Center Dot;}{I}}_{C} - {\overset{&Center Dot;}{I}}_{A} |$ is established,

若不成立则进入判断操作X；like If it is not established, enter the judgment operation X;

若成立则进一步判断是否成立，若成立则判定故障为A相接地，若不成立则进入判断操作X。like If it is established, it will be further judged Whether it is true, if it is true, it will be judged that the fault is A-phase grounding, if it is not true, it will enter the judgment operation X.

所述判断操作X，包括首先判断是否成立，若成立，则进一步判断是否成立，若成立则判定故障为AB两相短路，否则判定为AB两相短路接地；若不成立，同样进一步判断是否成立，若成立则判定故障为AC两相短路，否则判定为AC两相短路接地。The judgment operation X includes first judging Is it established, if established, further judge Whether it is established, if it is established, it is judged that the fault is AB two-phase short circuit, otherwise it is judged that AB two-phase short circuit is grounded; if If it is not established, it is also judged further Whether it is established, if established, it is determined that the fault is an AC two-phase short circuit, otherwise it is determined that an AC two-phase short circuit is grounded.

当 $| {\overset{\cdot}{I}}_{A} - {\overset{\cdot}{I}}_{B} | = \min {| {\overset{\cdot}{I}}_{A} - {\overset{\cdot}{I}}_{B} |, | {\overset{\cdot}{I}}_{B} - {\overset{\cdot}{I}}_{C} |, | {\overset{\cdot}{I}}_{C} - {\overset{\cdot}{I}}_{A} |}$ 或 $| {\overset{\cdot}{I}}_{C} - {\overset{\cdot}{I}}_{A} | = \min {| {\overset{\cdot}{I}}_{A} - {\overset{\cdot}{I}}_{B} |, | {\overset{\cdot}{I}}_{B} - {\overset{\cdot}{I}}_{C} |, | {\overset{\cdot}{I}}_{C} - {\overset{\cdot}{I}}_{A} |}$ 时的判定步骤与 $| {\overset{\cdot}{I}}_{A} - {\overset{\cdot}{I}}_{B} | = \min {| {\overset{\cdot}{I}}_{A} - {\overset{\cdot}{I}}_{B} |, | {\overset{\cdot}{I}}_{B} - {\overset{\cdot}{I}}_{C} |, | {\overset{\cdot}{I}}_{C} - {\overset{\cdot}{I}}_{A} |}$ 时类似，本发明不予赘述，图3中省略。when $| {\overset{&Center Dot;}{I}}_{A} - {\overset{\cdot}{I}}_{B} | = \min {| {\overset{\cdot}{I}}_{A} - {\overset{\cdot}{I}}_{B} |, | {\overset{&Center Dot;}{I}}_{B} - {\overset{&Center Dot;}{I}}_{C} |, | {\overset{&Center Dot;}{I}}_{C} - {\overset{&Center Dot;}{I}}_{A} |}$ or $| {\overset{\cdot}{I}}_{C} - {\overset{&Center Dot;}{I}}_{A} | = \min {| {\overset{&Center Dot;}{I}}_{A} - {\overset{&Center Dot;}{I}}_{B} |, | {\overset{\cdot}{I}}_{B} - {\overset{&Center Dot;}{I}}_{C} |, | {\overset{\cdot}{I}}_{C} - {\overset{&Center Dot;}{I}}_{A} |}$ When the judgment steps and $| {\overset{&Center Dot;}{I}}_{A} - {\overset{&Center Dot;}{I}}_{B} | = \min {| {\overset{\cdot}{I}}_{A} - {\overset{\cdot}{I}}_{B} |, | {\overset{\cdot}{I}}_{B} - {\overset{&Center Dot;}{I}}_{C} |, | {\overset{&Center Dot;}{I}}_{C} - {\overset{&Center Dot;}{I}}_{A} |}$ is similar, the present invention will not repeat it, and it is omitted in FIG. 3 .

步骤B3、若步骤B2判断结果为两端数据的故障类型不匹配，则直接判定两端数据不匹配，结束流程；若两端故障类型匹配，则根据步骤二中检测出的m端数据故障起点计算的m端电流和根据步骤二检测出的n端数据故障起点推算的m端电流其中包括运用线路分布参数模型由线路一侧的电量(设为n端，以下标n表示)推算对侧(设为m端，以下标m表示)电流 Step B3, if the judgment result of step B2 is that the fault types of the data at both ends do not match, then directly determine that the data at both ends do not match, and end the process; Calculated m-terminal current and the m-terminal current calculated based on the n-terminal data fault starting point detected in step 2 It includes using the distribution parameter model of the line to calculate the current on the opposite side (set to m-terminal, represented by the subscript m) from the electricity on one side of the line (set to n-terminal, represented by the subscript n)

两端均以A相为基准相，运用傅氏算法计算由检测而来的故障起点n_s1(m端相应的故障起点n_s)、n_s2(n端相应的故障起点n_s)前推一个周波的(即采样点n_s1-N与n_s2-N对应的)电流与电压相量，包括运用傅氏算法计算由步骤二所得m端的故障起点前推一个周波的电流相量，得到m端电流运用傅氏算法计算由步骤二所得n端的故障起点前推一个周波的电流相量和电压相量，分别得到n端电流和n端电压再运用线路分布参数模型由n端数据推算对应的m端电流 Both ends take phase A as the reference phase, and use Fourier algorithm to calculate the fault starting point n _s1 (corresponding fault starting point n _s at m terminal) and n _s2 (corresponding fault starting point n _s at n terminal) to push forward one The current and voltage phasors of the cycle (that is, corresponding to the sampling point n _s1 -N and n _s2 -N) include using the Fourier algorithm to calculate the current phasor of one cycle from the fault starting point of the m terminal obtained in step 2 to obtain the m terminal electric current Use the Fourier algorithm to calculate the current phasor and voltage phasor of one cycle forward from the fault starting point at the n-terminal obtained in step 2, and obtain the current at the n-terminal respectively and n-terminal voltage Then use the line distribution parameter model to calculate the corresponding m-terminal current from the n-terminal data

${\overset{\cdot &Center Dot;}{I I}}_{m m}^{' '} = = \frac{sh sh ((γ γ \times \times L L))}{{Z Z}_{c c}} \times \times {\overset{\cdot \cdot}{U u}}_{n no} - - ch ch ((γ γ \times \times L L)) \times \times {\overset{\cdot \cdot}{I I}}_{n no}$

其中，γ和z_c分别为线路的传播常数与特性阻抗，L为线路长度。Among them, γ and z _c are the propagation constant and characteristic impedance of the line respectively, and L is the line length.

步骤B4、根据负荷电流判据判断两端负荷电流是否匹配。负荷电流判据为：若|ρ-1|≤λ，则负荷电流匹配，否则负荷电流不匹配。其中，参数λ为阀值，I_m、I′_m是和的有效值。具体实施时本领域技术人员可自行预设阀值λ的取值。Step B4, judging whether the load currents at both ends match according to the load current criterion. The load current criterion is: if |ρ-1|≤λ, then the load current matches, otherwise the load current does not match. Among them, the parameter λ is the threshold, I _m and I′ _m are and valid value for . During specific implementation, those skilled in the art can preset the value of the threshold λ by themselves.

步骤B5、若步骤B4判断结果为两端负荷电流不匹配，则判定两端数据不匹配，结束流程；若两端负荷电流匹配，则利用合闸角判据判断两端数据是否匹配。合闸角判据为：若参数则合闸角匹配，否则合闸角不匹配。其中，为阀值，具体实施时本领域技术人员可自行预设阀值的取值。Step B5. If the result of step B4 is that the load currents at both ends do not match, it is determined that the data at both ends do not match, and the process ends; if the load currents at both ends match, use the closing angle criterion to determine whether the data at both ends match. The closing angle criterion is: if the parameter Then the closing angle matches, otherwise the closing angle does not match. in, is the threshold value, and those skilled in the art can preset the threshold value by themselves during specific implementation value of .

步骤B6、若步骤B5判断结果为合闸角匹配，则说明三重判据判定结果均匹配，则最终判定两端数据匹配，否则最终仍判定两端数据不匹配。Step B6. If the judgment result of step B5 is that the closing angle matches, it means that the judgment results of the three criteria all match, and it is finally judged that the data at both ends match, otherwise it is finally judged that the data at both ends do not match.

本文中所描述的具体实施例仅仅是对本发明精神作举例说明。本发明所属技术领域的技术人员可以对所描述的具体实施例做各种各样的修改或补充或采用类似的方式替代，但并不会偏离本发明的精神或者超越所附权利要求书所定义的范围。The specific embodiments described herein are merely illustrative of the spirit of the invention. Those skilled in the art to which the present invention belongs can make various modifications or supplements to the described specific embodiments or adopt similar methods to replace them, but they will not deviate from the spirit of the present invention or go beyond the definition of the appended claims range.

Claims

1. A power system fault data matching method based on a mutation point detection combination algorithm is characterized in that: comprises the following steps of (a) carrying out,

step one, acquiring fault data, wherein the fault data comprises randomly extracting a group of fault data at two ends of a certain line from a fault recording database in a communication system, and recording one end of the line as an m end and the other end as an n end;

step two, respectively calculating the corresponding fault starting points at the two ends by utilizing a catastrophe point detection combination algorithm, comprising the following substeps of respectively executing the two ends of the line,

step A1, generating a basic signal sequence based on the fault data group at one end of a certain line extracted in the step I, wherein the implementation mode is that the phase-mode conversion is carried out on three-phase current by using the K conversion:

[\begin{matrix} i_{0} \\ i_{1} \\ i_{2} \end{matrix}] = [\begin{matrix} 1 & 1 & 1 \\ 1 & 2 & - 3 \\ 1 & - 3 & 2 \end{matrix}] [\begin{matrix} i_{a} \\ i_{b} \\ i_{c} \end{matrix}]

wherein i_a、i_b、i_cFor three-phase currents, i₀、i₁、i₂Selecting one modulus component as a basic signal for the corresponding three modulus components to generate a basic signal sequence;

step A2, classifying the basic signal sequence obtained in the step A1 by using a Bayesian classification algorithm to find out suspicious mutation points;

step A3, obtaining the mutation quantity current of the basic signal sequence obtained in the step A1, and calculating the current mutation point n by using a mutation quantity current algorithm_s0；

Step A4, determining the fault starting point by applying a mutation point detection combination algorithm, comprising the step A3 of obtaining n_s0Constructing a time window [ n ] for reference_s0-Δn,n_s0]Wherein, the delta n is the preset window width, the time window is utilized to screen the fault starting point from the suspicious mutation points obtained in the step A2, the screening mode is as follows,

1) when there is only one suspected discontinuity in the time window, the fault starting point n_sFor this reason, suspected mutation points;

2) when there are two or more suspected discontinuities in the time window, the fault starting point n_sAveraging and rounding the suspicious mutation points;

3) when there is no suspicious abrupt change point in the time window, the starting point n of the fault_sThe current mutation point n obtained for the step A3_s0；

Step three, related criterion parameters are obtained according to the two-end fault starting points calculated in the step two, and whether the two-end fault data are matched or not is judged according to triple criteria of data matching, and the method comprises the following substeps:

step B1, starting point n of fault corresponding to two ends_sRespectively calculating normal and fault current and voltage phasors of each phase of data at two ends by using a Fourier algorithm as a reference;

b2, respectively applying a sudden change current phase selection method to the two ends to perform fault phase selection according to the result obtained in the step B1, and judging whether the fault types corresponding to the data at the two ends are matched;

step B3, if the result of step B2 is both end dataIf the barrier types are not matched, directly judging that the data at the two ends are not matched, and ending the process; if the current is matched, m-end current is calculatedAndas follows below, the following description will be given,

taking the A phase as a reference phase at both ends, calculating the current phasor of a cycle forward of the fault starting point of the m end obtained in the step two by using a Fourier algorithm to obtain the current phasor of the m endCalculating the current phasor and the voltage phasor of a cycle forward from the fault starting point of the n end obtained in the step two by using a Fourier algorithm to respectively obtain the current phasor and the voltage phasor of the n endAnd n terminal voltage

Then, the n-terminal current is used to generate a line distribution parameter modelAnd n terminal voltageCalculate the corresponding m-terminal current

Wherein, γ and z_cRespectively is the propagation constant and characteristic impedance of the line, and L is the line length;

b4, judging whether the load currents at two ends are matched according to a load current criterion, wherein if | rho-1 | is less than or equal to lambda, the load currents are matched; otherwise, the load currents do not match, wherein the parametersLambda is a preset threshold value;

step B5, if the result of the step B4 is that the load currents at the two ends are not matched, judging that the data at the two ends are not matched, and ending the process; if the data are matched, judging whether the data at the two ends are matched by using a closing angle criterion, wherein the parameters comprise a normal parameterThe closing angles match, otherwise the closing angles do not match, wherein,is a preset threshold value;

and step B6, if the judgment result in the step B5 is that the closing angle is matched, the judgment results of the three criteria are both matched, the data at the two ends are finally judged to be matched, otherwise, the data at the two ends are not matched.

2. The power system fault data matching method based on the mutation point detection combination algorithm as claimed in claim 1, wherein: the step A3 is realized by searching a current mutation point by using a mutation quantity current algorithm as follows:

a) calculating the abrupt variable quantity current delta i (k) according to the formula delta i (k) (i (k)) -i (k-N), wherein N is the number of sampling points of one period of the fault recording device, and i (k) is the current at the sampling point k of the basic signal i;

b) determining the current mutation point according to the detection criterion, wherein the criterion is as follows,

<math> <mfenced open='{' close=''> <mtable> <mtr> <mtd> <mo>|</mo> <mi>ψ</mi> <mrow> <mo>(</mo> <mi>k</mi> <mo>)</mo> </mrow> <mo>|</mo> <mo>></mo> <mi>ξ</mi> </mtd> </mtr> <mtr> <mtd> <mi>card</mi> <mrow> <mo>(</mo> <mi>A</mi> <mo>)</mo> </mrow> <mo>&GreaterEqual;</mo> <mi>β</mi> </mtd> </mtr> </mtable> </mfenced> </math>

wherein the parametersAnd A ═ psi (n) | | psi (n) | > xi, k is less than or equal to n and less than k + alpha }, alpha, beta and xi are preset parameters, and the first k value meeting the criterion is a mutation point corresponding to the detected signal.