CN111596167A - Fault section positioning method and device based on fuzzy c-means clustering algorithm - Google Patents
Fault section positioning method and device based on fuzzy c-means clustering algorithm Download PDFInfo
- Publication number
- CN111596167A CN111596167A CN202010407081.2A CN202010407081A CN111596167A CN 111596167 A CN111596167 A CN 111596167A CN 202010407081 A CN202010407081 A CN 202010407081A CN 111596167 A CN111596167 A CN 111596167A
- Authority
- CN
- China
- Prior art keywords
- fault
- detected
- sample
- class
- preprocessed
- 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.)
- Pending
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
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/23—Clustering techniques
- G06F18/232—Non-hierarchical techniques
- G06F18/2321—Non-hierarchical techniques using statistics or function optimisation, e.g. modelling of probability density functions
- G06F18/23211—Non-hierarchical techniques using statistics or function optimisation, e.g. modelling of probability density functions with adaptive number of clusters
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- Data Mining & Analysis (AREA)
- Life Sciences & Earth Sciences (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Bioinformatics & Computational Biology (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Evolutionary Biology (AREA)
- Evolutionary Computation (AREA)
- General Engineering & Computer Science (AREA)
- Artificial Intelligence (AREA)
- Probability & Statistics with Applications (AREA)
- Mathematical Physics (AREA)
- Locating Faults (AREA)
Abstract
The invention discloses a fault section positioning method and a fault section positioning device based on a fuzzy c-means clustering algorithm, wherein the method comprises the following steps: acquiring fault characteristic vectors of n nodes distributed in the whole network, and constructing a sample set to be tested; carrying out standardized pretreatment on each sample to be detected in a sample set to be detected; dividing the n preprocessed samples to be detected into a fault class and a non-fault class based on a fuzzy c-means clustering algorithm, and calculating to obtain clustering centers of the fault class and the non-fault class; calculating the distance between each preprocessed sample to be tested and the clustering centers of the fault class and the non-fault class, and dividing the node corresponding to each sample to be tested into the fault class or the non-fault class according to the size relation of the two; searching a fault characteristic information transmission path, determining the position of a fault information source, and realizing the positioning of a fault section. The idea of comprehensively analyzing the multi-source fault characteristic quantity is adopted, and the reliability of the data is improved.
Description
Technical Field
The invention relates to the technical field of power distribution networks, in particular to a fault section positioning method and device based on a fuzzy c-means clustering algorithm.
Background
The core idea of the traditional protection method is to compare a fault characteristic quantity extracted from real-time information of a measuring point with a preset protection setting value or a setting curve, and once the size of the measured characteristic quantity exceeds a range limited by the setting value, the fault is judged to occur and a protection device acts. However, the effect of these methods based on protection setting values in practical applications is not very ideal. The method is mainly limited by the following aspects: when a small current earth fault occurs, the system impedance is large, so that when a single-phase earth fault occurs, the fault current is small, the characteristic quantity serving as a protection criterion is weak, and the detection has great difficulty; the operation mode of the power distribution network is complex and changeable, and the preset protection setting value is difficult to be adaptively adjusted under different operation conditions; the fault characteristics expressed under different fault conditions are different, the change range of the fault characteristic quantity is large, and accurate detection is difficult; in a field environment, fault signals are affected by electromagnetic interference, unbalanced current of a system, load current submergence and the like, and the signal-to-noise ratio is not high; distributed Generation (DG) is gradually improved in permeability in a power distribution network, a passive power distribution network with a traditional unidirectional power flow is gradually changed into an active power distribution network with a bidirectional power flow, and the change range of fault characteristic quantity is further increased.
With the development of power distribution network fault section positioning research, a fault section positioning method based on an intelligent algorithm becomes a new research hotspot. For the intelligent algorithm with self-learning capability, the description and judgment of the feeder line running state can be realized by training a large number of fault samples on line under the condition of not predicting the mathematical model of a study object, and the positioning precision of a fault section is improved. The artificial neural network algorithm has the advantages that the data training process is convenient to operate, the application range is wide, and a plurality of scholars develop research in this respect. Related documents propose a scheme for realizing fault location by means of an artificial neural network model, firstly, wavelet packets are adopted to process extracted transient process signals, and then, the artificial neural network is utilized to perform off-line training on fault information, so that a fault location result is obtained. The related documents carry out the specialization processing on the structure of the BP neural network, and organically integrate a wavelet transformation layer and a fuzzy processing layer structure on the basis of the original network structure, so that the fault identification accuracy is improved. Relevant documents deeply analyze transient current of a non-fault phase, and a novel single-phase earth fault positioning method is provided, wherein fault positioning is realized on the basis of non-fault phase current transient quantity of a fault line, transient fault information is processed by utilizing wavelet transformation, and then analysis is carried out by matching with an artificial neural network, so that the influence of factors such as change of a system operation mode, change of fault conditions and the like on fault positioning can be effectively reduced. Based on the above analysis, it can be seen that based on these new mathematical methods, experts and scholars at home and abroad make intensive studies on the positioning of the fault section of the low-current grounding system, and a new breakthrough idea is continuously introduced into the field of positioning of the fault section of the low-current grounding system, such as introducing wavelet analysis, Prony, genetic algorithm, expert system, artificial neural network, fuzzy theory, etc. into the field of positioning of the fault section of the low-current system.
Although the accuracy and reliability of fault section positioning are improved to a certain extent by introducing an intelligent algorithm, the problem of fault section positioning of a low-current grounding system is not properly solved so far. The inventor finds that for specific engineering practice problems, even if an advanced intelligent mathematical method is used as a support, the research on the mechanism and principle of the ground fault of the power distribution network is not neglected, and how to research a new fault section positioning method suitable for various fault conditions under different operation modes is particularly important from the fault characteristics of the ground fault.
Disclosure of Invention
The invention provides a fault section positioning method and device based on a fuzzy c-means clustering algorithm, and aims to solve the problem of low accuracy and reliability caused by the fact that the ground fault characteristics of a power distribution network are not deeply explored in the prior art.
In a first aspect, a fault section positioning method based on a fuzzy c-means clustering algorithm is provided, which includes:
acquiring a plurality of fault characteristic quantities distributed on n nodes of the whole network, wherein the plurality of fault characteristic quantities of each node form a fault characteristic vector, and the fault characteristic vector is used as a sample to be detected to construct a sample set to be detected;
carrying out standardized pretreatment on each sample to be detected in a sample set to be detected;
constructing a high-dimensional fault feature space, projecting n samples to be detected into the high-dimensional fault feature space based on a plurality of fault feature quantities of each preprocessed sample to be detected, dividing the n preprocessed samples to be detected into a fault class and a non-fault class based on a fuzzy c-means clustering algorithm, and calculating to obtain clustering centers of the fault class and the non-fault class;
respectively calculating the distance d between each preprocessed sample to be detected and the clustering centers of the fault class and the non-fault classk1And dk2According to dk1And dk2The size relation divides the nodes corresponding to each preprocessed sample to be tested into a fault class or a non-fault class;
and searching a fault characteristic information transmission path based on the fault category to which each node belongs, determining the position of the fault information source, and realizing the positioning of the fault section.
Further, the step of performing a standardized preprocessing on each sample to be tested in the sample set to be tested includes:
carrying out standardization pretreatment on a sample to be detected according to the following formula:
wherein x isk′jThe jth fault characteristic quantity value of the kth sample to be tested,is the mean value of the jth fault characteristic quantity, S (x)j') standard deviation of the jth fault characteristic quantity, xkjAnd s is the total number of the fault characteristic quantities acquired by each node for the jth fault characteristic quantity value of the kth sample to be detected after the normalization pretreatment.
Further, the dividing n preprocessed samples to be tested into a fault class and a non-fault class based on the fuzzy c-means clustering algorithm, and calculating to obtain the clustering centers of the fault class and the non-fault class includes:
dynamically clustering all preprocessed samples to be detected by the aid of balance iterative equations (5) and (6) through the following optimization objective functions (4), and obtaining clustering centers of fault classes and non-fault classes:
wherein c is the number of clustering categories, and c is 2; mu.sik∈[0,1]Represents the k-th sample x to be measured after the pretreatmentkMembership degree belonging to the ith cluster typepiFor cluster centers, i takes 1 or 2, p1For fault class centers, p2Is a non-fault class center; the | l | · | is a matrix norm representing the spatial distance between the preprocessed sample to be detected and the clustering center; m is a weighting index, and m is 2; u is a membership matrix formed by membership degrees of all preprocessed samples to be detected, and P is all clustering centers PiFormed cluster center matrix, Jm(U, P) is a clustering loss function, MfcA space is divided for the fuzzy c of the preprocessed sample to be detected,Rsall membership matrices generated in the iterative process; the end conditions of the iterative process of optimizing the objective function (4) are as follows:and w is the current iteration frequency and is a preset iteration stop threshold value, and 1.0e-6 is taken.
Further, the distance d between each preprocessed sample to be detected and the clustering centers of the fault class and the non-fault class is calculated respectivelyk1And dk2According to dk1And dk2The size relationship of (2) divides the nodes corresponding to each preprocessed sample to be tested into a fault class or a non-fault class, and comprises the following steps:
calculating the Euclidean distance between each preprocessed sample to be detected and the clustering centers of the fault class and the non-fault class according to a formula (7):
wherein d isk1Representing the distance between the preprocessed sample to be tested and the center of the fault class, dk2Representing the distance, x, between the preprocessed sample to be tested and the non-fault centerkjFor normalizing the jth fault characteristic quantity value p of the kth sample to be tested after pretreatmentijRepresenting the clustering center of the jth fault characteristic quantity of the fault type or the non-fault type, wherein s is the total number of the fault characteristic quantities acquired by each node;
if d isk1>dk2If so, the node corresponding to the preprocessed sample to be detected belongs to the non-fault class;
if d isk1<dk2And if so, the node corresponding to the preprocessed sample to be detected belongs to the fault class.
Further, the plurality of fault characteristic quantities include a plurality of steady-state characteristic quantities and a plurality of transient characteristic quantities; the steady state characteristic quantities at least comprise two of a zero sequence admittance angle xk1, a negative sequence current xk2, a zero sequence current xk3, amplitude variation of each phase voltage and phase angle variation of each phase voltage; the plurality of transient characteristic quantities at least comprise two of a zero-sequence current modulus maximum value xk4 after db3 wavelet transformation, a zero-sequence energy function value xk5 in a half power frequency cycle after the start of a fault and a zero-sequence active value xk6 in a power frequency cycle after the fault.
Further, the plurality of fault characteristic quantities comprise three steady-state characteristic quantities and two transient-state characteristic quantities; the three steady state characteristic quantities comprise a zero sequence admittance angle xk1, a negative sequence current xk2 and a zero sequence current xk 3; the two transient characteristic quantities comprise a zero-sequence current modulus maximum value xk4 after db3 wavelet transformation and a zero-sequence energy function value xk5 in a half power frequency cycle after the start of a fault.
Further, before the acquiring the plurality of fault characteristic quantities distributed to the n nodes of the whole network, the method further includes:
acquiring zero sequence voltage in a power distribution network system in real time;
and when the zero sequence voltage in the power distribution network system is larger than 15% of the phase voltage, entering a single-phase earth fault section positioning process of the power distribution network.
Further, the plurality of fault characteristic quantities of the n nodes are extracted by n FTUs (distribution automation terminal units) distributed corresponding to the n node positions, and each FTU is arranged at an outlet of a corresponding protected feeder.
In a second aspect, a fault section locating device based on a fuzzy c-means clustering algorithm is provided, which includes:
a sample set acquisition unit to be tested: acquiring a plurality of fault characteristic quantities distributed on n nodes of the whole network, wherein the plurality of fault characteristic quantities of each node form a fault characteristic vector, and the fault characteristic vector is used as a sample to be detected to construct a sample set to be detected;
a pretreatment unit: carrying out standardized pretreatment on each sample to be detected in a sample set to be detected;
a clustering center obtaining unit: constructing a high-dimensional fault feature space, projecting n samples to be detected into the high-dimensional fault feature space based on a plurality of fault feature quantities of each preprocessed sample to be detected, dividing the n preprocessed samples to be detected into a fault class and a non-fault class based on a fuzzy c-means clustering algorithm, and calculating to obtain clustering centers of the fault class and the non-fault class;
a node category division unit: respectively calculating the distance between each preprocessed sample to be tested and the clustering centers of the fault class and the non-fault classFrom dk1And dk2According to dk1And dk2The size relation divides the nodes corresponding to each preprocessed sample to be tested into a fault class or a non-fault class;
a fault location unit: and searching a fault characteristic information transmission path based on the fault category to which each node belongs, determining the position of the fault information source, and realizing the positioning of the fault section.
Advantageous effects
The invention provides a fault section positioning method and a fault section positioning device based on a fuzzy c-means clustering algorithm, and the method and the device have the following advantages:
(1) the existing method for positioning the ground fault section of the distribution network system usually takes single fault characteristic quantity as a fault criterion, the misjudgment rate is high, but the scheme adopts the idea of comprehensively analyzing the multi-source fault characteristic quantity, so that the reliability in the aspect of data is greatly improved, and the accuracy of the method for positioning the ground fault section of the scheme is higher than that of the similar method;
(2) according to the ground fault section positioning technology, the fault characteristics are amplified through a comprehensive judgment system of multiple fault characteristic quantities, the influence of accidental factors on fault judgment is weakened, the influence of interference factors such as nonlinear load and the like is avoided, and the anti-interference capability is strong;
(3) according to the scheme, a historical sample set is not required to be utilized, independent data processing is not required to be carried out on all feeder line related fault characteristic data, the samples to be tested in the whole system can be unified and standardized at one time, and the complexity of section positioning calculation is small.
Drawings
FIG. 1 is a flowchart of a method for locating a fault section based on a fuzzy c-means clustering algorithm according to an embodiment of the present invention;
FIG. 2 is a diagram of a ground fault information extraction provided by the present invention;
FIG. 3 is a schematic diagram of cluster analysis of a to-be-detected fault feature sample according to an embodiment of the present invention;
fig. 4 is a ground fault simulation model of a power distribution system according to an embodiment of the present invention.
Detailed Description
The present invention will be described in detail below with reference to the accompanying drawings and specific embodiments.
Example 1
As shown in fig. 1, the present embodiment provides a method for locating a fault section based on a fuzzy c-means clustering algorithm, including:
s01: and acquiring zero sequence voltage in the power distribution network system in real time.
S02: when the zero sequence voltage in the power distribution network system is larger than the phase voltage of 15%, the step S03 is carried out; otherwise, return to step S01; in the figure, U0Representing zero sequence voltage, UsetRepresenting 15% of the phase voltage.
S03: obtaining s fault characteristic quantities distributed on n nodes of the whole network, wherein the fault characteristic quantities of each node form a fault characteristic vector x'k(1×s)(k ═ 1, 2...., n), will fail feature vector x'k(1×s)Sample set X 'to be detected is constructed as sample to be detected'(n×s)(ii) a N is determined according to the condition of an actual power distribution network system, s is a preset value, and the s fault characteristic quantities comprise a plurality of steady-state characteristic quantities and a plurality of transient characteristic quantities; the steady state characteristic quantities at least comprise two of a zero sequence admittance angle xk1, a negative sequence current xk2, a zero sequence current xk3, amplitude variation of each phase voltage and phase angle variation of each phase voltage; the plurality of transient characteristic quantities at least comprise two of a zero-sequence current modulus maximum value xk4 after db3 wavelet transformation, a zero-sequence energy function value xk5 in a half power frequency cycle after the start of a fault and a zero-sequence active value xk6 in a power frequency cycle after the fault. The specific selection of the s fault characteristic quantities can be automatically selected and adjusted according to actual conditions. In this embodiment, the plurality of fault characteristic quantities include three steady-state characteristic quantities and two transient-state characteristic quantities; the three steady state characteristic quantities comprise a zero sequence admittance angle xk1, a negative sequence current xk2 and a zero sequence current xk 3; the two transient characteristic quantities comprise a zero-sequence current modulus maximum value xk4 after db3 wavelet transformation and a zero-sequence energy function value xk5 in a half power frequency cycle after the start of a fault. Extracting a plurality of fault characteristic quantities of the n nodes by n FTUs (distribution automation terminal units) distributed corresponding to the n node positions, each FTU being disposed at an outlet corresponding to a protected feeder, as shown in fig. 2An example graph of ground fault information extraction is provided.
S04: collecting X 'to-be-detected sample'(n×s)Carrying out standardized pretreatment on each sample to be detected; each sample to be tested or fault feature vector x'kCan be represented by x'k=(x′k1,…,x′kj,…,x′ks) Wherein, x'k1、…、x′ksThe normalization preprocessing comprises the following steps of:
carrying out standardization pretreatment on a sample to be detected according to the following formula:
wherein, x'kjThe jth fault characteristic quantity value of the kth sample to be tested,is the mean value of the jth fault characteristic quantity, S (x'j) Is the standard deviation, x, of the jth fault characteristic quantitykjAnd s is the total number of the fault characteristic quantities acquired by each node for the jth fault characteristic quantity value of the kth sample to be detected after the normalization pretreatment. The normalized kth sample to be tested can be expressed as: x is the number ofk=(xk1,…,xkj,…,xks) Forming a standardized section positioning sample matrix X by the standardized sample set to be tested(n×s)。
S05: constructing a high-dimensional fault feature space, projecting n samples to be detected into the high-dimensional fault feature space based on a plurality of fault feature quantities of each preprocessed sample to be detected, dividing the n preprocessed samples to be detected into a fault class and a non-fault class based on a fuzzy c-means clustering algorithm, and calculating to obtain clustering centers of the fault class and the non-fault class;
the clustering process specifically comprises the following steps: dynamically clustering all preprocessed samples to be detected by the aid of balance iterative equations (5) and (6) through the following optimization objective functions (4), and obtaining clustering centers of fault classes and non-fault classes:
wherein c is the number of clustering categories, and c is 2; mu.sik∈[0,1]Represents the k-th sample x to be measured after the pretreatmentkMembership degree belonging to the ith cluster typepiFor cluster centers, i takes 1 or 2, p1For fault class centers, p2Is a non-fault class center; the | l | · | is a matrix norm representing the spatial distance between the preprocessed sample to be detected and the clustering center; m is a weighting index, and m is 2; u is a membership matrix formed by membership degrees of all preprocessed samples to be detected, and P is all clustering centers PiFormed cluster center matrix, Jm(U, P) is a clustering loss function, MfcA space is divided for the fuzzy c of the preprocessed sample to be detected,Rsall membership matrices generated in the iterative process; the end conditions of the iterative process of optimizing the objective function (4) are as follows:w is the current iteration number, isThe iteration stop threshold is set first, and is taken as 1.0e-6 in the embodiment.
S06: respectively calculating the distance d between each preprocessed sample to be detected and the clustering centers of the fault class and the non-fault classk1And dk2(ii) a The method specifically comprises the following steps:
calculating the Euclidean distance between each preprocessed sample to be detected and the clustering centers of the fault class and the non-fault class according to a formula (7):
wherein d isk1Representing the distance between the preprocessed sample to be tested and the center of the fault class, dk2Representing the distance, x, between the preprocessed sample to be tested and the non-fault centerkjFor normalizing the jth fault characteristic quantity value p of the kth sample to be tested after pretreatmentijCluster center, p, representing the jth fault characteristic quantity of a fault class or of a non-fault class1jCluster center, p, of jth fault feature quantity representing fault class or2jRepresenting a clustering center of the jth fault characteristic quantity of the non-fault class or the jth fault characteristic quantity, wherein s is the total number of the fault characteristic quantities acquired by each node;
s07: according to dk1And dk2The size relation divides the nodes corresponding to each preprocessed sample to be tested into a fault class or a non-fault class; the method specifically comprises the following steps: if d isk1>dk2If so, the node corresponding to the preprocessed sample to be detected belongs to the non-fault class; if d isk1<dk2And if so, the node corresponding to the preprocessed sample to be detected belongs to the fault class. As shown in fig. 3, this is an example of cluster analysis of samples to be measured.
The fuzzy clustering algorithm can realize soft division of the multiple samples, obtain the uncertainty degree of each sample belonging to different classes during classification, and objectively reflect the fuzziness of the fault characteristic information of the power distribution system. Therefore, the scheme adopts the fuzzy clustering algorithm to carry out fuzzy clustering analysis on all samples to be detected, and the sample set to be detected is divided into two types of fault type and non-fault type in a fuzzy mode. Different from the traditional protection scheme based on single characteristic quantity, due to the fact that multiple fault characteristic judgment indexes are fused, the scheme realizes multi-angle description of the line operation state and can make judgment according to different fault characteristics shown by different fault types under various operation conditions of the power distribution network. By combining the factors, the scheme can excavate more detailed information which cannot be presented by the traditional method by means of the fuzzy recognition function and the information fusion performance of the scheme.
S08: and searching a fault characteristic information transmission path based on the fault category to which each node belongs, determining the position of the fault information source, and realizing the positioning of the fault section.
Example 2
The embodiment provides a fault section positioning device based on a fuzzy c-means clustering algorithm, which comprises:
a sample set acquisition unit to be tested: acquiring a plurality of fault characteristic quantities distributed on n nodes of the whole network, wherein the plurality of fault characteristic quantities of each node form a fault characteristic vector, and the fault characteristic vector is used as a sample to be detected to construct a sample set to be detected;
a pretreatment unit: carrying out standardized pretreatment on each sample to be detected in a sample set to be detected;
a clustering center obtaining unit: constructing a high-dimensional fault feature space, projecting n samples to be detected into the high-dimensional fault feature space based on a plurality of fault feature quantities of each preprocessed sample to be detected, dividing the n preprocessed samples to be detected into a fault class and a non-fault class based on a fuzzy c-means clustering algorithm, and calculating to obtain clustering centers of the fault class and the non-fault class;
a node category division unit: respectively calculating the distance d between each preprocessed sample to be detected and the clustering centers of the fault class and the non-fault classk1And dk2According to dk1And dk2The size relation divides the nodes corresponding to each preprocessed sample to be tested into a fault class or a non-fault class;
a fault location unit: and searching a fault characteristic information transmission path based on the fault category to which each node belongs, determining the position of the fault information source, and realizing the positioning of the fault section.
Please refer to the method for locating a fault section based on the fuzzy c-means clustering algorithm provided in embodiment 1 for other specific implementation schemes of this embodiment, which are not described herein again.
A specific simulation case is further described below.
PSCAD/EMTDC simulation software is adopted to carry out simulation analysis on a 35kV power distribution network, and a simulation model is shown in figure 4. The system adopts a mode that a neutral point is grounded through an arc suppression coil, the detuning degree is 5%, a 110kV system supplies power to a bus through a delta/Y transformer, and the influence of a nonlinear load on the system is considered. The four feeder lines connected with the bus comprise three overhead lines and a cable line, and specific parameters of the lines are shown in table 1.
TABLE 1
In order to realize wide area information acquisition of a power distribution system, a certain number of measuring nodes are arranged in each section of a network, and each measuring node is provided with a corresponding measuring device FTU to extract fault characteristic quantity of each section of line. And 3 kinds of steady-state fault characteristic quantities and 2 kinds of transient fault characteristic quantities with strong similarity in the propagation and distribution modes are selected to jointly depict the operation state of the system. Wherein the steady state characteristic quantities are respectively: zero sequence admittance angle xk1, negative sequence current xk2, and zero sequence current xk 3; the transient characteristic quantity is: db3 wavelet transformed zero sequence current modulus maximum value xk4, zero sequence energy function value xk5 in half power frequency cycle after fault start.
The method has the advantages that the fault position of the grounding point is changed under the condition of different detuning degrees of the power distribution network, comprehensive simulation calculation is completed according to various fault types (metallic grounding, high-resistance grounding, arc grounding and the like), and the positioning result of the fault section is 100% accurate. The simulation result is described as follows by a typical example: the single-phase earth fault is arranged between the detection points 17 and 18, and the earth fault resistor RfThe values of the characteristic quantities of the faults collected by the detection units are shown in table 2.
TABLE 2
After each sample to be tested in table 2 is subjected to standardization processing according to formulas (1) - (3), the sample set to be tested is divided into fault classes and non-fault classes in a fuzzy manner by means of a fuzzy clustering algorithm, and the fault class and non-fault class sample clustering centers are obtained by balance iteration, as shown in table 3.
TABLE 3
The distribution of the fault type samples and the non-fault type samples has obvious difference in spatial position: the spatial distance between the homogeneous samples is small, and the spatial distance between the heterogeneous samples is large. The clustering center intensively reflects the distribution condition of each sample to be detected in the high-dimensional fault characteristic space, and reflects the distribution rule of the class attribute of the sample to be detected.
All samples to be tested are divided into a fault type and a non-fault type by comparing Euclidean distances from the samples to be tested to the centers of the two types of types. As shown in table 4, the samples to be measured classified as non-failure classes are: { x '1, x'2, x '3, x'4, x '5, x'6, x '7, x'8, x '9, x'10, x '11, x'12, x '16, x'18 }; the samples to be tested that are classified as fault classes are: { x '13, x'14, x '15, x'17 }; the propagation path of the fault characteristic quantity is 13-14-15-17-18. According to the method for locating a fault section proposed herein, the location of the fault section as the section where the fault characteristic information source is located, i.e., sections 17-18, coincides with the actual fault situation. The PSCAD/EMTDC simulation result proves the effectiveness of the power distribution network fault section positioning scheme based on fuzzy clustering analysis.
TABLE 4
The invention provides a fault section positioning method and a fault section positioning device based on a fuzzy c-means clustering algorithm, and the method and the device have the following advantages:
(1) the existing method for positioning the ground fault section of the distribution network system usually takes single fault characteristic quantity as a fault criterion, the misjudgment rate is high, but the scheme adopts the idea of comprehensively analyzing the multi-source fault characteristic quantity, so that the reliability in the aspect of data is greatly improved, and the accuracy of the method for positioning the ground fault section of the scheme is higher than that of the similar method;
(2) according to the ground fault section positioning technology, the fault characteristics are amplified through a comprehensive judgment system of multiple fault characteristic quantities, the influence of accidental factors on fault judgment is weakened, the influence of interference factors such as nonlinear load is avoided, and the anti-interference capability is strong;
(3) the method has the characteristic of no influence of system operation mode change, better solves the problem of ground fault protection in the dynamic operation environment of the power distribution network, and can also be used for the section positioning problem of other various types of faults of the power system;
(4) according to the scheme, a historical sample set is not required to be utilized, independent data processing is not required to be carried out on all feeder line related fault characteristic data, the samples to be tested in the whole system can be unified and standardized at one time, and the complexity of section positioning calculation is small.
The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention, and various modifications and changes may be made by those skilled in the art. Any modification, equivalent replacement, or improvement made within the spirit and principle of the present invention should be included in the protection scope of the present invention.
Claims (9)
1. A fault section positioning method based on a fuzzy c-means clustering algorithm is characterized by comprising the following steps:
acquiring a plurality of fault characteristic quantities distributed on n nodes of the whole network, wherein the plurality of fault characteristic quantities of each node form a fault characteristic vector, and the fault characteristic vector is used as a sample to be detected to construct a sample set to be detected;
carrying out standardized pretreatment on each sample to be detected in a sample set to be detected;
constructing a high-dimensional fault feature space, projecting n samples to be detected into the high-dimensional fault feature space based on a plurality of fault feature quantities of each preprocessed sample to be detected, dividing the n preprocessed samples to be detected into a fault class and a non-fault class based on a fuzzy c-means clustering algorithm, and calculating to obtain clustering centers of the fault class and the non-fault class;
respectively calculating the distance d between each preprocessed sample to be detected and the clustering centers of the fault class and the non-fault classk1And dk2According to dk1And dk2The size relation divides the nodes corresponding to each preprocessed sample to be tested into a fault class or a non-fault class;
and searching a fault characteristic information transmission path based on the fault category to which each node belongs, determining the position of the fault information source, and realizing the positioning of the fault section.
2. The method for locating the fault section based on the fuzzy c-means clustering algorithm according to claim 1, wherein the step of performing the standardized preprocessing on each sample to be tested in the sample set to be tested comprises:
carrying out standardization pretreatment on a sample to be detected according to the following formula:
wherein, x'kjThe jth fault characteristic quantity value of the kth sample to be tested,is the mean value of the jth fault characteristic quantity, S (x'j) Is the standard deviation, x, of the jth fault characteristic quantitykjAnd s is the total number of the fault characteristic quantities acquired by each node for the jth fault characteristic quantity value of the kth sample to be detected after the normalization pretreatment.
3. The method for locating the fault section based on the fuzzy c-means clustering algorithm according to claim 1, wherein the fuzzy c-means clustering algorithm divides n preprocessed samples to be tested into a fault class and a non-fault class, and calculates the clustering centers of the fault class and the non-fault class, and comprises:
dynamically clustering all preprocessed samples to be detected by the aid of balance iterative equations (5) and (6) through the following optimization objective functions (4), and obtaining clustering centers of fault classes and non-fault classes:
wherein c is the number of clustering categories, and c is 2; mu.sik∈[0,1]Represents the k-th sample x to be measured after the pretreatmentkMembership degree belonging to the ith cluster typepiFor cluster centers, i takes 1 or 2, p1For fault class centers, p2Is a non-fault class center; the | l | · | is a matrix norm representing the spatial distance between the preprocessed sample to be detected and the clustering center; m is a weighting index, and m is 2; u is a membership matrix formed by the membership degrees of all the preprocessed samples to be detected,p is all cluster centers PiFormed cluster center matrix, Jm(U, P) is a clustering loss function, MfcDividing a space for the fuzzy c of the preprocessed sample to be detected; the end conditions of the iterative process of optimizing the objective function (4) are as follows:and w is the current iteration frequency and is a preset iteration stop threshold.
4. The method for locating the fault section based on the fuzzy c-means clustering algorithm as claimed in claim 3, wherein the distance d between each preprocessed sample to be tested and the clustering centers of the fault class and the non-fault class is calculated respectivelyk1And dk2According to dk1And dk2The size relationship of (2) divides the nodes corresponding to each preprocessed sample to be tested into a fault class or a non-fault class, and comprises the following steps:
calculating the Euclidean distance between each preprocessed sample to be detected and the clustering centers of the fault class and the non-fault class according to a formula (7):
wherein d isk1Representing the distance between the preprocessed sample to be tested and the center of the fault class, dk2Representing the distance, x, between the preprocessed sample to be tested and the non-fault centerkjFor normalizing the jth fault characteristic quantity value p of the kth sample to be tested after pretreatmentijRepresenting the clustering center of the jth fault characteristic quantity of the fault type or the non-fault type, wherein s is the total number of the fault characteristic quantities acquired by each node;
if d isk1>dk2If so, the node corresponding to the preprocessed sample to be detected belongs to the non-fault class;
if d isk1<dk2And if so, the node corresponding to the preprocessed sample to be detected belongs to the fault class.
5. The method for locating a fault section based on the fuzzy c-means clustering algorithm according to claim 1, wherein the plurality of fault characteristic quantities comprise a plurality of steady-state characteristic quantities and a plurality of transient characteristic quantities; the steady state characteristic quantities at least comprise two of a zero sequence admittance angle xk1, a negative sequence current xk2, a zero sequence current xk3, amplitude variation of each phase voltage and phase angle variation of each phase voltage; the plurality of transient characteristic quantities at least comprise two of a zero-sequence current modulus maximum value xk4 after db3 wavelet transformation, a zero-sequence energy function value xk5 in a half power frequency cycle after the start of a fault and a zero-sequence active value xk6 in a power frequency cycle after the fault.
6. The method according to claim 5, wherein the plurality of fault characteristic quantities include three steady-state characteristic quantities and two transient characteristic quantities; the three steady state characteristic quantities comprise a zero sequence admittance angle xk1, a negative sequence current xk2 and a zero sequence current xk 3; the two transient characteristic quantities comprise a zero-sequence current modulus maximum value xk4 after db3 wavelet transformation and a zero-sequence energy function value xk5 in a half power frequency cycle after the start of a fault.
7. The method for locating a fault section based on the fuzzy c-means clustering algorithm according to claim 1, further comprising, before the obtaining the plurality of fault feature quantities distributed in n nodes of the whole network:
acquiring zero sequence voltage in a power distribution network system in real time;
and when the zero sequence voltage in the power distribution network system is larger than 15% of the phase voltage, entering a single-phase earth fault section positioning process of the power distribution network.
8. The method according to claim 1, wherein the plurality of fault feature quantities of the n nodes are extracted by n FTUs distributed according to the n node positions.
9. A fault section positioning device based on a fuzzy c-means clustering algorithm is characterized by comprising the following components:
a sample set acquisition unit to be tested: acquiring a plurality of fault characteristic quantities distributed on n nodes of the whole network, wherein the plurality of fault characteristic quantities of each node form a fault characteristic vector, and the fault characteristic vector is used as a sample to be detected to construct a sample set to be detected;
a pretreatment unit: carrying out standardized pretreatment on each sample to be detected in a sample set to be detected;
a clustering center obtaining unit: constructing a high-dimensional fault feature space, projecting n samples to be detected into the high-dimensional fault feature space based on a plurality of fault feature quantities of each preprocessed sample to be detected, dividing the n preprocessed samples to be detected into a fault class and a non-fault class based on a fuzzy c-means clustering algorithm, and calculating to obtain clustering centers of the fault class and the non-fault class;
a node category division unit: respectively calculating the distance d between each preprocessed sample to be detected and the clustering centers of the fault class and the non-fault classk1And dk2According to dk1And dk2The size relation divides the nodes corresponding to each preprocessed sample to be tested into a fault class or a non-fault class;
a fault location unit: and searching a fault characteristic information transmission path based on the fault category to which each node belongs, determining the position of the fault information source, and realizing the positioning of the fault section.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202010407081.2A CN111596167A (en) | 2020-05-14 | 2020-05-14 | Fault section positioning method and device based on fuzzy c-means clustering algorithm |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202010407081.2A CN111596167A (en) | 2020-05-14 | 2020-05-14 | Fault section positioning method and device based on fuzzy c-means clustering algorithm |
Publications (1)
Publication Number | Publication Date |
---|---|
CN111596167A true CN111596167A (en) | 2020-08-28 |
Family
ID=72180628
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN202010407081.2A Pending CN111596167A (en) | 2020-05-14 | 2020-05-14 | Fault section positioning method and device based on fuzzy c-means clustering algorithm |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN111596167A (en) |
Cited By (11)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN112964959A (en) * | 2021-02-03 | 2021-06-15 | 国网河南省电力公司电力科学研究院 | Power distribution network earth fault line selection method based on correlation coefficient measurement criterion |
CN113156267A (en) * | 2021-04-27 | 2021-07-23 | 福州大学 | Power distribution network ground fault section selection method and system |
CN113419140A (en) * | 2021-08-06 | 2021-09-21 | 广东电网有限责任公司 | Distribution network line fault diagnosis method and device considering human-computer-object cooperation |
CN113434317A (en) * | 2021-03-11 | 2021-09-24 | 同济大学 | Fault diagnosis method and system based on unbalanced data |
CN113608066A (en) * | 2021-07-16 | 2021-11-05 | 国网江西省电力有限公司电力科学研究院 | Distribution network overhead line single-phase earth fault positioning method based on magnetic field detection |
CN113674106A (en) * | 2021-08-11 | 2021-11-19 | 国网山东省电力公司平阴县供电公司 | Combined positioning method for ground fault of medium-low voltage distribution network |
CN113792824A (en) * | 2021-11-17 | 2021-12-14 | 广东电网有限责任公司佛山供电局 | Power distribution network fault power failure influence area classification method and system |
CN114488785A (en) * | 2021-08-09 | 2022-05-13 | 宁波大学 | MFC actuator trajectory tracking method |
CN114660418A (en) * | 2022-05-24 | 2022-06-24 | 国网山西省电力公司大同供电公司 | Power distribution network single-phase earth fault line selection method based on feature fusion and clustering |
CN114705953A (en) * | 2022-05-25 | 2022-07-05 | 国网山西省电力公司大同供电公司 | WPT-HHT-based small current ground fault section positioning method |
CN116087692A (en) * | 2023-04-12 | 2023-05-09 | 国网四川省电力公司电力科学研究院 | Distribution network tree line discharge fault identification method, system, terminal and medium |
Citations (5)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN106370981A (en) * | 2016-09-30 | 2017-02-01 | 中国石油化工股份有限公司 | Fuzzy clustering analysis-based power distribution network fault line selection method |
CN107589341A (en) * | 2017-08-16 | 2018-01-16 | 珠海许继芝电网自动化有限公司 | A kind of online Fault Locating Method of single-phase earthing based on power distribution automation main station |
CN107607836A (en) * | 2017-09-22 | 2018-01-19 | 中国矿业大学 | A kind of electric network fault area positioning method based on multidimensional information |
CN108459240A (en) * | 2018-03-27 | 2018-08-28 | 国网江苏省电力有限公司苏州供电分公司 | A kind of distribution network line fault deciding field method |
CN110609200A (en) * | 2019-09-20 | 2019-12-24 | 长沙理工大学 | Power distribution network earth fault protection method based on fuzzy metric fusion criterion |
-
2020
- 2020-05-14 CN CN202010407081.2A patent/CN111596167A/en active Pending
Patent Citations (5)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN106370981A (en) * | 2016-09-30 | 2017-02-01 | 中国石油化工股份有限公司 | Fuzzy clustering analysis-based power distribution network fault line selection method |
CN107589341A (en) * | 2017-08-16 | 2018-01-16 | 珠海许继芝电网自动化有限公司 | A kind of online Fault Locating Method of single-phase earthing based on power distribution automation main station |
CN107607836A (en) * | 2017-09-22 | 2018-01-19 | 中国矿业大学 | A kind of electric network fault area positioning method based on multidimensional information |
CN108459240A (en) * | 2018-03-27 | 2018-08-28 | 国网江苏省电力有限公司苏州供电分公司 | A kind of distribution network line fault deciding field method |
CN110609200A (en) * | 2019-09-20 | 2019-12-24 | 长沙理工大学 | Power distribution network earth fault protection method based on fuzzy metric fusion criterion |
Non-Patent Citations (2)
Title |
---|
曾祥君等: "基于聚类分析的配电网无整定保护技术", 《电力科学与技术学报》 * |
王小匆: "基于PMU数据的配电网故障定位方法研究", 《中国优秀硕士学位论文全文数据库(电子期刊)》 * |
Cited By (16)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN112964959A (en) * | 2021-02-03 | 2021-06-15 | 国网河南省电力公司电力科学研究院 | Power distribution network earth fault line selection method based on correlation coefficient measurement criterion |
CN113434317A (en) * | 2021-03-11 | 2021-09-24 | 同济大学 | Fault diagnosis method and system based on unbalanced data |
CN113156267A (en) * | 2021-04-27 | 2021-07-23 | 福州大学 | Power distribution network ground fault section selection method and system |
CN113156267B (en) * | 2021-04-27 | 2024-02-02 | 福州大学 | Power distribution network ground fault section selection method and system |
CN113608066B (en) * | 2021-07-16 | 2023-11-07 | 国网江西省电力有限公司电力科学研究院 | Distribution network overhead line single-phase grounding fault positioning method based on magnetic field detection |
CN113608066A (en) * | 2021-07-16 | 2021-11-05 | 国网江西省电力有限公司电力科学研究院 | Distribution network overhead line single-phase earth fault positioning method based on magnetic field detection |
CN113419140A (en) * | 2021-08-06 | 2021-09-21 | 广东电网有限责任公司 | Distribution network line fault diagnosis method and device considering human-computer-object cooperation |
CN113419140B (en) * | 2021-08-06 | 2023-01-24 | 广东电网有限责任公司 | Distribution network line fault diagnosis method and device considering human-computer-object cooperation |
CN114488785A (en) * | 2021-08-09 | 2022-05-13 | 宁波大学 | MFC actuator trajectory tracking method |
CN114488785B (en) * | 2021-08-09 | 2023-11-14 | 宁波大学 | MFC actuator track tracking method |
CN113674106A (en) * | 2021-08-11 | 2021-11-19 | 国网山东省电力公司平阴县供电公司 | Combined positioning method for ground fault of medium-low voltage distribution network |
CN113792824A (en) * | 2021-11-17 | 2021-12-14 | 广东电网有限责任公司佛山供电局 | Power distribution network fault power failure influence area classification method and system |
CN114660418A (en) * | 2022-05-24 | 2022-06-24 | 国网山西省电力公司大同供电公司 | Power distribution network single-phase earth fault line selection method based on feature fusion and clustering |
CN114705953A (en) * | 2022-05-25 | 2022-07-05 | 国网山西省电力公司大同供电公司 | WPT-HHT-based small current ground fault section positioning method |
CN116087692B (en) * | 2023-04-12 | 2023-06-23 | 国网四川省电力公司电力科学研究院 | Distribution network tree line discharge fault identification method, system, terminal and medium |
CN116087692A (en) * | 2023-04-12 | 2023-05-09 | 国网四川省电力公司电力科学研究院 | Distribution network tree line discharge fault identification method, system, terminal and medium |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN111596167A (en) | Fault section positioning method and device based on fuzzy c-means clustering algorithm | |
CN110082640B (en) | Distribution network single-phase earth fault identification method based on long-time memory network | |
CN110609200B (en) | Power distribution network earth fault protection method based on fuzzy metric fusion criterion | |
CN109459669B (en) | 10kV power distribution network single-phase earth fault section positioning method | |
CN112147462A (en) | Power transmission line fault identification method based on deep learning | |
CN111598166A (en) | Single-phase earth fault classification method and system based on principal component analysis and Softmax function | |
CN110927527A (en) | Transient fault sensing method based on transient energy double channels and information fusion | |
CN115267428B (en) | LCC-MMC monopole ground fault positioning method based on VMD-ET feature selection | |
CN111612056A (en) | Low-pressure customer variation relation identification method based on fuzzy clustering and zero-crossing offset | |
CN112964959A (en) | Power distribution network earth fault line selection method based on correlation coefficient measurement criterion | |
CN112684295A (en) | Power distribution network fault line selection method and system under high permeability based on similarity separation degree | |
CN115618249A (en) | Low-voltage power distribution station area phase identification method based on LargeVis dimension reduction and DBSCAN clustering | |
CN116845971A (en) | Automatic identification method for topological structure of photovoltaic grid-connected low-voltage transformer area | |
CN110492470B (en) | Power distribution network multi-dimensional typical scene generation method based on load clustering and network equivalence | |
CN113659564B (en) | Low-voltage distribution network topology identification method and system based on voltage fluctuation feature clustering | |
CN116990633A (en) | Fault studying and judging method based on multiple characteristic quantities | |
CN109684749B (en) | Photovoltaic power station equivalent modeling method considering operating characteristics | |
CN110232168B (en) | Electrical distance-based electrical coordinate system construction method, medium and equipment | |
CN116466185A (en) | Intelligent threshold-free positioning method for single-phase ground fault section of power distribution network | |
CN115545422A (en) | Platform area user variation relation identification method based on improved decision mechanism | |
Dehghani et al. | Distribution feeder classification based on self organized maps (case study: Lorestan province, Iran) | |
Li et al. | Faulty Feeders Identification for Single-phase-to-ground Fault Based on Multi-features and Machine Learning | |
Wei et al. | Fault location method for active distribution network based on SVM and feature search algorithm | |
CN113687292A (en) | System and method for detecting electric energy meter wrong wiring in big data and cloud environment | |
Deng et al. | Topology Estimation of Power Distribution Grids Using Graphical Model |
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 | ||
RJ01 | Rejection of invention patent application after publication | ||
RJ01 | Rejection of invention patent application after publication |
Application publication date: 20200828 |