CN-121978464-A - Distribution network ground fault positioning method based on bidirectional translation cross-correlation and waveform polarity recognition

CN121978464ACN 121978464 ACN121978464 ACN 121978464ACN-121978464-A

Abstract

A power distribution network ground fault positioning method based on bidirectional translation cross correlation and waveform polarity recognition relates to the technical field of relay protection of power systems and power distribution network fault diagnosis. The application aims to solve the problem that the positioning accuracy of the existing power distribution network ground fault positioning method cannot meet the requirement. When zero sequence voltage mutation occurs in the power distribution network, three-phase current data of all measuring nodes in the power distribution network are collected to form an original data table, the topological connection relation of the power distribution network is read, each element in the original data table is subjected to Carnbiotic transformation and wavelet denoising processing to obtain a pure transient zero mode current sequence, the pure transient zero mode current sequences of two adjacent measuring nodes are subjected to bidirectional translational sliding to find an optimal time shifting point, the waveform comprehensive characteristics of all measuring node pairs are obtained through the energy effective values of the sequences corresponding to the two adjacent measuring nodes, and whether links corresponding to all measuring node pairs are fault intervals is judged based on the waveform comprehensive characteristics and a preset fault criterion.

Inventors

WEI QIANG
YANG ZIQI
WANG KEQIANG
LI DEYUAN
XU WENLONG
ZHU WENLONG
GAO GUANGYU
TAN JINGPENG
XU DAMING
LIU JIZHE

Assignees

国网吉林省电力有限公司白城供电公司

Dates

Publication Date: 20260505
Application Date: 20260310

Claims (10)

1. The power distribution network ground fault positioning method based on bidirectional translation cross correlation and waveform polarity recognition is characterized by comprising the following steps of: When a zero sequence voltage mutation occurs in a power distribution network, acquiring three-phase current data of each measuring node in the power distribution network to form an original data table, and reading a topological connection relation of the power distribution network, wherein each element in the original data table comprises a terminal number and a three-phase current array, the terminal number is used for representing the topological position of the measuring node, and the three-phase current array is three-phase high-frequency current data of the measuring node in an acquisition time window; carrying out the Carleno's conversion and wavelet denoising treatment on each element in the original data table to obtain a pure transient zero-mode current sequence; Traversing the topological connection relation of the power distribution network, carrying out bidirectional translational sliding on the pure transient zero-mode current sequences of two adjacent measuring nodes to find an optimal time shift point, and obtaining the waveform comprehensive characteristics of all measuring node pairs through the energy effective values of the corresponding sequences of the two adjacent measuring nodes; Based on the waveform comprehensive characteristics, whether the link corresponding to each measuring node pair is a fault interval or not is judged by combining with a preset fault criterion.
2. The method for positioning a power distribution network ground fault based on bidirectional translational cross correlation and waveform polarity recognition according to claim 1, wherein the performing a karnbauer transformation and wavelet denoising process on each element in the original data table to obtain a pure transient zero-mode current sequence comprises: The first data table is provided with Three-phase high-frequency current data of each element is input to a Carleno Baud conversion and wavelet denoising module to carry out the Carleno Baud conversion and wavelet denoising processing to obtain the first element Transient zero-mode currents corresponding to the individual elements; Using the first of the original data tables The terminal numbers of the individual elements and the transient zero-mode current construct a pure transient zero-mode current element; repeating the steps until all elements in the original data table obtain corresponding pure transient zero-mode current elements; and forming a pure transient zero-mode current sequence by using pure transient zero-mode current elements corresponding to all elements in the original data table.
3. The method for positioning a power distribution network ground fault based on bi-directional translational cross-correlation and waveform polarity recognition according to claim 2, wherein the karun Bao E transform and wavelet denoising module comprises: The Karen Bao E transformation unit is used for performing Karenbauer transformation on the three-phase high-frequency current data of each element in the original data table, and extracting an initial transient zero-mode current sequence through a decoupling transformation matrix; A resolution wavelet transform unit for decomposing the initial transient zero-mode current sequence into discrete approximation and detail components using a preset wavelet basis function, And the reconstruction unit is used for carrying out signal reconstruction after the detail component is set to zero and filtering high-frequency harmonic waves and noise.
4. The method for positioning a power distribution network ground fault based on bidirectional translational cross correlation and waveform polarity recognition according to claim 1, wherein traversing the topological connection relation of the power distribution network, performing bidirectional translational sliding on pure transient zero-mode current sequences of two adjacent measurement nodes to find an optimal time shift point, and obtaining waveform comprehensive characteristics of all measurement node pairs by energy effective values of corresponding sequences of the two adjacent measurement nodes, comprises: reading the first topological connection relation of the power distribution network Upstream terminal number and downstream terminal number of the link; searching elements with terminal numbers equal to the upstream terminal number and the downstream terminal number in the pure transient zero-mode current sequence; Inputting the two elements to a waveform similarity cross-correlation calculation module to perform bidirectional translation sliding calculation to obtain the first element Waveform synthesis characteristics corresponding to the links; Repeating the steps until all links in the topological connection relation of the power distribution network obtain the corresponding waveform comprehensive characteristics.
5. The method for positioning a power distribution network ground fault based on bidirectional translational cross-correlation and waveform polarity recognition according to claim 4, wherein the two elements searched for are input to a waveform similarity cross-correlation calculation module to perform bidirectional translational sliding calculation to obtain the first element The waveform synthesis characteristic corresponding to the link comprises: Taking the two searched elements as an upstream sequence and a downstream sequence respectively, and carrying out bidirectional translation sliding calculation on the upstream sequence and the downstream sequence to obtain cross correlation coefficients under different time shifts; Taking the time shift point corresponding to the time shift point with the maximum absolute value of the cross correlation coefficient as an optimal time shift point, extracting the cross correlation coefficient corresponding to the optimal time shift point as an original cross correlation coefficient, and further obtaining a polarity symbol of the original cross correlation coefficient; Respectively calculating energy effective values of the upstream sequence and the downstream sequence; taking the deviation degree between the energy effective values of the upstream sequence and the downstream sequence as an amplitude correction coefficient; Carrying out amplitude weight correction on the original cross-correlation coefficient by using the amplitude correction coefficient to obtain an absolute comprehensive similarity coefficient considering only waveform shape approximation; and constructing polarity symbols of the absolute integrated similarity coefficient and the original cross-correlation coefficient as waveform integrated characteristics.
6. The method for positioning a power distribution network ground fault based on bidirectional translational cross-correlation and waveform polarity recognition according to claim 5, wherein the performing bidirectional translational sliding calculation on the upstream sequence and the downstream sequence to obtain cross-correlation coefficients under different time shifts comprises: performing a bi-directional translational slip calculation on the upstream and downstream sequences according to the following formula: ; Wherein, the Is a translation step length, and , The number of time shift points is the maximum allowable synchronization error, Representing translation step size The cross-correlation coefficient below; And Representing an upstream sequence score and a downstream sequence score, respectively, , For the number of sampling points of the elements found, And The decomposition is to the average of the upstream and downstream sequences.
7. The method for positioning a power distribution network ground fault based on bi-directional translational cross-correlation and waveform polarity recognition according to claim 5, wherein the expression of the polarity sign of the original cross-correlation coefficient is: , Wherein, the A sign of the polarity is indicated, The sign function is represented by a sign function, Is the original cross-correlation coefficient.
8. The method for positioning a ground fault of a power distribution network based on bi-directional translational cross-correlation and waveform polarity recognition according to claim 6, wherein said calculating energy effective values of said upstream sequence and downstream sequence, respectively, comprises: Calculating the energy efficiency values of the upstream and downstream sequences according to the following formula: , , Wherein, the And Respectively is And Is a function of the energy efficiency value of (a).
9. The method for positioning a power distribution network ground fault based on bi-directional translational cross-correlation and waveform polarity recognition according to claim 8, wherein the amplitude correction factor is calculated according to the following formula: , Wherein, the For the magnitude correction factor, And Respectively representing the minimum value and the maximum value; and carrying out amplitude weight correction on the original cross-correlation coefficient according to the following steps: , Wherein, the To consider only the absolute integrated similarity coefficient of waveform shape approximation, Is the original cross-correlation coefficient.
10. The method for positioning a power distribution network ground fault based on bidirectional translational cross-correlation and waveform polarity recognition according to claim 9, wherein the determining whether the link corresponding to each measurement node pair is a fault interval based on the waveform comprehensive characteristics in combination with a preset fault criterion comprises: the preset fault criteria include: If any one of the following inequality terms is satisfied, it is determined that the link is abnormal: Condition one: , Is a safety threshold; Condition II: , Representing the polarity symbol.

Description

Distribution network ground fault positioning method based on bidirectional translation cross-correlation and waveform polarity recognition Technical Field The application belongs to the technical field of relay protection of a power system and fault diagnosis of a power distribution network. Background With the deep construction of smart power grids, the requirements of the whole society on power supply reliability are increasingly improved. However, high-resistance ground faults frequently occur in the power distribution network, serious safety accidents such as equipment burnout and even forest fires are extremely easy to occur, and a great threat is formed to the safety and social stable operation of the power grid. The traditional manual blind line inspection and obstacle removal mode can not meet the urgent requirement of high-efficiency operation in the modern society. The main means for locating the ground fault of the power distribution network by using the monitoring data is to collect the operation data deployed on the line by using a sensor and analyze the data by using a feature extraction or algorithm model. The main methods adopted at present comprise: 1) Processing is performed using single point detection or a conventional fault indicator. Because the high-resistance fault signal is very weak, a large number of distributed capacitors exist in the power distribution network, the local detection is very easy to be subjected to serious noise interference to cause missed judgment, and the characteristic extraction is very difficult compared with the complex refraction and reflection interference condition of the travelling wave head, the processing mode can only judge approximately branch sections, and the positioning accuracy is low. 2) The method can relatively effectively position based on the difference of upstream and downstream waveforms by utilizing the transient zero-mode current waveform similarity, but the traditional algorithm does not consider the situation that the correlation coefficient is negative, namely polarity is reversed, and the problem of asynchronous is only processed by signal unidirectional translation, the theoretical result of the mode under ideal working conditions is usually better, but in practical application, the method cannot eliminate complex bidirectional communication synchronous errors to cause positioning failure. Especially, under the complex condition of small difference of upstream and downstream capacitance to ground, the current frequency and amplitude difference at two sides of the fault point are extremely small and only have opposite polarities, and the effects of the two methods are poor. 3) The method has the greatest problem that objective influence of amplitude attenuation of upstream and downstream signals is completely ignored, a multidimensional correction mechanism aiming at high-resistance grounding is lacked, and the model is extremely easy to be misled by complex working condition changes, so that misoperation or positioning failure phenomenon is caused, and the final positioning precision cannot meet the requirement. Disclosure of Invention The application aims to solve the problem that the positioning precision of the existing power distribution network ground fault positioning method can not meet the requirement, and provides the power distribution network ground fault positioning method based on bidirectional translational cross-correlation and waveform polarity recognition, which utilizes a wide-area high-precision synchronous measurement terminal to acquire transient information, through bidirectional translation cross-correlation calculation and waveform polarity identification, correlation coefficient polarity and amplitude weight correction are comprehensively considered to adapt to complex working conditions such as high-resistance grounding of a power distribution network and synchronous errors caused by communication delay, and accurate positioning of power distribution network ground faults is achieved. The first aspect of the application provides a power distribution network ground fault positioning method based on bidirectional translation cross correlation and waveform polarity recognition, which comprises the following steps: When a zero sequence voltage mutation occurs in a power distribution network, acquiring three-phase current data of each measuring node in the power distribution network to form an original data table, and reading a topological connection relation of the power distribution network, wherein each element in the original data table comprises a terminal number and a three-phase current array, the terminal number is used for representing the topological position of the measuring node, and the three-phase current array is three-phase high-frequency current data of the measuring node in an acquisition time window; carrying out the Carleno's conversion and wavelet denoising treatment on each element in the