CN108267673B - Distribution network fault line selection big data dimension reduction method and device - Google Patents
Distribution network fault line selection big data dimension reduction method and device Download PDFInfo
- Publication number
- CN108267673B CN108267673B CN201810063390.5A CN201810063390A CN108267673B CN 108267673 B CN108267673 B CN 108267673B CN 201810063390 A CN201810063390 A CN 201810063390A CN 108267673 B CN108267673 B CN 108267673B
- Authority
- CN
- China
- Prior art keywords
- fault
- line
- data
- distribution network
- mic
- 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
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
Landscapes
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Engineering & Computer Science (AREA)
- Mathematical Physics (AREA)
- Theoretical Computer Science (AREA)
- Supply And Distribution Of Alternating Current (AREA)
Abstract
The invention discloses a distribution network fault line selection big data dimension reduction method and a distribution network fault line selection big data dimension reduction device, wherein the distribution network fault line selection big data is obtained firstly, and fault characteristic variables are extracted; then analyzing the correlation degree between the extracted fault characteristic variables and the line fault result, and calculating the MIC value between each fault characteristic variable and the line fault result; and comparing the MIC value between each fault characteristic variable and the line fault result with the set MIC threshold value, and screening the fault characteristic variables. The invention reduces the big data dimension when the big data method is used for carrying out distribution network fault line selection, and greatly reduces the big data processing difficulty.
Description
Technical Field
The invention relates to the technical field of power grid fault line selection, in particular to a distribution network fault line selection big data dimension reduction method and device based on fault feature extraction and a maximum information coefficient theory.
Background
The power grid is a support of national economy, the development of economy cannot be separated from the development of the power grid, and once a large-area power failure accident occurs, immeasurable loss can be generated. The power distribution network is a network directly transmitting electric energy to users, the normal operation of the power distribution network has great significance to social development, and the line selection of the grounding fault of the power distribution network is always a difficult problem.
At present, the neutral point of a medium-voltage distribution network at home and abroad widely adopts a low-current grounding mode to avoid power supply interruption caused by tripping when single-phase grounding faults occur. For a small-current ground fault, due to the reasons of weak fault current, unstable electric arc, influence of random factors and the like, the line selection of the ground fault is difficult. In the traditional various line selection methods, only one or more fault characteristic variables are considered, the accuracy of the obtained line selection result is low, and the requirement of fault line selection of the power distribution network cannot be met.
With the development of smart grids, a large number of smart monitoring devices are being applied to distribution grids, thereby generating a large amount of data of different types. The data processing method comprises various electrical quantity data generated by a real-time monitoring device in the power distribution network and various data outside the power distribution network, such as weather, traffic, geographic information and other non-electrical quantity data.
In order to comprehensively analyze a fault line by using various data and improve the fault line selection accuracy of the low-current grounding system, a large amount of data needs to be analyzed and processed by using a big data technology. A big data fault line selection method is adopted at present, a neural network model needs to be established, a large amount of data is used as input, a fault line is used as output, data mining and model training are carried out. Due to the fact that various fault characteristic information is utilized, the method can effectively improve the fault line selection accuracy of the low-current grounding system, and is of great significance in improving the distribution network automation level.
In summary, in the prior art, due to the fact that data dimensionality is too high and noise often exists, the problems of insufficient result precision and overlong data mining time are caused when data mining is performed, and an effective solution is not provided.
Disclosure of Invention
In order to overcome the defects of the prior art, the invention provides a large data dimension reduction method and device for distribution network fault line selection, which reduces the large data dimension when the large data method is used for distribution network fault line selection and greatly reduces the large data processing difficulty.
The technical scheme adopted by the invention is as follows:
a distribution network fault line selection big data dimension reduction method comprises the following steps:
the method comprises the following steps: acquiring large data of distribution network fault line selection, and extracting fault characteristic variables;
step two: analyzing the correlation degree between the fault characteristic variables extracted in the step one and the line fault result, and calculating an MIC (minimum integrated coefficient) value between each fault characteristic variable and the line fault result;
step three: and comparing the MIC value between each fault characteristic variable and the line fault result with the set MIC threshold value, and screening the fault characteristic variables.
Further, the distribution network fault line selection big data comprise current and voltage data, zero sequence voltage and current data, steady-state electrical data, transient electrical data, remote telemetry data and fault recording data.
Further, in the step one, a specific method for acquiring the distribution network fault line selection big data is as follows:
building a simulation model, simulating faults of different lines, different positions and different types, and generating a plurality of groups of fault recording data;
recording fault recording data of each line through a fault recorder;
and acquiring fault recording data of each line as distribution network fault line selection big data.
Further, the fault characteristic variables comprise fault types, phase voltage drop values of all phases after fault, zero sequence current amplitude and phase angle after fault, wavelet packet energy of all outgoing line zero sequence currents, zero sequence current quintic harmonic amplitude and direction, first half-wave amplitude and polarity after fault, energy function after fault, transient state zero mode characteristic current SFB amplitude and polarity and zero sequence current amplitude and phase angle at FTU installation position.
Further, in the step one, a specific method for extracting the fault characteristic variable is as follows:
selecting a fault line, and representing different types of fault results by using different numbers;
calculating voltage drop values of each phase before and after the fault by using voltage effective values before and after the fault in the distribution network fault line selection big data;
extracting the amplitude and phase angle of the zero-sequence current after the fault, the amplitude and polarity of the transient zero-mode characteristic current SFB and the amplitude and phase angle of the zero-sequence current at the installation position of the FTU from the large data of the distribution network fault line selection;
according to the big data of the fault line selection of the distribution network, calculating the wavelet packet energy of the zero-sequence current of each outgoing line by using a dB15 basis function, a three-layer decomposition function and a fourth scale energy function;
and calculating an energy function after the fault according to the distribution network fault line selection big data, and extracting the quintic harmonic component of the zero-sequence current and the amplitude and the polarity of the first half wave after the fault.
Further, in the second step, the specific method of analyzing the degree of association between the fault characteristic variables extracted in the first step and the line fault result and calculating the MIC value between each fault characteristic variable and the line fault result includes:
constructing a two-dimensional data set based on each fault characteristic variable and a line fault result;
and respectively calculating MIC values of all two-dimensional data sets, namely MIC values between all fault characteristic variables and line fault results according to the maximum information coefficient MIC calculation formula.
Further, the maximum information coefficient MIC is calculated by the following formula:
MIC(D)＝maxxy<B(n)M(D)xy＝maxxy<B(n)I(D,x,y)·log-1(min(x,y))
wherein D is a two-dimensional dataset, I (D, x, y) ═ max (I (D-G)), G represents a partition of the two-dimensional dataset D, D-G represents a distribution of the two-dimensional dataset D in the grid G, and I (D, x, y) represents mutual information based on the two-dimensional dataset D; b (n) the power of 0.6 or 0.55 of the total amount of data taken is an empirical value; x is a fault characteristic variable; and y is a line fault result.
Further, in the third step, the MIC value between each fault characteristic variable and the line fault result is compared with the set MIC threshold value, and the specific method for screening the fault characteristic variables includes:
comparing the MIC value between each fault characteristic variable and the line fault result with the set MIC threshold value;
if the MIC value between the fault characteristic variable and the line fault result is greater than or equal to the set MIC threshold value, indicating that the correlation degree between the fault characteristic variable and the line fault result is large, and keeping the fault characteristic variable;
and if the MIC value between the fault characteristic variable and the line fault result is smaller than the set MIC threshold value, indicating that the correlation degree between the fault characteristic variable and the line fault result is small, and rejecting the fault characteristic variable.
A device for realizing the distribution network fault line selection big data dimension reduction method comprises the following steps:
the fault recording data simulation module is used for simulating and generating fault recording data of each line;
the distribution network fault line selection big data acquisition module is used for acquiring fault recording data of each line from the fault recording data simulation module and taking the fault recording data of each line as distribution network fault line selection big data;
the fault characteristic variable extraction module is used for extracting fault characteristic variables of the large data of the distribution network fault line selection;
the maximum information coefficient analysis module is used for analyzing the correlation degree between the extracted fault characteristic variables and the line fault result and calculating an MIC value between the fault characteristic variables and the line fault result;
and the fault characteristic variable screening module is used for reserving the fault characteristic variable with large association degree and eliminating the fault characteristic variable with small association degree according to the MIC value between the fault characteristic variable and the line fault result.
Furthermore, the fault recording data simulation module comprises a low-current grounding system with a plurality of outgoing lines, wherein each outgoing line of the low-current grounding system is provided with a feeder line terminal device, and parameters are set through the feeder line terminal devices; each outgoing line of the low-current grounding system is provided with a plurality of fault positions, a fault recorder is installed on each fault position, and fault recording data of each line during fault are recorded through the fault recorder.
Compared with the prior art, the invention has the beneficial effects that:
(1) the method extracts fault characteristic data which is significant to fault line selection from the large data of the fault line selection of the distribution network, performs primary dimension reduction on the large data of the fault line selection of the distribution network, performs maximum information coefficient algorithm analysis on the correlation degree between fault characteristic variables and whether a line has a fault, obtains MIC values of various fault characteristic variables and whether the line has the fault, retains the fault characteristics with larger correlation degree, eliminates the fault characteristics with smaller correlation degree, performs further dimension reduction on the large data, effectively reduces the dimension of the large data when the large data method is used for fault line selection of the distribution network, and greatly reduces the difficulty in processing the large data;
(2) according to the method, the useful fault characteristic variables are extracted based on the fault line selection method, and the preliminary dimension reduction of the big data of the distribution network fault line selection is effectively carried out;
(3) the invention adopts an advanced correlation maximum information coefficient analysis algorithm, accurately and comprehensively researches the correlation degree of various fault characteristic variables and line faults, and effectively carries out further dimension reduction on big data based on the correlation degree;
(4) the method and the device screen the original big data of the fault line selection of the distribution network, retain useful data and remove irrelevant data, thereby improving the fault line selection precision and the operation speed of the big data.
Drawings
The accompanying drawings, which are incorporated in and constitute a part of this application, illustrate embodiments of the application and, together with the description, serve to explain the application and are not intended to limit the application.
FIG. 1 is a flow chart of a distribution network fault line selection big data dimension reduction method disclosed in the embodiment of the present invention;
FIG. 2 is a simplified diagram of a simulation model disclosed in an embodiment of the present invention;
FIG. 3 is a flowchart of a line selection method based on big data of line selection of a distribution network fault, which is disclosed by the embodiment of the invention;
Detailed Description
It should be noted that the following detailed description is exemplary and is intended to provide further explanation of the disclosure. Unless defined otherwise, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this application belongs.
It is noted that the terminology used herein is for the purpose of describing particular embodiments only and is not intended to be limiting of example embodiments according to the present application. As used herein, the singular forms "a", "an" and "the" are intended to include the plural forms as well, and it should be understood that when the terms "comprises" and/or "comprising" are used in this specification, they specify the presence of stated features, steps, operations, devices, components, and/or combinations thereof, unless the context clearly indicates otherwise.
As introduced by the background technology, in the prior art, due to the fact that data dimensionality is too high and noise often exists, the problems of insufficient result precision and overlong data mining time are caused when data mining is carried out.
As shown in fig. 1, the present embodiment provides a distribution network fault line selection big data dimension reduction method, which includes the following steps:
step 101: and acquiring the large data of the fault line selection of the distribution network, and extracting fault characteristics of the large data of the fault line selection of the distribution network to obtain fault characteristic variables. Extracting fault characteristic data which has significance on fault line selection according to experience of a line selection method, and performing preliminary large data dimension reduction on the fault line selection of the distribution network;
step 102: maximum Information Coefficient (MIC) algorithm analysis is performed on the correlation degree between the fault feature variable extracted in step 101 and the line fault result, MIC values between various fault feature data and line fault are obtained, fault features with large correlation degree are reserved, fault features with small correlation degree are removed, and further large data dimension reduction is performed.
The distribution network fault line selection big data comprise various electric quantity data and non-electric quantity data generated by a real-time monitoring device in the distribution network, and most of the data are obtained from an internal system of the power system, such as current and voltage data, zero sequence voltage and current data, steady-state electric data, transient electric data, remote-end remote-measuring data and fault recording data. For these data, in order to secure the power system information, a forward isolation device needs to be installed.
In step 101, the fault feature extraction is to extract feature variables with important physical significance or statistical significance from the original data, and can perform effective dimension reduction on the original data with higher dimension. Since the original data, especially the electrical data, is too high in dimensionality, too large in data size, and contains some useless information, it is necessary to perform feature extraction to extract useful fault feature variables for analysis.
There are many methods for extracting fault features, and in order to ensure the validity of extracted data, the feature data to be extracted is generally determined according to a line selection method. In the fault line selection, different numbers are used for representing different types of fault results, voltage drops of each phase are calculated by using voltage effective values before and after the fault, wavelet packet energy of zero-sequence current of each outgoing line is calculated by using a dB15 basis function, three-layer decomposition and a fourth scale energy function, the amplitude and phase angle of each zero-sequence current are recorded, the energy function is calculated, and a quintic harmonic component of the zero-sequence current and the amplitude and polarity of a first half wave after the fault are extracted; according to fault recording data, extracting amplitude and polarity of transient zero-mode characteristic current SFB (setting sampling frequency of a neutral point ungrounded system to be 0-2 kHz and sampling frequency of an arc suppression coil grounded system to be 200 Hz-2000 kHz), and extracting zero-sequence current amplitude and phase angle of an FTU installation position. The above variables are used as fault characteristic variables, so that the dimension can be effectively reduced.
In the second step, the maximum information coefficient MIC is an effective measure for analyzing the degree of correlation among a large amount of data, and is an algorithm provided based on a mutual information technology, and the reliability is extremely high. The MIC can effectively dig out the strength of the relationship between two data items, which provides a basis for further data mining.
MIC has commonality and fairness. Wherein, the universality means that the method is not only suitable for functional correlation but also suitable for non-functional correlation, and is suitable for linear correlation and non-linear correlation; fairness refers to the fact that when two different functional relationships are interfered by the same noise, MIC values of the two different functional relationships are consistent, namely, the influence of the noise on the MIC is irrelevant to the functional relationship between variables.
When the number of samples is large enough, the fairness and universality of the MIC can ensure that it can capture a variety of complex hidden correlations, without being limited to a specific functional form, known as 21 st century correlation. And the MIC is used for analyzing the correlation degree between the accurate and comprehensive fault characteristic data and the line fault, and further useless data are eliminated for dimensionality reduction, so that the line selection precision and the operation speed of the big data model can be effectively improved.
The Maximum Information Coefficient (MIC) principle is as follows:
for a certain data variable, the amount of information contained in the random variable is measured and defined as follows:
wherein p isiRepresents the probability of occurrence of the event i,
the mutual information between two random variables is:
I(X,Y)＝H(X)+H(Y)-H(X,Y)
for a finite ordered two-dimensional data set D (x, y), the Maximum Information Coefficient (MIC) is calculated as follows:
MIC(D)＝maxxy<B(n)M(D)xy＝maxxy<B(n)I(D,x,y)·log-1(min(x,y))
where I (D, x, y) ═ max (I (D-G)), G represents a partition of the finite data set D, D-G represents the distribution of the data set D in the grid G, and I (D, x, y) represents mutual information based on the data set D. B (n) the power of 0.6 or 0.55 of the total amount of data taken is an empirical value; x, y are two vectors and D is the data set.
And importing a fault characteristic extraction result into the MATLAB, wherein the fault characteristic extraction result comprises a phase voltage drop value of each phase, the amplitude and the phase angle of zero-sequence current after fault, the wavelet packet energy of zero-sequence current of each outgoing line, the amplitude and the direction of quintic harmonic of the zero-sequence current, the amplitude and the polarity of the first half-wave after fault, an energy function after fault, the amplitude and the polarity of the SFB of transient zero-mode characteristic current and the amplitude and the phase angle of zero-sequence current at an FTU installation position. Taking each fault characteristic variable as a row vector, and taking a line fault result as another row vector to form a two-dimensional data set;
and (3) calculating MIC values between all fault characteristic variables and line fault results one by using a principle algorithm of a Maximum Information Coefficient (MIC) principle to obtain MIC values of every two vectors.
And comparing the MIC value between each fault characteristic variable and the line fault result with the set MIC threshold value. According to the MIC principle, the MIC value between two variables is between 0 and 1, the closer to 1, the greater the correlation degree, and the closer to 0, the smaller the correlation degree. Since the method involves many variables, combining with practical situations, it is an empirical value to use MIC of 0.1 as a threshold value for dividing the correlation degree.
If the MIC value between the fault characteristic variable and the line fault result is greater than or equal to the set MIC threshold value, indicating that the correlation degree between the fault characteristic variable and the line fault result is large, and keeping the fault characteristic variable; if the MIC value between the fault characteristic variable and the line fault result is smaller than the set MIC threshold value, indicating that the correlation degree between the fault characteristic variable and the line fault result is smaller, and rejecting the fault characteristic variable; and further large data dimension reduction is realized.
Another exemplary embodiment of the present application provides a device for implementing the above distribution network fault line selection big data dimension reduction method, where the device includes:
the fault recording data simulation module is used for simulating and generating fault recording data of each line;
the distribution network fault line selection big data acquisition module is used for acquiring fault recording data of each line from the fault recording data simulation module and taking the fault recording data of each line as distribution network fault line selection big data;
the fault characteristic variable extraction module is used for extracting fault characteristic variables of the large data of the distribution network fault line selection;
the maximum information coefficient analysis module is used for carrying out maximum information coefficient analysis on the extracted fault characteristic variables and the line fault result to obtain an MIC value between the fault characteristic variables and the line fault result;
and the fault characteristic variable screening module is used for reserving the fault characteristic variable with large association degree and eliminating the fault characteristic variable with small association degree according to the MIC value between the fault characteristic variable and the line fault result.
In this embodiment, the fault recording data simulation module includes a low-current grounding system having a plurality of outgoing lines, each outgoing line of the low-current grounding system is provided with a feeder terminal device, and parameters are set through the feeder terminal devices; each outgoing line of the low-current grounding system is provided with a plurality of fault positions, a fault recorder is installed on each fault position, and fault recording data of each line during fault are recorded through the fault recorder.
The technical solution of the present invention will be described in detail with reference to a specific embodiment.
Example one
The distribution network fault line selection big data dimension reduction method provided by the embodiment comprises the following steps:
acquiring large data of distribution network fault line selection, and extracting fault characteristics of the large data of the distribution network fault line selection. Extracting fault characteristic data which has significance on fault line selection by adopting a reasonable method according to experience of a line selection method, and performing preliminary big data dimension reduction;
and step two, carrying out Maximum Information Coefficient (MIC) algorithm analysis on the correlation degree between the fault characteristic data variable extracted in the step one and the fault result of the line, obtaining the MIC value between various fault characteristic data and the fault result of the line, retaining the fault characteristic with larger correlation degree, eliminating the fault characteristic with smaller correlation degree, and carrying out further big data dimension reduction.
The first step is that the distribution network fault line selection big data comprise various electric quantity data and non-electric quantity data generated by a real-time monitoring device in the distribution network, wherein the electric quantity data and the non-electric quantity data comprise current and voltage data, zero sequence voltage and current data, steady state electric data, transient state electric data, remote-end remote-measuring data and fault wave-recording data. Most data are obtained from the internal system of the power system, and for the data, a forward isolation device needs to be installed in order to ensure the information security of the power system.
And simulating the electrical data in the distribution network fault line selection big data by adopting a simulation model. A simulation model is built by using Power Systems Computer Aided Design (PSCAD) software of a Power system, the simulation model is a low-current grounding system with five outgoing lines, and each outgoing line is provided with a Feeder Terminal Unit (FTU). And setting parameters of a line, a load, a power supply and a transformer. The parameters of the arc suppression coil are 6.0H, and the parameters of the positive sequence of the circuit are as follows: positive sequence resistance R11.7E-4 Ω/m, positive sequence inductance XL13.8E-4 Ω/m, positive sequence capacitance XC132.84M Ω/M; the zero sequence parameters of the line are as follows: zero sequence resistance R02.3E-4 omega/m, zero sequence inductance XL01.72E-3 omega/m, zero sequence capacitance XC053.08M Ω/M. Each outgoing line is provided with five fault positions, a fault recorder is arranged, and voltage and current fault recording data of each line during fault are recorded. A simplified diagram of the model is shown in fig. 2.
Two grounding systems of which the neutral point is not grounded and is grounded through an arc suppression coil are arranged; setting the fault types as three conditions of A-phase fault grounding AG, B-phase fault grounding BG and C-phase fault grounding CG respectively; setting fault angles to be 0 degree, 45 degrees and 90 degrees respectively; setting fault resistances to be 0 omega, 10 omega and 50 omega respectively; the model is set to run circularly, and 1350 sets of fault information are generated in total.
As shown in fig. 3, the simulated electrical data is subjected to fault feature extraction. 225 groups of fault data are generated in each line during simulation, fault recording data are processed, and a reasonable method is adopted to extract useful characteristic data. The following 14 fault feature data are extracted: the fault type, each phase voltage drop after the fault, zero sequence current amplitude and phase angle after the fault, wavelet packet energy of zero sequence current of each outgoing line, zero sequence current quintuple harmonic amplitude and direction, first half-wave amplitude and polarity after the fault, energy function after the fault, transient zero mode characteristic current (SFB) amplitude and polarity, and zero sequence current amplitude and phase angle at the FTU installation position. And acquiring a fault characteristic variable according to each outgoing line, and normalizing. The important fault characteristics are extracted from the original electrical data, and data dimension reduction is effectively carried out.
In the second step, MIC is used for analyzing the correlation degree between the fault characteristic data variable and the fault result of the line accurately and comprehensively, useless data are further removed for dimension reduction, and the line selection precision and the operation speed of the big data model can be effectively improved.
And researching the correlation degree of the fault characteristic data variable and the line fault by using an MIC algorithm. Using the fault characteristic data as one vector and the line fault result as another vector, calculating MIC values of the two vectors by using Matrix Laboratory (MATLAB) software, and averaging 5 lines, the results are as follows:
TABLE 1 MIC values for fault characteristic variables and line faults
From the above results, in the two grounding systems, the zero-sequence current amplitude and phase angle, the wavelet packet energy, the first half-wave amplitude, the transient zero-mode characteristic current SFB amplitude and polarity, and the FTU amplitude are fault characteristic variables having a large correlation with the line fault, and the result is consistent with the advantages and disadvantages of the conventional line selection method. In an ungrounded system, the wavelet packet energy, the amplitude of the first half wave and the amplitude and polarity of the transient zero-mode characteristic current SFB are the most important variables; in a crowbar grounding system, the first half wave amplitude and the FTU polarity are more important variables. In two grounding systems, fault types and first half-wave polarities are fault characteristic variables with small correlation with line faults, and can be removed during data processing, and further dimension reduction is carried out, so that only 12 fault characteristic variables are left, namely each phase voltage is reduced after a fault, the amplitude and the phase angle of zero-sequence current after the fault, the wavelet packet energy of zero-sequence current of each outgoing line, the amplitude and the direction of quintic harmonic of the zero-sequence current, the amplitude and the direction of first half-wave after the fault, the energy function after the fault, the amplitude and the polarity of transient zero-mode characteristic current SFB, the amplitude and the phase angle of zero-sequence current at an FTU installation position, and the large data dimension is further reduced.
And establishing a big data model to verify the effectiveness of the dimension reduction method. A Statistical Product and Service Solutions Modeler (SPSS Modeler) of a data mining platform is used to build a big data routing model. And taking various fault characteristic vectors as input data and fault lines as output data, establishing a neural network, setting parameters of the neural network, and training the neural network.
In order to verify that the method provided by the invention can effectively reduce the dimension of data, improve the fault line selection accuracy and the operation speed, and add a plurality of irrelevant fault characteristic variables to add interference: amplitude and polarity of the third harmonic current and the seventh harmonic current of the zero-sequence current are calculated by using energy functions of the 1 st, 2 nd, 3 rd and 5 th scales to calculate the wavelet packet energy of the zero-sequence current of each outgoing line. And adding 14 original fault characteristic variables, wherein 22 fault characteristic variables are in total, and the 12 fault characteristic variables subjected to dimensionality reduction by the text method are respectively used as the input of a neural network, and a radial basis function is used as a core function to establish a neural network model and verify the accuracy and the operation speed of the model.
The result shows that the fault characteristic variables are reduced from 22 to 12, the accuracy rate of the big data model is increased from 95.6% to 99.6%, the operation time is reduced from 561ms to 425ms, and the number of neurons in the hidden layer is reduced from 5 to 2. The results show that the number of fault characteristic variables is obviously reduced after the dimension reduction by the method provided by the invention, the network structure is simpler when a big data neural network model is established, the construction and operation speed is accelerated, and the accuracy is further improved.
Although the embodiments of the present invention have been described with reference to the accompanying drawings, it is not intended to limit the scope of the present invention, and it should be understood by those skilled in the art that various modifications and variations can be made without inventive efforts by those skilled in the art based on the technical solution of the present invention.
Claims (8)
1. A big data dimension reduction method for distribution network fault line selection is characterized by comprising the following steps:
the method comprises the following steps: acquiring large data of distribution network fault line selection, and extracting fault characteristic variables;
step two: analyzing the correlation degree between the fault characteristic variables extracted in the step one and the line fault result, and calculating an MIC (minimum integrated coefficient) value between each fault characteristic variable and the line fault result;
step three: comparing the MIC value between each fault characteristic variable and the line fault result with the set MIC threshold value, and screening the fault characteristic variables;
the fault characteristic variables comprise voltage drop of each phase after the fault, amplitude and phase angle of zero-sequence current after the fault, wavelet packet energy of zero-sequence current of each outgoing line, amplitude and direction of quintuple harmonic of the zero-sequence current, amplitude of first half-wave after the fault, energy function after the fault, amplitude and polarity of transient zero-mode characteristic current SFB and amplitude and phase angle of zero-sequence current at an FTU installation position;
in the second step, the specific method for calculating the MIC value between each fault characteristic variable and the line fault result for the correlation degree between the fault characteristic variable extracted in the first step and the line fault result is as follows:
constructing a two-dimensional data set based on each fault characteristic variable and a line fault result;
respectively calculating MIC values of all two-dimensional data sets, namely MIC values between all fault characteristic variables and line fault results according to a maximum information coefficient MIC calculation formula;
the maximum information coefficient MIC is calculated according to the formula:
MIC(D)＝maxxy<B(n)M(D)xy＝maxxy<B(n)I(D,x,y)·log-1(min(x,y))
wherein D is a two-dimensional dataset, I (D, x, y) ═ max (I (D-G)), G represents a partition of the two-dimensional dataset D, D-G represents a distribution of the two-dimensional dataset D in the grid G, and I (D, x, y) represents mutual information based on the two-dimensional dataset D; b (n) the power of 0.6 or 0.55 of the total amount of data taken is an empirical value; x is a fault characteristic variable; and y is a line fault result.
2. The distribution network fault line selection big data dimension reduction method according to claim 1, wherein the distribution network fault line selection big data comprises current and voltage data, zero sequence voltage and current data, steady state electrical data, transient electrical data, remote telemetry data and fault recording data.
3. The distribution network fault line selection big data dimension reduction method according to claim 1, wherein in the first step, the specific method for obtaining the distribution network fault line selection big data is as follows:
building a simulation model, simulating faults of different lines, different positions and different types, and generating a plurality of groups of fault recording data;
recording fault recording data of each line through a fault recorder;
and acquiring fault recording data of each line as distribution network fault line selection big data.
4. The big data dimension reduction method for distribution network fault line selection according to claim 1, wherein the fault characteristic variables further include fault type and first half-wave polarity after fault.
5. The big data dimension reduction method for distribution network fault line selection according to claim 1, wherein in the first step, the specific method for extracting fault characteristic variables is as follows:
selecting a fault line, and representing different types of fault results by using different numbers;
calculating voltage drop values of each phase before and after the fault by using voltage effective values before and after the fault in the distribution network fault line selection big data;
extracting the amplitude and phase angle of the zero-sequence current after the fault, the amplitude and polarity of the transient zero-mode characteristic current SFB and the amplitude and phase angle of the zero-sequence current at the installation position of the FTU from the large data of the distribution network fault line selection;
according to the big data of the fault line selection of the distribution network, calculating the wavelet packet energy of the zero-sequence current of each outgoing line by using a dB15 basis function, a three-layer decomposition function and a fourth scale energy function;
and calculating an energy function after the fault according to the distribution network fault line selection big data, and extracting the quintic harmonic component of the zero-sequence current and the amplitude and the polarity of the first half wave after the fault.
6. The distribution network fault line selection big data dimension reduction method according to claim 1, wherein in the third step, the MIC value between each fault characteristic variable and the line fault result is compared with the set MIC threshold value, and the specific method for screening the fault characteristic variables is as follows:
comparing the MIC value between each fault characteristic variable and the line fault result with the set MIC threshold value;
if the MIC value between the fault characteristic variable and the line fault result is greater than or equal to the set MIC threshold value, indicating that the correlation degree between the fault characteristic variable and the line fault result is large, and keeping the fault characteristic variable;
and if the MIC value between the fault characteristic variable and the line fault result is smaller than the set MIC threshold value, indicating that the correlation degree between the fault characteristic variable and the line fault result is small, and rejecting the fault characteristic variable.
7. An apparatus for implementing the distribution network fault line selection big data dimension reduction method of any one of claims 1 to 6, characterized by comprising:
the fault recording data simulation module is used for simulating and generating fault recording data of each line;
the distribution network fault line selection big data acquisition module is used for acquiring fault recording data of each line from the fault recording data simulation module and taking the fault recording data of each line as distribution network fault line selection big data;
the fault characteristic variable extraction module is used for extracting fault characteristic variables of the large data of the distribution network fault line selection;
the maximum information coefficient analysis module is used for analyzing the correlation degree between the extracted fault characteristic variables and the line fault result and calculating the MIC value between each fault characteristic variable and the line fault result;
and the fault characteristic variable screening module is used for reserving the fault characteristic variable with large association degree and eliminating the fault characteristic variable with small association degree according to the MIC value between the fault characteristic variable and the line fault result.
8. The apparatus of claim 7, wherein the fault recording data simulation module comprises a low current grounding system having a plurality of outgoing lines, each outgoing line of the low current grounding system is provided with a feeder terminal device, and parameters are set through the feeder terminal devices; each outgoing line of the low-current grounding system is provided with a plurality of fault positions, a fault recorder is installed on each fault position, and fault recording data of each line during fault are recorded through the fault recorder.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810063390.5A CN108267673B (en) | 2018-01-23 | 2018-01-23 | Distribution network fault line selection big data dimension reduction method and device |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810063390.5A CN108267673B (en) | 2018-01-23 | 2018-01-23 | Distribution network fault line selection big data dimension reduction method and device |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN108267673A CN108267673A (en) | 2018-07-10 |
| CN108267673B true CN108267673B (en) | 2021-02-02 |
Family
ID=62776239
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201810063390.5A Active CN108267673B (en) | 2018-01-23 | 2018-01-23 | Distribution network fault line selection big data dimension reduction method and device |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN108267673B (en) |
Families Citing this family (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109581142B (en) * | 2019-01-25 | 2020-06-09 | 广东电网有限责任公司 | Novel multi-dimensional fusion high-voltage generator set stator single-phase earth fault detection method |
| CN110544158B (en) * | 2019-09-09 | 2022-07-08 | 北京百度网讯科技有限公司 | Information pushing method, device, equipment and readable storage medium |
| CN112964967B (en) * | 2021-03-19 | 2023-02-07 | 云南电网有限责任公司昆明供电局 | Line selection method for single-phase earth fault line of power distribution network switching power supply |
| CN113589098B (en) * | 2021-07-12 | 2024-06-07 | 国网河南省电力公司灵宝市供电公司 | Power grid fault prediction and diagnosis method based on big data driving |
Family Cites Families (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN105548744A (en) * | 2015-12-04 | 2016-05-04 | 国网技术学院 | Substation equipment fault identification method based on operation-detection large data and system thereof |
| CN107478963A (en) * | 2017-09-30 | 2017-12-15 | 山东海兴电力科技有限公司 | Single-phase ground fault line selecting method of small-electric current grounding system based on power network big data |
-
2018
- 2018-01-23 CN CN201810063390.5A patent/CN108267673B/en active Active
Also Published As
| Publication number | Publication date |
|---|---|
| CN108267673A (en) | 2018-07-10 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN108267673B (en) | Distribution network fault line selection big data dimension reduction method and device | |
| CN107451557B (en) | Power transmission line short-circuit fault diagnosis method based on empirical wavelet transform and local energy | |
| EP3043186B1 (en) | Method and system for identifying full parameters of element by fault recorder, and fault locating method | |
| CN104237731B (en) | Single-phase earth fault line selection method on basis of EEMD (ensemble empirical mode decomposition) and energy method for resonance earthed power distribution network | |
| CN109523165B (en) | Voltage sag state estimation method based on limited information | |
| CN103018632A (en) | Small current grounding system single-phase ground fault line selection method based on fisher information | |
| CN109782126B (en) | Power distribution network early fault detection method based on humanoid concept learning | |
| CN110873847A (en) | Single-phase earth fault positioning method based on SOGI (system-in-package) fundamental wave removal | |
| Yuan et al. | Study on fault line detection methods based on multi‐feature fusion in distribution systems | |
| CN103823158B (en) | Use the malfunction route selection method for resonant grounded system of not bending moment | |
| CN106646138B (en) | Distribution net work earthing fault localization method based on the conversion of more sample frequency wavelet character energy | |
| CN110703134B (en) | Small current grounding line selection and phase selection method based on fault sequence component | |
| CN113092933A (en) | LSTM-based single-phase earth fault line selection method and system | |
| CN117741336A (en) | Fault line selection method and device for composite grounding system | |
| CN114966324B (en) | Single-phase earth fault positioning method based on improved variational modal decomposition | |
| Sun et al. | A combined method for line selection of single phase to ground fault in compensated distributions based on evidence theory | |
| CN113419139B (en) | High-resistance grounding fault positioning method and related device for small-resistance grounding system | |
| Mishra et al. | A fast-executing single ended fault location method using transmission line characteristics for wind parks integrated HVDC network | |
| CN112345876A (en) | Fault positioning method and system suitable for interval DTU | |
| CN104793105A (en) | Power distribution network fault line selection method based on optimal FIR (finite impulse response) filter and hierarchical clustering | |
| CN106443352A (en) | Method for single-phase earth fault line selection of neutral non-grounding power grid | |
| Shi et al. | Fault Diagnosis of Transmission Lines Based on Improved Complete Ensemble Empirical Mode Decomposition | |
| Yang et al. | Multi-criteria Fusion Method for Line Selection of Single-phase-to-ground Fault with Small Current based on GRU Neural Network | |
| Ponukumati et al. | Unbalanced Distribution Network Cross-Country Fault Diagnosis Method with Emphasis on High-Impedance Fault Syndrome | |
| Cheng et al. | Single phase grounding fault line selection method of coal mine power grid based on smooth decomposition modified threshold denoising and waveform diversity |
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 |