CN108267673A - A kind of Distribution Network Failure route selection big data dimension reduction method and device - Google Patents
A kind of Distribution Network Failure route selection big data dimension reduction method and device Download PDFInfo
- Publication number
- CN108267673A CN108267673A CN201810063390.5A CN201810063390A CN108267673A CN 108267673 A CN108267673 A CN 108267673A CN 201810063390 A CN201810063390 A CN 201810063390A CN 108267673 A CN108267673 A CN 108267673A
- Authority
- CN
- China
- Prior art keywords
- fault
- distribution network
- route selection
- big data
- data
- 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.)
- Granted
Links
Classifications
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01R—MEASURING ELECTRIC VARIABLES; MEASURING MAGNETIC VARIABLES
- G01R31/00—Arrangements for testing electric properties; Arrangements for locating electric faults; Arrangements for electrical testing characterised by what is being tested not provided for elsewhere
- G01R31/08—Locating faults in cables, transmission lines, or networks
- G01R31/081—Locating faults in cables, transmission lines, or networks according to type of conductors
- G01R31/086—Locating faults in cables, transmission lines, or networks according to type of conductors in power transmission or distribution networks, i.e. with interconnected conductors
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01R—MEASURING ELECTRIC VARIABLES; MEASURING MAGNETIC VARIABLES
- G01R31/00—Arrangements for testing electric properties; Arrangements for locating electric faults; Arrangements for electrical testing characterised by what is being tested not provided for elsewhere
- G01R31/08—Locating faults in cables, transmission lines, or networks
- G01R31/088—Aspects of digital computing
Abstract
The invention discloses a kind of Distribution Network Failure route selection big data dimension reduction method and devices, first obtain Distribution Network Failure route selection big data, extract fault signature variable；Then the correlation degree between extracted fault signature variable and line fault result is analyzed, calculates the MIC value between each fault signature variable and line fault result；Compare the MIC value and the size of the MIC threshold values of setting between each fault signature variable and line fault result, screen fault signature variable.Present invention reduces big data dimensions when big data method being used to carry out Distribution Network Failure route selection, substantially reduce big data intractability.
Description
Technical field
The present invention relates to electric network fault route selection technical fields, and in particular to one kind is based on fault signature extraction and maximum information
The Distribution Network Failure route selection big data dimension reduction method and device of coefficient theory.
Background technology
Power grid is the mainstay of the national economy, and expanding economy be unable to do without the development of power grid, once occurrence of large-area power failure thing
Therefore immeasurable loss will be generated.Power distribution network is the network that electric energy is directly conveyed for user, and the normal operation of power distribution network is to society
It can develop and have great significance, distribution earth fault line selection is always a problem.
Currently, the widely used small current neutral grounding mode of domestic and international medium voltage distribution network neutral point, to avoid single-phase earthing occurs
Tripping causes power failure during failure.For small current grounding fault, due to fault current is faint, electric arc is unstable and it is random because
Reasons, the earth fault line selections such as element influence are relatively difficult.Traditional various selection methods only consider one or more of fault signatures
Variable, obtained route selection result precision are relatively low, it is impossible to meet the requirement of distribution network fault line selection.
With the development of intelligent grid, a large amount of intelligent monitoring devices are applied to power distribution network every aspect, thereby produce
A large amount of different types of data.Including the various electrical datas that power distribution network inside real-time monitoring device generates, also wrap
Include the various data such as non-electric quantities data such as weather, traffic and geography information outside power distribution network.
In order to analyze faulty line comprehensively using a variety of data, low current neutral grounding system fault route selecting accuracy rate is improved, is needed
Above a large amount of data are analyzed and processed with big data technology.Now using big data fault-line selecting method, need
Neural network model is established, using mass data as input, faulty line carries out data mining, model training as output.By
In various faults characteristic information is utilized, the method can effectively improve low current neutral grounding system fault route selecting accuracy rate, right
It is significant to improve distribution automation level.
In conclusion in the prior art since data dimension is too high, and often there is noise, when carrying out data mining
The problem of causing inadequate result precision, data mining overlong time, still lack effective solution.
Invention content
In order to overcome the above-mentioned deficiencies of the prior art, the present invention provides a kind of Distribution Network Failure route selection big data dimension reduction methods
And device, big data dimension when Distribution Network Failure route selection is carried out using big data method is reduced, is substantially reduced at big data
Manage difficulty.
The technical solution adopted in the present invention is：
A kind of Distribution Network Failure route selection big data dimension reduction method, includes the following steps：
Step 1：Distribution Network Failure route selection big data is obtained, extracts fault signature variable；
Step 2：Correlation degree between the fault signature variable and line fault result that are extracted in analytical procedure one,
Calculate the MIC value between each fault signature variable and line fault result；
Step 3：Compare the big of MIC value between each fault signature variable and line fault result and the MIC threshold values that set
It is small, screen fault signature variable.
Further, the Distribution Network Failure route selection big data include electric current and voltage data, residual voltage and current data,
Stable state electric data, transient state electric data, distal end telemetry and fault recorder data.
Further, in the step 1, the specific method for obtaining Distribution Network Failure route selection big data is：
Simulation model is built, simulates different circuits, different location, different types of faults, generates multigroup fault recorder data；
The fault recorder data of each circuit is recorded by fault oscillograph；
The fault recorder data of each circuit is obtained as Distribution Network Failure route selection big data.
Further, the fault signature variable includes zero sequence after each phase voltage drop value, failure after fault type, failure
Current amplitude and head after phase angle, the wavelet-packet energy of each outlet zero-sequence current, zero-sequence current quintuple harmonics amplitude and direction, failure
Half wave amplitude and energy function, zero modular character current amplitude of transient state and polarity and FTU installation places zero-sequence current width after polarity, failure
Value and phase angle.
Further, in the step 1, the specific method of extraction fault signature variable is：
A faulty line is selected, different types of faults result is represented with different digital；
Utilize each phase voltage drop value before and after voltage effective value calculating failure before and after failure in Distribution Network Failure route selection big data；
Zero-sequence current amplitude and phase angle, zero modular character electric current width of transient state after failure are extracted from Distribution Network Failure route selection big data
Value and the zero-sequence current amplitude and phase angle at polarity and FTU installings；
According to Distribution Network Failure route selection big data, decomposed using dB15 basic functions, three layers and the 4th Scale energy function calculates
The wavelet-packet energy of each outlet zero-sequence current；
According to Distribution Network Failure route selection big data, calculate energy function after failure, extraction zero-sequence current quintuple harmonics component and
The amplitude and polarity of first half-wave after failure.
Further, in the step 2, fault signature variable and the line fault result extracted in analytical procedure one
Between correlation degree, the specific method for calculating the MIC value between each fault signature variable and line fault result is：
Based on each fault signature variable and line fault result structure two-dimentional data set；
According to maximum information coefficient MIC calculation formula, the MIC value of all two-dimentional data sets is calculated respectively, i.e., institute is faulty
MIC value between characteristic variable and line fault result.
Further, maximum information coefficient MIC calculation formula are：
MIC (D)=maxXy ＜ B (n)M(D)xy=maxXy ＜ B (n)I(D,x,y)·log-1(min(x,y))
Wherein, D is two-dimentional data set, and I (D, x, y)=max (I (D │ G)), G represents a kind of division of two-dimentional data set D, D
│ G represent distributions of the two-dimentional data set D in grid G, and I (D, x, y) represents the mutual information based on two-dimentional data set D；B (n) fetches
It is an empirical value according to 0.6 or 0.55 power of total amount；X is fault signature variable；Y is line fault result.
Further, in the step 3, the MIC value between more each fault signature variable and line fault result is with setting
The size of fixed MIC threshold values, the specific method of screening fault signature variable are：
Compare the MIC value and the size of the MIC threshold values of setting between each fault signature variable and line fault result；
If the MIC value between fault signature variable and line fault result is greater than or equal to the MIC threshold values of setting, show
Correlation degree is big between fault signature variable and line fault result, retains the fault signature variable；
If the MIC value between fault signature variable and line fault result is less than the MIC threshold values of setting, show failure spy
Correlation degree is small between sign variable and line fault result, rejects the fault signature variable.
A kind of device for realizing above-mentioned Distribution Network Failure route selection big data dimension reduction method, including：
Fault recorder data analog module, for simulating the fault recorder data for generating each circuit；
Distribution Network Failure route selection big data acquisition module, for obtaining the event of each circuit from fault recorder data analog module
Hinder recorder data, using the fault recorder data of each circuit i.e. as Distribution Network Failure route selection electrical quantity big data；
Fault signature variable extraction module, for carrying out fault signature variable extraction to Distribution Network Failure route selection big data；
Maximum information coefficient analysis module is closed for analyzing between extracted fault signature variable and line fault result
Connection degree, the MIC value between calculating fault features variable and line fault result；
Fault signature Variable Selection module, for according to the MIC value between fault signature variable and line fault result, protecting
The fault signature variable that correlation degree is big is stayed, rejects the small fault signature variable of correlation degree.
Further, the fault recorder data analog module includes the small current neutral grounding system with a plurality of outlet, institute
The each outlet for stating small current neutral grounding system is installed with Feeder Terminal Unit, passes through Feeder Terminal Unit arrange parameter；It is described
Multiple abort situation are provided in every outlet of small current neutral grounding system, fault oscillograph is installed in abort situation, is passed through
Fault oscillograph records fault recorder data during each line fault.
Compared with prior art, the beneficial effects of the invention are as follows：
(1) present invention extracts the fault signature number significant to failure line selection from Distribution Network Failure route selection big data
According to, carry out preliminary Distribution Network Failure route selection big data dimensionality reduction, to fault signature variable and circuit whether the association journey between failure
Degree carries out maximum information Coefficient Algorithm analysis, obtain various fault signature variables and circuit whether the MIC value between failure, retain
The larger fault signature of correlation degree rejects the smaller fault signature of correlation degree, carries out further big data dimensionality reduction, effectively
Big data dimension when Distribution Network Failure route selection is carried out using big data method is reduced, substantially reduces big data intractability；
(2) the present invention is based on failure line selection meal methods, extract useful fault signature variable, have effectively carried out distribution event
Hinder the preliminary dimensionality reduction of route selection big data；
(3) present invention uses advanced relevance maximum information coefficient analysis algorithm, accurately comprehensively has studied various events
Hinder the correlation degree of characteristic variable and line fault, carry out the further dimensionality reduction of effective big data on this basis；
(4) present invention carries out Screening Treatment to the original big data of Distribution Network Failure route selection, retains useful data, rejects unrelated number
According to so as to improve big data failure line selection precision and arithmetic speed.
Description of the drawings
The accompanying drawings which form a part of this application are used for providing further understanding of the present application, and the application's shows
Meaning property embodiment and its explanation do not form the improper restriction to the application for explaining the application.
Fig. 1 is Distribution Network Failure route selection big data dimension reduction method flow chart disclosed by the embodiments of the present invention；
Fig. 2 is simulation model simplification figure disclosed by the embodiments of the present invention；
Fig. 3 is the selection method flow chart disclosed by the embodiments of the present invention based on Distribution Network Failure route selection big data；
Specific embodiment
It is noted that following detailed description is all illustrative, it is intended to provide further instruction to the application.It is unless another
It indicates, all technical and scientific terms used herein has usual with the application person of an ordinary skill in the technical field
The identical meanings of understanding.
It should be noted that term used herein above is merely to describe specific embodiment, and be not intended to restricted root
According to the illustrative embodiments of the application.As used herein, unless the context clearly indicates otherwise, otherwise singulative
It is also intended to include plural form, additionally, it should be understood that, when in the present specification using term "comprising" and/or " packet
Include " when, indicate existing characteristics, step, operation, device, component and/or combination thereof.
As background technology is introduced, in the prior art since data dimension is too high, and often there is noise, into
The problem of inadequate result precision, data mining overlong time can be caused during row data mining, present applicant proposes a kind of distribution events
Hinder route selection big data dimension reduction method and device, including the extraction of fault signature variable and correlation analysis two parts, can carry significantly
High big data route selection model accuracy and arithmetic speed.
As shown in Figure 1, present embodiments providing a kind of Distribution Network Failure route selection big data dimension reduction method, include the following steps：
Step 101：Distribution Network Failure route selection big data is obtained, fault signature extraction is carried out to Distribution Network Failure route selection big data,
Obtain fault signature variable.According to selection method experience, the fault signature data significant to failure line selection are extracted, into
The preliminary Distribution Network Failure route selection big data dimensionality reduction of row；
Step 102：To the fault signature variable and circuit that are extracted in step 101 whether the association between fail result
Degree carries out maximum information coefficient (Maximal Information Coefficient, MIC) Algorithm Analysis, obtains various events
Hinder characteristic and circuit whether the MIC value between failure, retain the larger fault signature of correlation degree, reject correlation degree compared with
Small fault signature carries out further big data dimensionality reduction.
The Distribution Network Failure route selection big data includes the various electrical datas that real-time monitoring device generates inside power distribution network
With non-electric quantity data, most of data are all obtained from electric system built-in system, such as electric current, voltage data, zero sequence electricity
Pressure, current data, stable state electric data, transient state electric data, distal end telemetry and fault recorder data.For these numbers
According to, in order to ensure power system information safety, need that positive isolating device is installed.
In a step 101, fault signature extraction refers to be extracted from initial data with important physical meaning or statistics
The characteristic variable of meaning, initial data that can be higher to dimension carry out effective dimensionality reduction.Since initial data especially electrically counts
Too high according to dimension, data volume is too big, and containing some useless information, it is therefore necessary to carry out feature extraction, has extracted
Fault signature variable is analyzed.
There are many methods for fault signature extraction, and to ensure the validity of extraction data, generally being determined according to selection method will
The characteristic of extraction.In failure line selection, different types of faults are represented using different digital as a result, using voltage before and after failure
Virtual value calculates each phase voltage landing, and each outlet is calculated using dB15 basic functions, three layers of decomposition and the 4th Scale energy function
The wavelet-packet energy of zero-sequence current records zero-sequence current amplitude and phase angle everywhere, computation energy function, extraction zero-sequence current five times
The amplitude and polarity of first half-wave after harmonic component and failure；According to fault recorder data, zero modular character current amplitude of transient state is extracted
(setting isolated neutral system sample frequency is 0~2kHz, and setting arc suppression coil earthing system sample frequency is with polarity
200Hz~2000kHz), extract the zero-sequence current amplitude and phase angle at FTU installings.By more than variable as fault signature variable,
It can effectively dimensionality reduction.
In step 2, maximum information coefficient MIC is a kind of measurement of correlation degree between effective analysis mass data
Index, is a kind of algorithm proposed based on mutual information technology, and confidence level is high.MIC can be excavated effectively between two data item
Relationship it is strong and weak, this provides the foundation for further data mining.
MIC has versatility and fairness.Wherein, versatility refer to it be applicable not only to functional dependence be also applied for it is non-
Functional dependence is not only suitable for linear dependence and is also applied for non-linear dependencies；Fairness refers to that two different functions close
System is still consistent by their MIC value during identical noise jamming namely noise is influenced caused by MIC between variable
Functional relation it is unrelated.
When sample size is sufficiently large, it is hidden that the fairness and versatility of MIC can ensure that it can capture various complexity
The association of Tibetan, and specific functional form is not limited to, it is referred to as the correlation of 21 century.It is carried out with MIC accurate comprehensive
Correlation degree analysis between fault signature data and line fault further rejects hash and carries out dimensionality reduction, can be effective
Improve big data model route selection precision and arithmetic speed.
Maximum information coefficient (MIC) principle is as follows：
For some data variable, information content contained in stochastic variable is weighed, is defined as follows：
Wherein piThe probability of happening of event i is represented,Mutual information between two stochastic variables is：
I (X, Y)=H (X)+H (Y)-H (X, Y)
The orderly two-dimentional data set D limited to one:(x, y), maximum information coefficient (MIC) computational methods are as follows：
MIC (D)=maxXy ＜ B (n)M(D)xy=maxXy ＜ B (n)I(D,x,y)·log-1(min(x,y))
Wherein I (D, x, y)=max (I (D │ G)), G represents a kind of division of finite data collection D, and D │ G represent data set D and exist
Distribution in grid G, I (D, x, y) represent the mutual information based on data set D.B (n) takes 0.6 or 0.55 time of total amount of data
Side is an empirical value；X, y are two vectors, and D is data set.
Fault signature extraction is imported in MATLAB as a result, including each phase voltage drop value, zero-sequence current amplitude after failure
With half wave amplitude first after phase angle, the wavelet-packet energy of each outlet zero-sequence current, zero-sequence current quintuple harmonics amplitude and direction, failure
With energy function, zero modular character current amplitude of transient state and polarity after polarity, failure and FTU installation places zero-sequence current amplitude and phase
Angle.Using each fault signature variable as a row vector, whether fail result as another row vector, forms two dimension to circuit
Data set；
Using the principle algorithm of maximum information coefficient (MIC) principle, all fault signature variables and circuit event are calculated one by one
Hinder the MIC value between result, obtain the MIC value of each two vector.
Compare the MIC value and the size of the MIC threshold values of setting between each fault signature variable and line fault result.By
MIC principles understand the value of the MIC between two variables between 0 to 1, represent that correlation degree is bigger closer to 1, closer to 0
Then represent that correlation degree is smaller.Since this method is related to a variety of variables, with reference to actual conditions, it is associated with MIC=0.1 as division
The threshold value of degree is an empirical value.
If the MIC value between fault signature variable and line fault result is greater than or equal to the MIC threshold values of setting, show
Correlation degree is larger between fault signature variable and line fault result, retains the fault signature variable；If fault signature variable
MIC value between line fault result is less than the MIC threshold values of setting, then show fault signature variable and line fault result it
Between correlation degree it is smaller, reject the fault signature variable；It realizes and carries out further big data dimensionality reduction.
Another exemplary embodiment of the application provides and a kind of realizes above-mentioned Distribution Network Failure route selection big data dimensionality reduction
The device of method, the device include：
Fault recorder data analog module, for simulating the fault recorder data for generating each circuit；
Distribution Network Failure route selection big data acquisition module, for obtaining the event of each circuit from fault recorder data analog module
Hinder recorder data, using the fault recorder data of each circuit i.e. as Distribution Network Failure route selection electrical quantity big data；
Fault signature variable extraction module, for carrying out fault signature variable extraction to Distribution Network Failure route selection big data；
Maximum information coefficient analysis module, it is maximum for being carried out to the fault signature variable extracted and line fault result
Information coefficient is analyzed, and obtains the MIC value between fault signature variable and line fault result；
Fault signature Variable Selection module, for according to the MIC value between fault signature variable and line fault result, protecting
The fault signature variable that correlation degree is big is stayed, rejects the small fault signature variable of correlation degree.
In the present embodiment, the fault recorder data analog module includes the small current neutral grounding system with a plurality of outlet
System, each outlet of the small current neutral grounding system is installed with Feeder Terminal Unit, passes through Feeder Terminal Unit arrange parameter；
Multiple abort situation are provided in every outlet of the small current neutral grounding system, fault oscillograph is installed in abort situation,
Fault recorder data during each line fault is recorded by fault oscillograph.
Technical scheme of the present invention is described in detail with reference to a specific embodiment.
Embodiment one
A kind of Distribution Network Failure route selection big data dimension reduction method that the present embodiment proposes, includes the following steps：
Step 1 obtains Distribution Network Failure route selection big data, and fault signature extraction is carried out to Distribution Network Failure route selection big data.Root
According to selection method experience, using rational method, the fault signature data significant to failure line selection are extracted, are carried out preliminary
Big data dimensionality reduction；
Step 2, the fault signature data variable and circuit extracted to step 1 whether the association journey between fail result
Degree carries out maximum information coefficient (MIC) Algorithm Analysis, obtain various fault signature data and circuit whether the MIC between failure
Value retains the larger fault signature of correlation degree, rejects the smaller fault signature of correlation degree, carry out further big data drop
Dimension.
The step 1, Distribution Network Failure route selection big data include the various electrical of power distribution network inside real-time monitoring device generation
Data and non-electric quantity data are measured, including electric current, voltage data, residual voltage, current data, stable state electric data, transient state electricity
Destiny evidence, distal end telemetry and fault recorder data.Most of data are all obtained from electric system built-in system, right
In these data, in order to ensure power system information safety, need that positive isolating device is installed.
Electric data in Distribution Network Failure route selection big data is simulated using simulation model.It is calculated using electrical power system transient
Machine Computer Aided Design (Power Systems Computer Aided Design, PSCAD) software builds simulation model, model
To have the small current neutral grounding system of five outlets, each outlet is installed with Feeder Terminal Unit (Feeder Terminal
Unit, FTU).Circuit, load, power supply, transformer parameter are set.Arc suppression coil parameter is 6.0H, and the positive order parameter of circuit is：Just
Sequence resistance R1=1.7E-4 Ω/m, positive sequence inductance XL1=3.8E-4 Ω/m, positive sequence capacitance XC1=32.84M Ω/m；Circuit zero sequence is joined
Number is：Zero sequence resistance R0=2.3E-4 Ω/m, zero sequence inductance XL0=1.72E-3 Ω/m, zero sequence capacitance XC0=53.08M Ω/m.Often
Outlet sets five abort situation, installs fault oscillograph, voltage, current failure recording number when recording each line fault
According to.Model simplification figure is as shown in Figure 2.
Isolated neutral is set, through two kinds of earthed systems of grounding through arc；It is respectively A phase faults to set fault type
It is grounded AG, B phase fault ground connection BG, C phase fault ground connection tri- kinds of situations of CG；It is respectively 0 degree, 45 degree, 90 degree to set fault angle；Setting
Fault resstance is respectively 0 Ω, 10 Ω, 50 Ω；Mold cycle is set to run, common property gives birth to 1350 groups of fault messages.
As shown in figure 3, fault signature extraction is carried out to emulation electric data.Every circuit produces 225 groups of events during emulation
Hinder data, fault recorder data is handled, using rational method, extract useful feature data.Extract following 14 kinds
Fault signature data：Fault type, each phase voltage landing after failure, zero-sequence current amplitude and phase angle after failure, each outlet zero sequence
The wavelet-packet energy of electric current, zero-sequence current quintuple harmonics amplitude and direction, first half wave amplitude and polarity after failure, energy after failure
Function, zero modular character electric current of transient state (Special Frequency Band, SFB) amplitude and polarity, FTU installation places zero-sequence current
Amplitude and phase angle.Fault signature variable is obtained, and normalize according to each outlet.From original electrical data extract more than weight
Fault signature is wanted, has effectively carried out Data Dimensionality Reduction.
In the step 2, with MIC carry out accurate comprehensive fault signature data variable and circuit whether fail result
Between correlation degree analysis, further reject hash carry out dimensionality reduction, big data model route selection precision can be effectively improved
With arithmetic speed.
The correlation degree of fault signature data variable and line failure is studied using MIC algorithms.By each failure
Characteristic uses matrix labotstory (Matrix as a vector, line fault result as another vector
Laboratory, MATLAB) software calculate two vector MIC values, ask for the average value of 5 circuits, it is as a result as follows：
The MIC value of 1 each fault signature variable of table and line failure
As can be seen from the above results, in two kinds of earthed systems, zero-sequence current amplitude and phase angle, wavelet-packet energy, head
Half wave amplitude, zero modular character current amplitude of transient state and polarity and FTU amplitudes are the failure larger with line fault correlation
Characteristic variable, this result and the advantage and disadvantage of traditional selection method are consistent.Wherein in isolated neutral system, wavelet packet energy
Amount, first half wave amplitude and zero modular character current amplitude of transient state and polarity are mostly important variables；And it is connect through arc suppression coil
In ground system, first half wave amplitude and FTU polarity are even more important variables.In two kinds of earthed systems, fault type and first half
Wave polarity be with the smaller fault signature variable of line fault correlation, can be rejected in data processing, into traveling
The dimensionality reduction of one step only leaves 12 kinds of fault signature variables in this way, i.e., each phase voltage landing after failure, zero-sequence current width after failure
Value and phase angle, the wavelet-packet energy of each outlet zero-sequence current, zero-sequence current quintuple harmonics amplitude and direction, first half wave amplitude after failure
It is worth, energy function, zero modular character electric current SFB amplitudes of transient state and polarity, FTU installation places zero-sequence current amplitude and phase angle after failure,
Big data dimension has obtained further reduction.
Establish the validity that big data model verify the dimension reduction method.Use data mining platform statistical product and clothes
Be engaged in solution modeling device (Statistical Product and Service Solutions Modeler, SPSS
Modeler) big data route selection model is built.Using various fault feature vectors as input data, faulty line is as output
Data establish neural network, set neural network parameter, and neural network is trained.
To verify that method proposed by the present invention can effectively carry out Data Dimensionality Reduction really, improve failure line selection accuracy with
Arithmetic speed adds in a few unrelated fault signature variable to add in interference：Zero-sequence current three times, the seventh harmonic current amplitude with
Polarity calculates the wavelet-packet energy of each outlet zero-sequence current with the 1st, 2,3,5 Scale energy functions.In addition 14 kinds of original events
Hinder characteristic variable, one shares 22 kinds of fault signature variables, distinguishes with 12 kinds of fault signature variables after context of methods dimensionality reduction
As the input of neural network, using radial basis function as core function, neural network model is established, verifies model accuracy
With arithmetic speed.
The result shows that fault signature variable is reduced to 12 kinds from 22 kinds, big data model accuracy rate rises to from 95.6%
99.6%, operation time drops to 425ms from 561ms, and hidden layer neuron quantity drops to 2 by 5, it was demonstrated that this present invention proposes
Method can effectively improve really and effectively improve big data model route selection precision and arithmetic speed.It passes through as can be seen from the results
After crossing method dimensionality reduction proposed by the present invention, fault signature variable quantity significantly reduces, net when establishing big data neural network model
Network structure is simpler, and structure is accelerated with arithmetic speed, and accuracy is further improved.
Above-mentioned, although the foregoing specific embodiments of the present invention is described with reference to the accompanying drawings, not protects model to the present invention
The limitation enclosed, those skilled in the art should understand that, based on the technical solutions of the present invention, those skilled in the art are not
Need to make the creative labor the various modifications or changes that can be made still within protection scope of the present invention.
Claims (10)
1. a kind of Distribution Network Failure route selection big data dimension reduction method, it is characterized in that, include the following steps：
Step 1：Distribution Network Failure route selection big data is obtained, extracts fault signature variable；
Step 2：Correlation degree between the fault signature variable and line fault result that are extracted in analytical procedure one calculates
MIC value between each fault signature variable and line fault result；
Step 3：Compare the MIC value and the size of the MIC threshold values of setting between each fault signature variable and line fault result,
Screen fault signature variable.
2. Distribution Network Failure route selection big data dimension reduction method according to claim 1, it is characterized in that, the Distribution Network Failure route selection
Big data is distant including electric current and voltage data, residual voltage and current data, stable state electric data, transient state electric data, distal end
Measured data and fault recorder data.
3. Distribution Network Failure route selection big data dimension reduction method according to claim 1, it is characterized in that, in the step 1, obtain
The specific method for taking Distribution Network Failure route selection big data is：
Simulation model is built, simulates different circuits, different location, different types of faults, generates multigroup fault recorder data；
The fault recorder data of each circuit is recorded by fault oscillograph；
The fault recorder data of each circuit is obtained as Distribution Network Failure route selection big data.
4. Distribution Network Failure route selection big data dimension reduction method according to claim 1, it is characterized in that, the fault signature variable
Including zero-sequence current amplitude after phase voltage drop value each after fault type, failure, failure and phase angle, each outlet zero-sequence current it is small
It is first half wave amplitude and energy function after polarity, failure after wave packet energy, zero-sequence current quintuple harmonics amplitude and direction, failure, temporary
Zero modular character current amplitude of state and polarity and FTU installation places zero-sequence current amplitude and phase angle.
5. Distribution Network Failure route selection big data dimension reduction method according to claim 1, it is characterized in that, in the step 1, carry
The specific method for taking fault signature variable is：
A faulty line is selected, different types of faults result is represented with different digital；
Utilize each phase voltage drop value before and after voltage effective value calculating failure before and after failure in Distribution Network Failure route selection big data；
Extracted from Distribution Network Failure route selection big data zero-sequence current amplitude and phase angle after failure, zero modular character current amplitude of transient state with
Zero-sequence current amplitude and phase angle at polarity and FTU installings；
According to Distribution Network Failure route selection big data, decomposed using dB15 basic functions, three layers and the calculating of the 4th Scale energy function respectively goes out
The wavelet-packet energy of line zero-sequence current；
According to Distribution Network Failure route selection big data, energy function after failure is calculated, extracts zero-sequence current quintuple harmonics component and failure
The amplitude and polarity of first half-wave afterwards.
6. Distribution Network Failure route selection big data dimension reduction method according to claim 1, it is characterized in that, it is right in the step 2
Correlation degree between fault signature variable and line fault result that analytical procedure one is extracted, calculates each fault signature variable
The specific method of MIC value between line fault result is：
Based on each fault signature variable and line fault result structure two-dimentional data set；
According to maximum information coefficient MIC calculation formula, the MIC value of all two-dimentional data sets, i.e., all fault signatures are calculated respectively
MIC value between variable and line fault result.
7. Distribution Network Failure route selection big data dimension reduction method according to claim 6, it is characterized in that, maximum information coefficient MIC
Calculation formula is：
MIC (D)=maxXy ＜ B (n)M(D)xy=maxXy ＜ B (n)I(D,x,y)·log-1(min(x,y))
Wherein, D is two-dimentional data set, and I (D, x, y)=max (I (D │ G)), G represents a kind of division of two-dimentional data set D, D │ G generations
Distributions of the table two-dimentional data set D in grid G, I (D, x, y) represent the mutual information based on two-dimentional data set D；B (n) access is according to total
0.6 or 0.55 power of amount is an empirical value；X is fault signature variable；Y is line fault result.
8. Distribution Network Failure route selection big data dimension reduction method according to claim 1, it is characterized in that, in the step 3, than
MIC value and the size of the MIC threshold values of setting between more each fault signature variable and line fault result, screening fault signature become
The specific method of amount is：
Compare the MIC value and the size of the MIC threshold values of setting between each fault signature variable and line fault result；
If the MIC value between fault signature variable and line fault result is greater than or equal to the MIC threshold values of setting, show failure
Correlation degree is big between characteristic variable and line fault result, retains the fault signature variable；
If the MIC value between fault signature variable and line fault result is less than the MIC threshold values of setting, show that fault signature becomes
Correlation degree is small between amount and line fault result, rejects the fault signature variable.
9. a kind of device of Distribution Network Failure route selection big data dimension reduction method realized described in any one of claim 1-8, special
Sign is, including：
Fault recorder data analog module, for simulating the fault recorder data for generating each circuit；
Distribution Network Failure route selection big data acquisition module, for obtaining the failure record of each circuit from fault recorder data analog module
Wave number evidence, using the fault recorder data of each circuit i.e. as Distribution Network Failure route selection electrical quantity big data；
Fault signature variable extraction module, for carrying out fault signature variable extraction to Distribution Network Failure route selection big data；
Maximum information coefficient analysis module, for analyzing being associated between extracted fault signature variable and line fault result
Degree calculates the MIC value between each fault signature variable and line fault result；
Fault signature Variable Selection module, for according to the MIC value between fault signature variable and line fault result, retaining and closing
The big fault signature variable of connection degree rejects the small fault signature variable of correlation degree.
10. the device according to claim 9 for realizing Distribution Network Failure route selection big data dimension reduction method, it is characterized in that, it is described
Fault recorder data analog module include with a plurality of outlet small current neutral grounding system, the small current neutral grounding system it is each
Outlet is installed with Feeder Terminal Unit, passes through Feeder Terminal Unit arrange parameter；Every of the small current neutral grounding system goes out
Multiple abort situation are provided on line, fault oscillograph is installed in abort situation, each circuit is recorded by fault oscillograph
Fault recorder data during failure.
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 true CN108267673A (en) | 2018-07-10 |
| CN108267673B 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) |
Cited By (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109581142A (en) * | 2019-01-25 | 2019-04-05 | 广东电网有限责任公司 | A kind of High-voltage generator set stator single-phase earthing fault detection method of novel multi-vitamin degree fusion |
| CN110544158A (en) * | 2019-09-09 | 2019-12-06 | 北京百度网讯科技有限公司 | information pushing method, device, equipment and readable storage medium |
| CN112964967A (en) * | 2021-03-19 | 2021-06-15 | 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 |
| CN113589098A (en) * | 2021-07-12 | 2021-11-02 | 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 drive |
Citations (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
Patent Citations (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 |
Non-Patent Citations (2)
| Title |
|---|
| ZHENG SHAO等: "A Fault Line Selection Method for Small Current Grounding System Based on Big Data", 《2016 IEEE PES ASIA-PACIFIC POWER AND ENERGY CONFERENCE-XI"AN-CHINA》 * |
| 陈凯俊: "自适应快速最大信息系数算法实现", 《微电子学与计算机》 * |
Cited By (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109581142A (en) * | 2019-01-25 | 2019-04-05 | 广东电网有限责任公司 | A kind of High-voltage generator set stator single-phase earthing fault detection method of novel multi-vitamin degree fusion |
| CN110544158A (en) * | 2019-09-09 | 2019-12-06 | 北京百度网讯科技有限公司 | information pushing method, device, equipment and readable storage medium |
| 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 |
| CN112964967A (en) * | 2021-03-19 | 2021-06-15 | 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 |
| 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 |
| CN113589098A (en) * | 2021-07-12 | 2021-11-02 | 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 drive |
Also Published As
| Publication number | Publication date |
|---|---|
| CN108267673B (en) | 2021-02-02 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Zhang et al. | Transmission line boundary protection using wavelet transform and neural network | |
| CN108267673A (en) | A kind of Distribution Network Failure route selection big data dimension reduction method and device | |
| CN107451557A (en) | Transmission line short-circuit fault diagnostic method based on experience wavelet transformation and local energy | |
| CN101924354B (en) | Artificially neural network routing method for distribution network failure by using S-transforming energy sampling property | |
| CN106771735A (en) | The treating method and apparatus of simulated failure | |
| Usama et al. | Design and implementation of a wavelet analysis‐based shunt fault detection and identification module for transmission lines application | |
| Stojkovic | Computer-aided design in power engineering: Application of software tools | |
| CN108647479A (en) | A kind of surge arrester failure transient-wave diagnostic method and device | |
| CN107515339A (en) | A kind of Risk Identification Method and system based on DC current distribution situation | |
| Dey et al. | Rough-granular approach for impulse fault classification of transformers using cross-wavelet transform | |
| Mehinović et al. | Application of artificial intelligence methods for determination of transients in the power system | |
| Ngaopitakkul et al. | Application of probabilistic neural network with transmission and distribution protection schemes for classification of fault types on radial, loop, and underground structures | |
| CN103823158A (en) | Resonant earthed system fault line selection method with invariant moments | |
| Wu et al. | Transmission line fault cause identification method based on transient waveform image and MCNN-LSTM | |
| CN113092933A (en) | LSTM-based single-phase earth fault line selection method and system | |
| Aljohani | Centralized fault detection and classification for motor power distribution centers utilizing MLP-NN and stockwell transform | |
| Mishra et al. | Classification of faults in a TCSC compensated transmission line using data mining algorithm | |
| Wali et al. | High impedance fault detection based on power spectrum technique | |
| Sima et al. | A framework for automatically cleansing overvoltage data measured from transmission and distribution systems | |
| Shu et al. | A detection method of high impedance arcing fault for distribution network with distributed generation based on CEEMDAN and TEO algorithm | |
| Ali et al. | Studies on the application of wavelet families for a high impedance fault location algorithm in a distribution network | |
| Sendilkumar et al. | A new technique to classify transient events in power transformer differential protection using S-transform | |
| CN112345876A (en) | Fault positioning method and system suitable for interval DTU | |
| Chen et al. | Identifying single‐phase to ground fault based on line multi‐dimensional data in cyber‐physical distribution system | |
| Patel et al. | Transient energy and its impact on transmission line faults |
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 |