CN111157850B - Mean value clustering-based power grid line fault identification method - Google Patents

Mean value clustering-based power grid line fault identification method Download PDF

Info

Publication number
CN111157850B
CN111157850B CN202010044361.1A CN202010044361A CN111157850B CN 111157850 B CN111157850 B CN 111157850B CN 202010044361 A CN202010044361 A CN 202010044361A CN 111157850 B CN111157850 B CN 111157850B
Authority
CN
China
Prior art keywords
fault
formula
line
line fault
mean value
Prior art date
Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
Active
Application number
CN202010044361.1A
Other languages
Chinese (zh)
Other versions
CN111157850A (en
Inventor
杨帆
邓一帆
李东东
徐波
林顺富
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Shanghai Electric Power University
Original Assignee
Shanghai Electric Power University
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by Shanghai Electric Power University filed Critical Shanghai Electric Power University
Priority to CN202010044361.1A priority Critical patent/CN111157850B/en
Publication of CN111157850A publication Critical patent/CN111157850A/en
Application granted granted Critical
Publication of CN111157850B publication Critical patent/CN111157850B/en
Active legal-status Critical Current
Anticipated expiration legal-status Critical

Links

Images

Classifications

    • G—PHYSICS
    • G01—MEASURING; TESTING
    • G01R—MEASURING ELECTRIC VARIABLES; MEASURING MAGNETIC VARIABLES
    • G01R31/00—Arrangements for testing electric properties; Arrangements for locating electric faults; Arrangements for electrical testing characterised by what is being tested not provided for elsewhere
    • G01R31/08—Locating faults in cables, transmission lines, or networks
    • G01R31/081—Locating faults in cables, transmission lines, or networks according to type of conductors
    • G01R31/086—Locating faults in cables, transmission lines, or networks according to type of conductors in power transmission or distribution networks, i.e. with interconnected conductors
    • G—PHYSICS
    • G01—MEASURING; TESTING
    • G01R—MEASURING ELECTRIC VARIABLES; MEASURING MAGNETIC VARIABLES
    • G01R31/00—Arrangements for testing electric properties; Arrangements for locating electric faults; Arrangements for electrical testing characterised by what is being tested not provided for elsewhere
    • G01R31/08—Locating faults in cables, transmission lines, or networks
    • G01R31/088—Aspects of digital computing

Abstract

The invention relates to a mean value clustering-based power grid line fault identification method, which extracts four abnormal state data according to a monthly report and abnormal working state information stored by a system, wherein the abnormal state of a line comprises a single-phase line short-circuit fault, a line re-trip after reclosing, a three-phase line short-circuit fault and a regulation misinformation. And extracting kurtosis, average distance percentage and variation coefficient fault characteristic quantity by fully mining and analyzing typical false alarm waveforms of all faults. And classifying the samples by using a mean value clustering algorithm, so as to realize fault identification of the line according to the abnormal state current data. Compared with the prior art, the method has the advantages of rapid diagnosis, accurate fault identification and the like.

Description

Mean value clustering-based power grid line fault identification method
Technical Field
The invention relates to the technical field of intelligent power grid fault identification, in particular to a mean value clustering-based power grid line fault identification method.
Background
The power grid dispatching fault warning system is an important support for guaranteeing safe and stable operation and dispatching of a power grid, timely identification is made on abnormal fault accidents by monitoring various data and indexes of the power grid in real time, and dispatching and fault processing real-time basis is provided for dispatching personnel and operation and maintenance personnel.
With the development of an electric power system, higher requirements are provided for automation and intellectualization of system operation, the traditional dispatching fault alarm system excessively depends on manual processing and cannot meet the requirements for power grid development, so that an intelligent fault identification system needs to be researched.
Disclosure of Invention
The invention aims to overcome the defects in the prior art and provide a mean value clustering-based power grid line fault identification method.
The purpose of the invention can be realized by the following technical scheme:
a power grid line fault identification method based on mean value clustering comprises the following steps:
step 1: extracting kurtosis, average distance percentage and variation coefficient as fault characteristic quantities according to waveform characteristics of each typical state data of the line, and constructing an intelligent line fault identification characteristic database;
step 2: clustering and dividing the identification feature database by adopting a K mean algorithm;
and step 3: and visually representing the recognition characteristic database after the clustering division is finished through a TSNE dimension reduction algorithm, and comparing an actual result with a visual result of the clustering division by operation and maintenance personnel to diagnose and recognize the power grid line fault.
Further, the kurtosis in step1 is calculated as:
Figure GDA0003394599780000021
in the formula, QXIs kurtosis, X, of a faulty sampleiIn order to sample the data at the sample points,
Figure GDA0003394599780000022
is the average value of the fault samples, and n is the number of sampling points.
Further, the calculation formula of the average distance percentage in step1 is as follows:
Figure GDA0003394599780000023
in the formula, YXIs the average distance percentage of the failed samples.
Further, the calculation formula of the coefficient of variation in step1 is:
Figure GDA0003394599780000024
in the formula, VXIs the coefficient of variation of the fault sample.
Further, in the K-means algorithm in step2, an euclidean distance calculation formula between any two of the state information quantities for each sample point is as follows:
Figure GDA0003394599780000025
in the formula, d (i, j)2Is the Euclidean distance, D, between any two of the quantities of state information for each sample pointiAnd DjRespectively, any two state information quantities of the state information quantities of each sample point.
Further, the line fault intelligent identification feature database in the step1 describes a formula as follows:
Figure GDA0003394599780000026
in the formula, D is a line fault intelligent identification characteristic database, D1,D2,…,DnThe individual elements in the feature database are intelligently identified for the line fault.
Further, the typical state data of the line in the step1 include a mediation false alarm data, a single-phase line short circuit fault data, a line reclosing re-trip fault data and a three-phase line short circuit fault data.
Further, the formula for calculating the conditional probability in the TSNE dimension reduction algorithm in step3 is as follows:
Figure GDA0003394599780000027
in the formula, qj|iFor conditional probabilities in TSNE dimension reduction algorithms, di，dj，dk，dlAnd intelligently identifying each element in the characteristic database for the line fault after dimension reduction.
Compared with the prior art, the invention has the following advantages:
(1) the method selects the TSNE dimension reduction to perform the clustering algorithm to perform the visualization of the clustering result of the high-dimensional characteristic quantity, and the TSNE is a nonlinear dimension reduction algorithm for finding the data intrinsic relation through the probability distribution of random walk on a neighborhood map for the two-dimensional visualization research. Through visual display, the division effect of fault identification can be clearly and visually displayed, and rapid diagnosis of operation and maintenance personnel is facilitated.
(2) Each fault sample comprises line current remote measurement values within 3s before and after the fault, 3000 sampling data points are counted, firstly, the waveform characteristics of each fault sample point are fully considered, three groups of characteristic quantities of kurtosis, average distance percentage and coefficient of variation are extracted, and the data characteristics of each fault type are fully mined; then, the invention adopts a mean value clustering algorithm to classify the data, thereby realizing the automatic classification and identification of the fault types.
Drawings
FIG. 1 is a flow chart of a method of the present invention;
fig. 2 is a waveform diagram of a typical reclosing re-trip fluctuation in an abnormal line condition in accordance with an embodiment of the present invention;
FIG. 3 is a waveform illustrating a three-phase fault in an abnormal line condition in accordance with an exemplary embodiment of the present invention;
FIG. 4 is a waveform diagram of a typical line fault in an abnormal condition of the present invention;
FIG. 5 is a waveform illustrating a single phase fault trip coincidence in an exemplary line anomaly condition in accordance with an embodiment of the present invention;
FIG. 6 is a TSNE dimension reduction visualization diagram of the clustering result of the present invention.
Detailed Description
The technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the drawings in the embodiments of the present invention, and it is obvious that the described embodiments are some, not all, embodiments of the present invention. All other embodiments, which can be obtained by a person skilled in the art without any inventive step based on the embodiments of the present invention, shall fall within the scope of protection of the present invention.
Examples
The method of the invention has a flow chart as shown in fig. 1, and specifically comprises the following process steps:
1. method for constructing intelligent line fault identification feature database
Four typical types of line anomaly conditions are illustrated in fig. 2, 3, 4 and 5, where telemetry monitored current amplitudes are analyzed. The amplitude, the period and the frequency of the current fluctuation degree under the same fault are similar, and the fluctuation difference under different faults is larger, so that fault characteristic quantities such as kurtosis, average distance percentage, variation coefficient and the like are extracted according to the waveform characteristics of each typical state.
Wherein, the kurtosis calculation formula is as follows:
Figure GDA0003394599780000041
in the formula, QXIs kurtosis, X, of a faulty sampleiIn order to sample the data at the sample points,
Figure GDA0003394599780000042
and n is the average value of the fault samples, and the number of sampling points.
Wherein, the average distance percentage calculation formula is as follows:
Figure GDA0003394599780000043
in the formula, YXIs the average distance percentage of the failed samples.
Wherein, the coefficient of variation computational formula is as follows:
Figure GDA0003394599780000044
in the formula, VXIs the coefficient of variation of the fault sample.
By the above feature quantity construction method, an intelligent alarm technical feature database is constructed:
Figure GDA0003394599780000045
in the formula, D is a line fault intelligent identification characteristic database, D1,D2,…,DnThe individual elements in the feature database are intelligently identified for the line fault.
2. Intelligent line fault identification technology
2.1, clustering and dividing the current fault multi-dimensional sample database by adopting a K mean algorithm
The mean clustering algorithm is a clustering algorithm based on Euclidean distance partitioning. The clustering algorithm classifies the same characteristics of the established fault current multi-dimensional samples into one class, divides the fault current multi-dimensional samples into different classes according to different characteristics, realizes the intelligent fault identification technology, and obtains the state information quantity D of each sample point1、D2……DkI.e. the amount of status information D of any two fault samplesiAnd DjThe Euclidean distance formula is as follows:
Figure GDA0003394599780000046
in the formula, d (i, j)2Is the Euclidean distance, D, between any two of the quantities of state information for each sample pointiAnd DjRespectively, any two state information quantities in the state information quantities of each sample point.
The K-means clustering algorithm comprises the following calculation steps:
(1) and randomly setting K-4 sample points from a fault sample database as an initial clustering center.
(2) And finding out a cluster suitable for each sample by calculating the Euclidean distance from the randomly set central point.
(3) And calculating the optimal clustering center of the cluster through the divided clusters.
(4) Repeating the steps 2 and 3 until the central point of the cluster is unchanged or reaches a set iteration number or reaches a set error range.
2.2 visualization of Fault identification
And the TSNE is selected for dimensionality reduction, the clustering algorithm is used for visualizing the clustering result of the high-dimensional characteristic quantity, and the TSNE is a nonlinear dimensionality reduction algorithm for finding the data intrinsic relation through the probability distribution of random walk on a neighborhood graph for two-dimensional visualization research. Through visual display, the division effect of fault identification can be clearly and visually displayed, the rapid diagnosis of operation and maintenance personnel is facilitated, and the specific algorithm is as follows:
(1) random adjacency embedding (SNE) begins by converting the high-dimensional euclidean distances between data points to conditional probabilities representing similarity, Di、DjConditional probability p between (arbitrary two feature quantities)j|iGiven by:
Figure GDA0003394599780000051
wherein sigmaiIs the data point xiA gaussian variance at the center.
(2) For d of lower dimensioniAnd djData, q can be calculated as wellj|iTo achieve dimension reduction. Set its variance
Figure GDA0003394599780000052
Figure GDA0003394599780000053
With a conditional probability of qj|i：
Figure GDA0003394599780000054
(3) T-SNE adopts T distribution to solve the problem of light data congestion in a low-dimensional space. Thus p isj|iThe formula after adopting T distribution is as follows:
Figure GDA0003394599780000055
in the formula, qj|iFor conditional probabilities in TSNE dimension reduction algorithms, di，dj，dk，dlAnd intelligently identifying each element in the characteristic database for the line fault after dimension reduction.
The practical implementation scheme is as follows:
the invention selects 44 groups of sample data, each sample contains fault information within 3s before and after the fault, and 3000 sampling data points are provided. The fault detection method comprises the following steps of respectively obtaining 21 groups of debugging and false alarm data, 10 groups of single-phase line short-circuit fault data, 6 groups of line reclosing and then tripping fault data and 7 groups of three-phase line short-circuit fault data.
Step 1: constructing a line fault intelligent identification feature database:
step 2: clustering and dividing a current fault multi-dimensional sample database by adopting a K mean algorithm;
step 3: the visual result of current fault data clustering by the TSNE dimension reduction algorithm is shown in FIG. 6
As shown in the figure, the scheme can rapidly and clearly identify each current fault state, the clustering division result of the scheme is consistent with the actual result, and operation and maintenance personnel can rapidly diagnose and identify the current fault.
While the invention has been described with reference to specific embodiments, the invention is not limited thereto, and various equivalent modifications and substitutions can be easily made by those skilled in the art within the technical scope of the invention. Therefore, the protection scope of the present invention shall be subject to the protection scope of the claims.

Claims (6)

1. A mean value clustering-based power grid line fault identification method is characterized by comprising the following steps:
step 1: extracting kurtosis, average distance percentage and variation coefficient as fault characteristic quantities according to waveform characteristics of each typical state data of the line, and constructing an intelligent line fault identification characteristic database;
step 2: clustering and dividing the identification feature database by adopting a K mean algorithm;
and 3, step 3: visually representing the recognition characteristic database after the clustering division is finished through a TSNE dimension reduction algorithm, and comparing an actual result with a visual result of the clustering division by operation and maintenance personnel to diagnose and recognize the power grid line fault;
each typical state data of the line in the step1 comprises regulation misinformation data, single-phase line short-circuit fault data, line reclosing re-trip fault data and three-phase line short-circuit fault data, and each typical state data is a line current remote measurement value in a corresponding typical state;
the kurtosis in the step1 is calculated by the following formula:
Figure FDA0003394599770000011
in the formula, QXIs kurtosis, X, of a faulty sampleiIn order to sample the data at the sample points,
Figure FDA0003394599770000012
is the average value of the fault samples, and n is the number of sampling points.
2. The method for identifying the grid line fault based on the mean value clustering as claimed in claim 1, wherein the calculation formula of the average distance percentage in the step1 is as follows:
Figure FDA0003394599770000013
in the formula, YXIs the average distance percentage of the failed samples.
3. The method for identifying the grid line fault based on the mean value clustering as claimed in claim 1, wherein the coefficient of variation in the step1 is calculated by the following formula:
Figure FDA0003394599770000014
in the formula, VXIs the coefficient of variation of the fault sample.
4. The method for identifying the power grid line fault based on the mean value clustering as claimed in claim 1, wherein the Euclidean distance calculation formula between any two of the state information quantities of each sample point in the K-means algorithm in the step2 is as follows:
Figure FDA0003394599770000021
in the formula, d (i, j)2Is the Euclidean distance, D, between any two of the quantities of state information for each sample pointiAnd DjRespectively, any two state information quantities of the state information quantities of each sample point.
5. The method for identifying the line fault of the power grid based on the mean value clustering as claimed in claim 1, wherein the line fault intelligent identification feature database in the step1 is described by the formula:
Figure FDA0003394599770000022
in the formula, D is a line fault intelligent identification characteristic database, D1,D2,…,DnThe individual elements in the signature database are intelligently identified for a line fault.
6. The method for identifying the grid line fault based on the mean value clustering according to claim 1, wherein the conditional probability in the TSNE dimension reduction algorithm in the step3 is calculated according to the formula:
Figure FDA0003394599770000023
in the formula, qj|iFor conditional probabilities in TSNE dimension reduction algorithms, di，dj，dk，dlAnd intelligently identifying each element in the characteristic database for the line fault after dimension reduction.
CN202010044361.1A 2020-01-15 2020-01-15 Mean value clustering-based power grid line fault identification method Active CN111157850B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202010044361.1A CN111157850B (en) 2020-01-15 2020-01-15 Mean value clustering-based power grid line fault identification method

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202010044361.1A CN111157850B (en) 2020-01-15 2020-01-15 Mean value clustering-based power grid line fault identification method

Publications (2)

Publication Number Publication Date
CN111157850A CN111157850A (en) 2020-05-15
CN111157850B true CN111157850B (en) 2022-06-21

Family

ID=70563248

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202010044361.1A Active CN111157850B (en) 2020-01-15 2020-01-15 Mean value clustering-based power grid line fault identification method

Country Status (1)

Country Link
CN (1) CN111157850B (en)

Families Citing this family (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN111913081B (en) * 2020-07-14 2023-05-02 上海电力大学 Mean shift clustering-based switch cabinet insulation state anomaly detection method
CN112505481A (en) * 2020-11-20 2021-03-16 云南电网有限责任公司普洱供电局 35kV power line fault traveling wave extraction method based on neighbor propagation clustering
CN113780354B (en) * 2021-08-11 2024-01-23 国网上海市电力公司 Remote measurement data anomaly identification method and device for dispatching automation master station system
CN116821834B (en) * 2023-08-29 2023-11-24 浙江北岛科技有限公司 Vacuum circuit breaker overhauling management system based on embedded sensor

Citations (7)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN101477172A (en) * 2009-02-18 2009-07-08 湖南大学 Analogue circuit fault diagnosis method based on neural network
CN101872963A (en) * 2009-04-24 2010-10-27 北京石油化工学院 Supply line transient state protection method based on wave shape identification and device thereof
CN103267941A (en) * 2013-05-07 2013-08-28 长沙学院 Method for testing fault modes of integrated switching current circuit
CN104198899A (en) * 2014-08-04 2014-12-10 西安交通大学 Partial discharge type identifying method for transformer under multi-discharge source situation
KR20170019831A (en) * 2015-08-13 2017-02-22 한국수력원자력 주식회사 Partial Discharge Defects Diagnostic System
CN106771898A (en) * 2016-11-27 2017-05-31 福州大学 Series fault arc detection device and its method based on Higher Order Cumulants identification
CN108808636A (en) * 2018-05-29 2018-11-13 西南交通大学 A kind of multi-terminal HVDC transmission line protection method based on transfer entropy

Family Cites Families (8)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
JP2003223474A (en) * 2002-01-30 2003-08-08 Pfu Ltd Bs cluster test diagnosis rate improving system
KR101268761B1 (en) * 2011-09-19 2013-05-29 한국전력공사 Method for deciding defect type of partial discharge and apparatus thereof
CN103941163B (en) * 2014-05-12 2017-01-18 福州大学 Resonant earthed system fault line selection method utilizing fuzzy K-means clustering
CN109100646A (en) * 2018-08-17 2018-12-28 国网江苏省电力有限公司检修分公司 A kind of Fault Diagnosis for HV Circuit Breakers method
CN109307824B (en) * 2018-09-18 2020-11-03 福州大学 Clustering-based power distribution network single-phase earth fault section positioning method
CN109765490B (en) * 2018-11-12 2020-08-04 北京理工大学 Power battery fault detection method and system based on high-dimensional data diagnosis
CN110133488B (en) * 2019-04-09 2021-10-08 上海电力学院 Switch cabinet health state evaluation method and device based on optimal grade number
CN110225107B (en) * 2019-06-04 2020-04-07 上海锐测电子科技有限公司 Cable comprehensive detection system

Patent Citations (7)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN101477172A (en) * 2009-02-18 2009-07-08 湖南大学 Analogue circuit fault diagnosis method based on neural network
CN101872963A (en) * 2009-04-24 2010-10-27 北京石油化工学院 Supply line transient state protection method based on wave shape identification and device thereof
CN103267941A (en) * 2013-05-07 2013-08-28 长沙学院 Method for testing fault modes of integrated switching current circuit
CN104198899A (en) * 2014-08-04 2014-12-10 西安交通大学 Partial discharge type identifying method for transformer under multi-discharge source situation
KR20170019831A (en) * 2015-08-13 2017-02-22 한국수력원자력 주식회사 Partial Discharge Defects Diagnostic System
CN106771898A (en) * 2016-11-27 2017-05-31 福州大学 Series fault arc detection device and its method based on Higher Order Cumulants identification
CN108808636A (en) * 2018-05-29 2018-11-13 西南交通大学 A kind of multi-terminal HVDC transmission line protection method based on transfer entropy

Non-Patent Citations (1)

* Cited by examiner, † Cited by third party
Title
Investigation of Feature Effectiveness in Polymer Electrolyte Membrane Fuel Cell Fault Diagnosis;Weitao Pan et al.;《2019 Prognostics and System Health Management Conference》;20191226;1-5 *

Also Published As

Publication number Publication date
CN111157850A (en) 2020-05-15

Similar Documents

Publication Publication Date Title
CN111157850B (en) Mean value clustering-based power grid line fault identification method
CN107742127B (en) Improved electricity stealing prevention intelligent early warning system and method
CN109933905B (en) Mechanical equipment health state assessment method based on multi-dimensional early warning analysis
CN108985380B (en) Point switch fault identification method based on cluster integration
CN110059845B (en) Metering device clock error trend prediction method based on time sequence evolution gene model
CN110580492A (en) Track circuit fault precursor discovery method based on small fluctuation detection
CN103593461A (en) Disturbance source recognition method based on typical early warning event waveform bank
CN107784393A (en) A kind of the defects of transmission line of electricity Forecasting Methodology and device
CN112464439A (en) Three-layer data mining-based power system fault type analysis method and device
CN113866455A (en) Bridge acceleration monitoring data anomaly detection method, system and device based on deep learning
CN110889441A (en) Distance and point density based substation equipment data anomaly identification method
CN111999591B (en) Method for identifying abnormal state of primary equipment of power distribution network
CN112036087A (en) Multi-strategy fused nuclear power key equipment fault diagnosis method and system
CN117148045A (en) Fault studying and judging management system for running state of power distribution network
CN109000924B (en) Method for monitoring state of ball screw pair based on K mean value
CN116910128A (en) Wind turbine generator hard and soft fault data analysis system and analysis method thereof
CN116502043A (en) Finish rolling motor state analysis method based on isolated forest algorithm
CN115267447A (en) GIS insulation defect identification method based on partial discharge statistical characteristic quantity and radar spectrogram
CN115524002A (en) Running state early warning method and system for power plant rotating equipment and storage medium
CN115561575A (en) Method for distinguishing electrical abnormal state of offshore wind farm based on multi-dimensional matrix profile
Zhai et al. Improved Fault Classification Method in Transmission Line based on K-means Clustering
CN113780354A (en) Telemetry data anomaly identification method and device for dispatching automation master station system
CN112733878A (en) Transformer fault diagnosis method based on kmeans-SVM algorithm
CN111476392A (en) Metering device clock error trend prediction method based on social perception
CN117560300B (en) Intelligent internet of things flow prediction and optimization system

Legal Events

Date Code Title Description
PB01 Publication
PB01 Publication
SE01 Entry into force of request for substantive examination
SE01 Entry into force of request for substantive examination
GR01 Patent grant
GR01 Patent grant