CN108562821A - A kind of method and system determining Single-phase Earth-fault Selection in Distribution Systems based on Softmax - Google Patents
A kind of method and system determining Single-phase Earth-fault Selection in Distribution Systems based on Softmax Download PDFInfo
- Publication number
- CN108562821A CN108562821A CN201810432276.5A CN201810432276A CN108562821A CN 108562821 A CN108562821 A CN 108562821A CN 201810432276 A CN201810432276 A CN 201810432276A CN 108562821 A CN108562821 A CN 108562821A
- Authority
- CN
- China
- Prior art keywords
- fault
- data
- regression models
- optimized parameter
- failure
- 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
- G01R31/50—Testing of electric apparatus, lines, cables or components for short-circuits, continuity, leakage current or incorrect line connections
-
- 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 method and system determining Single-phase Earth-fault Selection in Distribution Systems based on Softmax, including:Obtain the fault data for each presetting node in power distribution network in multiple default nodes of circuit;The fault data is handled, obtains fault signature data, and establish failure identification and fault signature database;Data in the failure identification and fault signature database are divided into training set and test set according to fault type, and carry out route selection using Softmax regression models and training set according to default iterations and recognize the determining optimized parameter of training, to determine the corresponding Softmax regression models of optimized parameter;Fault type and position are predicted according to the real time fail data of power grid using the corresponding Softmax regression models of the optimized parameter, to determine power distribution network one-way earth fault route selection.Maintenance data of the present invention is trained and verifies to Softmax regression models, improves data-handling capacity and route selection accuracy, promotes the research and development and application of new distribution terminal.
Description
Technical field
The present invention relates to power distribution network risk analysis and control technology fields, and are based on more particularly, to one kind
Softmax determines the method and system of Single-phase Earth-fault Selection in Distribution Systems.
Background technology
Small current neutral grounding problem is global problem, and the difficult point that fault identification is primarily present is at present:First, small
Electric current single-phase earthing, fault current are small, it is difficult to detect;Second is that system structure is complicated, neutral grounding mode is different, abort situation
Difference, fault type is different, and caused fault signature is also different, and the threshold value of detection is difficult to adjust;Third, fault characteristic signals
Pass through catadioptric, mutual superimposed interference, it is not easy to isolate significant characteristic quantity in network.
Meanwhile, there are following three points in the main reason for poor for existing methods validity:(1) without specific fault data
Analysis, only adjusts short circuit current, and small current neutral grounding current value is smaller, and simple limiting short-circuit current is difficult to be carried out to small current neutral grounding
Judge;(2) be based only on a certain characteristic signal and carry out route selection, due to different small current grounding faults because of neutral grounding mode not
Different with, ground resistance and diversification is presented, single features signal can not describe all types of fault signatures, therefore based on one kind
The method validity that characteristic signal carries out route selection is just limited to;(3) since practical distribution network can not largely carry out ground connection event
Barrier simulation, therefore the mode of the selection method generally use dynamic simulation based on multi signal data characteristics amount carries out, but Digital Simulation
The case where modeling can not accurately simulate true distribution net equipment completely, can ignore the influence factors such as transient characterisitics, simultaneously because imitative
The limitation of true step-length, has certain limitations the transient signal frequency range of offer, proposed method is caused to be answered in systems in practice
It is limited with effect.
Softmax regression algorithms are that a kind of typical more classification problem regression algorithms are widely used in machine before this
Learning areas among more classification purposes such as animals and plants classification, also has for Handwritten Digit Recognition and carries out hydraulic pump using the algorithm
Fault diagnosis application example.But Softmax regression algorithms have no application in field of power.In field of power,
In recent years Chinese each province configures a large amount of recording type fault detector, a large amount of true distribution network failure data of acquisition, while but
It is insufficient to the processing analysis ability of data.
If currently, wanting to propose the single-phase grounding selecting strategy based on authentic and valid data, need to solve the problems, such as following:
First, obtaining the Wave data under effective failure generating state, which will occur position with the type of failure with failure,
Location information is circuit number in route selection strategy, while the sample rate of enrolled data waveform wants sufficiently high, including number of faults
According to feature;It second is that the data waveform of admission cannot be used directly for fault identification, need to be pre-processed, be isolated and failure with reaching
The characteristic quantity of data characteristics strong correlation once carries out fault identification;Third, to propose the application process of fault signature, ordinary circumstance
Under, fault signature and fault location information and nonlinear correspondence relation will classify by regression algorithm, to reach effective choosing
Select the purpose of faulty line.
Therefore, because any of the above shortcomings and deficiencies make full use of true power distribution network to propose more effective route selection strategy
In fault data, a kind of wire selection method for power distribution network single phase earthing failure is needed, to solve accurately determine and break down
Route the problem of.
Invention content
The present invention proposes a kind of method and system determining Single-phase Earth-fault Selection in Distribution Systems based on Softmax, with solution
The problem of certainly can not accurately determining the route to break down.
To solve the above-mentioned problems, according to an aspect of the invention, there is provided one kind determining power distribution network based on Softmax
The method of single-phase earth fault line selection, which is characterized in that the method includes:
Step 1, the fault data for each presetting node in power distribution network in multiple default nodes of circuit is obtained, wherein described
Fault data includes:Three-phase voltage data, three-phase current data, the phase data of voltage, the phase information of electric current and failure
The absolute time information of generation;
Step 2, the fault data is handled, obtain fault signature data, and according to the fault signature data,
Fault type and abort situation mark establish failure identification and fault signature database, wherein the fault signature data include more
A characteristic quantity;
Step 3, by the data in the failure identification and fault signature database according to fault type be divided into training set and
Test set, and carry out route selection identification using Softmax regression models and training set according to default iterations and train determination optimal
Parameter, to determine the corresponding Softmax regression models of optimized parameter;
Step 4, using the corresponding Softmax regression models of the optimized parameter according to the real time fail data pair event of power grid
Barrier type and position are predicted, to determine power distribution network one-way earth fault route selection.
Preferably, wherein acquiring the fault data at each default node using recording type fault detector.
Preferably, wherein described handle the fault data, fault signature data are obtained, including:
The fault data is handled, synthesis obtains failure volume, wherein the failure volume includes:Frequency
Variable quantity, the variable quantity of voltage, the change rate of frequency, the change rate of voltage, the change rate of electric current, phase angle change rate, have
The change rate of power, the change rate of inactivity, frequency is distorted with change rate, the current harmonics of power, voltage harmonic is distorted, power
The each harmonic component of factor and voltage and current;
The failure volume is handled according to rough set theory, the low failure volume of the degree of correlation is got rid of, obtains
Take fault signature data.
Preferably, wherein the default iterations of the basis carry out route selection using Softmax regression models and training set and distinguish
Know training and determines optimized parameter, to determine the corresponding Softmax regression models of optimized parameter, including:
Determine that the cost function of Softmax regression algorithms, cost function are:
Wherein, the sample of Softmax regression models comes from k class, shares m, then the training set being made of these samples
For { (x(1),y(1)),(x(2),y(2)),…,(x(m),y(m)), whereinLabel:y(t)∈ { 1,2 ..., k }, for
Given input x, it is assumed that function estimates probability value p (y=j | x) for each classification j, for estimating each point of x
The probability that class result occurs;
Route selection identification training is carried out using gradient descent method to the cost function according to default iterations, is determined optimal
Parameter obtains the corresponding Softmax regression models of optimized parameter.
Preferably, wherein the method further includes:
The fault signature data of sample in test set are corresponded to using the optimized parameter corresponding Softmax regression models
Fault type and abort situation predicted;
Prediction accuracy is calculated, and judges whether prediction accuracy is more than or equal to default accuracy threshold value;Wherein,
If prediction accuracy is more than or equal to default accuracy threshold value, directly utilizes and have determined that optimized parameter is corresponding
Softmax regression models determine Single-phase Earth-fault Selection in Distribution Systems;
If prediction accuracy is less than default accuracy threshold value, increase iterations according to iteration step length is preset, and return
Step 3 carries out route selection using Softmax regression models and training set and recognizes the determining optimized parameter of training, to determine optimized parameter pair
The Softmax regression models answered.
According to another aspect of the present invention, it provides one kind and determining that one-phase earthing failure in electric distribution network selects based on Softmax
The system of line, which is characterized in that the system comprises:
Fault data acquiring unit, for obtaining the event for each presetting node in power distribution network in multiple default nodes of circuit
Hinder data, wherein the fault data includes:Three-phase voltage data, three-phase current data, the phase data of voltage, the phase of electric current
The absolute time information that position information and failure occur;
Database unit obtains fault signature data, and according to described for handling the fault data
Fault signature data, fault type and abort situation mark establish failure identification and fault signature database, wherein the failure
Characteristic includes multiple characteristic quantities;
Model determination unit, for dividing the data in the failure identification and fault signature database according to fault type
For training set and test set, and Softmax regression models and training set is utilized to carry out route selection identification instruction according to default iterations
Practice and determine optimized parameter, to determine the corresponding Softmax regression models of optimized parameter;
Failure line selection determination unit is used for the corresponding Softmax regression models of the optimized parameter according to the reality of power grid
When fault data fault type and position are predicted, to determine power distribution network one-way earth fault route selection.
Preferably, wherein acquiring the fault data at each default node using recording type fault detector.
Preferably, wherein the fault data acquiring unit, handles the fault data, fault signature number is obtained
According to, including:
The fault data is handled, synthesis obtains failure volume, wherein the failure volume includes:Frequency
Variable quantity, the variable quantity of voltage, the change rate of frequency, the change rate of voltage, the change rate of electric current, phase angle change rate, have
The change rate of power, the change rate of inactivity, frequency is distorted with change rate, the current harmonics of power, voltage harmonic is distorted, power
The each harmonic component of factor and voltage and current;
The failure volume is handled according to rough set theory, the low failure volume of the degree of correlation is got rid of, obtains
Take fault signature data.
Preferably, wherein the model determination unit, Softmax regression models and training are utilized according to default iterations
Collection carries out route selection identification training and determines optimized parameter, to determine the corresponding Softmax regression models of optimized parameter, including:
Determine that the cost function of Softmax regression algorithms, cost function are:
Wherein, the sample of Softmax regression models comes from k class, shares m, then the training set being made of these samples
For { (x(1),y(1)),(x(2),y(2)),…,(x(m),y(m)), whereinLabel:y(t)∈ { 1,2 ..., k }, for giving
Fixed input x, it is assumed that function estimates probability value p (y=j | x) for each classification j, for estimating each classification of x
As a result the probability occurred;
Route selection identification training is carried out using gradient descent method to the cost function according to default iterations, is determined optimal
Parameter obtains the corresponding Softmax regression models of optimized parameter.
Preferably, wherein the system also includes:Accuracy validation unit,
For the fault signature data using the corresponding Softmax regression models of the optimized parameter to sample in test set
Corresponding fault type and abort situation are predicted;
For calculating prediction accuracy, and judge whether prediction accuracy is more than or equal to default accuracy threshold value;Wherein,
If prediction accuracy is more than or equal to default accuracy threshold value, directly utilizes and have determined that optimized parameter is corresponding
Softmax regression models determine Single-phase Earth-fault Selection in Distribution Systems;
If prediction accuracy is less than default accuracy threshold value, increase iterations according to iteration step length is preset, and return
Model determination unit carries out route selection using Softmax regression models and training set and recognizes the determining optimized parameter of training, to determine most
The corresponding Softmax regression models of excellent parameter.
The present invention provides a kind of method and system determining Single-phase Earth-fault Selection in Distribution Systems based on Softmax, obtain
Take the fault data of circuit in power distribution network;The fault data is handled, fault signature data are obtained and establishes failure mark
Know and fault signature database;Data in the failure identification and fault signature database are divided into training according to fault type
Collection and test set, and carry out route selection using Softmax regression models and training set according to default iterations and recognize training determination
The corresponding Softmax regression models of optimized parameter;Finally, using the corresponding Softmax regression models of the optimized parameter according to
The real time fail data of power grid predict fault type and position, to determine power distribution network one-way earth fault circuit.This hair
It is bright to utilize truthful data, by processing extraction to data and the relevant characteristic quantity of abort situation, Softmax regression models are drawn
Enter single-phase earthing data identification field, maintenance data is trained and verifies to Softmax regression models, has hewed out with item event
Barrier identification new route, improves data-handling capacity and route selection accuracy, set for new distribution terminal protection algorithm, threshold value,
The verification of action logic strategy provides new approaches, promotes the research and development and application of new distribution terminal.
Description of the drawings
By reference to the following drawings, exemplary embodiments of the present invention can be more fully understood by:
Fig. 1 is the method for determining Single-phase Earth-fault Selection in Distribution Systems based on Softmax according to embodiment of the present invention
100 flow chart;
Fig. 2 is the rough set model figure according to embodiment of the present invention;
Fig. 3 is the power distribution network typical overhead gauze frame structure chart according to embodiment of the present invention;
Fig. 4 is 500 obtained data separating point diagrams of iteration according to embodiment of the present invention;
Fig. 5 is 1000 obtained data separating point diagrams of iteration according to embodiment of the present invention;And
Fig. 6 is to be based on what Softmax determined Single-phase Earth-fault Selection in Distribution Systems according to the school of embodiment of the present invention
The structural schematic diagram of system 600.
Specific implementation mode
Exemplary embodiments of the present invention are introduced referring now to the drawings, however, the present invention can use many different shapes
Formula is implemented, and is not limited to the embodiment described herein, and to provide these embodiments be to disclose at large and fully
The present invention, and fully convey the scope of the present invention to person of ordinary skill in the field.Show for what is be illustrated in the accompanying drawings
Term in example property embodiment is not limitation of the invention.In the accompanying drawings, identical cells/elements use identical attached
Icon is remembered.
Unless otherwise indicated, term (including scientific and technical terminology) used herein has person of ordinary skill in the field
It is common to understand meaning.Further it will be understood that with the term that usually used dictionary limits, should be understood as and its
The context of related field has consistent meaning, and is not construed as Utopian or too formal meaning.
Fig. 1 is the method for determining Single-phase Earth-fault Selection in Distribution Systems based on Softmax according to embodiment of the present invention
100 flow chart.As shown in Figure 1, embodiments of the present invention offer determines one-phase earthing failure in electric distribution network based on Softmax
The method of route selection utilizes truthful data, and by processing extraction to data and the relevant characteristic quantity of abort situation, Softmax is returned
Model is returned to introduce single-phase earthing data identification field, maintenance data is trained and verifies to Softmax regression models, hews out
With a fault identification new route, data-handling capacity and route selection accuracy are improved, is new distribution terminal protection algorithm, threshold
Value setting, the verification of action logic strategy provide new approaches, promote the research and development and application of new distribution terminal.The reality of the present invention
What the mode of applying provided determines the method 100 of Single-phase Earth-fault Selection in Distribution Systems since step 101 place based on Softmax,
Step 101 obtains the fault data for each presetting node in power distribution network in multiple default nodes of circuit, wherein the number of faults
According to including:What three-phase voltage data, three-phase current data, the phase data of voltage, the phase information of electric current and failure occurred
Absolute time information.
Preferably, wherein acquiring the fault data at each default node using recording type fault detector.
Preferably, in step 102, the fault data is handled, obtains fault signature data, and according to the event
Barrier characteristic, fault type and abort situation mark establish failure identification and fault signature database, wherein the failure is special
It includes multiple characteristic quantities to levy data.
Preferably, wherein described handle the fault data, fault signature data are obtained, including:To the event
Barrier data are handled, and synthesis obtains failure volume, wherein the failure volume includes:The variable quantity of frequency, voltage
Variable quantity, the change rate of frequency, the change rate of voltage, the change rate of electric current, the change rate of phase angle, the change rate for having power, nothing
The change rate of power, frequency be distorted with change rate, the current harmonics of power, voltage harmonic distortion, power factor and voltage and
The each harmonic component of electric current;
The failure volume is handled according to rough set theory, the low failure volume of the degree of correlation is got rid of, obtains
Take fault signature data.
In embodiments of the present invention, the fault data of record is handled, the variation delta f of frequency synthesisi, electricity
The variation delta V of pressurei, change rate (the Δ f/ Δs t) of frequencyi, change rate (the Δ V/ Δs t) of voltagei, change rate (the Δ I/ of electric current
Δt)i, the change rate of phase angleThere is change rate (the Δ P/ Δs t) of poweri, change rate (the Δ Q/ Δs t) of inactivityi,
Frequency with power change rate (Δ f/ Δ P)i, current harmonics distortion CTHD, voltage harmonic distortion VTHD, power factorWith
And the quantity of states such as each harmonic component of voltage and current.Then, the low quantity of state of the degree of correlation is got rid of according to rough set theory, remained
Remaining quantity of state, that is, fault signature data.
In embodiments of the present invention, ifX is fault message, and U is fault signature set.RBIt is one of U
Equivalence relation is judged based on existing knowledge, i.e. known fault location information, U/B={ E1,E2,…,EeIndicate
The degree of correspondence of each fault message and abort situation.X is about RBLower aprons collection, upper approximate set is defined respectively as:
Wherein, upper approximate set representations are according to existing knowledge RB, judge the object composition that centainly belongs to or may belong to X in U
Set, lower aprons set representations are according to existing knowledge RB, judge the set for the object composition for belonging to X in U certainly.
Obviously,The lower and upper approximations of X say that domain U is divided into three disjoint regions,
I.e. positive domain POSRB(X), Boundary Region BNDRB(X) and negative domain NEGRB(X), wherein
POSRB(X)=RB (X),
Fig. 2 is the rough set model figure according to embodiment of the present invention.As shown in Fig. 2, intuitively show the positive domain of X,
Boundary Region and negative domain, the characteristic for falling into positive domain is high with the abort situation degree of correlation, and Boundary Region has shown Partial Feature data
It is relevant with abort situation, the characteristic in negative domain is completely uncorrelated to abort situation.It is event with remaining quantity of state
Hinder characteristic, is combined with fault type and fault location information, establish failure identification and fault signature database.And it will be different
The failure of type is divided into two groups, and one group is trained for fault identification algorithm, another group of validity for being used for verification algorithm.
Preferably, in step 103, by the data in the failure identification and fault signature database according to fault type point
For training set and test set, and Softmax regression models and training set is utilized to carry out route selection identification instruction according to default iterations
Practice and determine optimized parameter, to determine the corresponding Softmax regression models of optimized parameter.
Preferably, wherein the default iterations of the basis carry out route selection using Softmax regression models and training set and distinguish
Know training and determines optimized parameter, to determine the corresponding Softmax regression models of optimized parameter, including:
Determine that the cost function of Softmax regression algorithms, cost function are:
Wherein, the sample of Softmax regression models comes from k class, shares m, then the training set being made of these samples
For { (x(1),y(1)),(x(2),y(2)),…,(x(m),y(m)), whereinLabel:y(t)∈ { 1,2 ..., k }, for giving
Fixed input x, it is assumed that function estimates probability value p (y=j | x) for each classification j, for estimating each classification of x
As a result the probability occurred;
Route selection identification training is carried out using gradient descent method to the cost function according to default iterations, is determined optimal
Parameter obtains the corresponding Softmax regression models of optimized parameter.
In embodiments of the present invention, the corresponding Softmax regression models of optimized parameter are determined, including:
Assuming that the sample of Softmax regression models comes from k class, m are shared, then the training set being made of these samples is
{(x(1),y(1)),(x(2),y(2)),…,(x(m),y(m))}.WhereinLabel:y(t)∈{1,2,…,k}.For giving
Fixed input x, it is assumed that function estimates probability value p (y=j | x) for each classification j, for estimating each classification of x
As a result the probability occurred, thus, it is supposed that function will export a k dimensional vector (vector element and be 1) to indicate that this k is estimated
Probability.Assuming that function hθ(x) form is as follows:
WhereinIt is model parameter, a θ is assigned to each characteristic quantity, to characterize this feature
Ability of the amount for failure judgement position.p(y(i)=j | x(i);θj) indicate sample x(i)Belong to the probability of jth class.
It is normalization probability distribution and the sum of to make all probability be 1.In order to facilitate expression, an indicative letter is expressed as using 1 { }
Number, i.e., 1 { true }=1,1 { false }=0.So, the cost function of Softmax regression algorithms may be defined as:
Softmax functions are added up k possible classifications, i.e., x are classified as classification J in Softmax recurrence
Probability be:
Classification wherein corresponding to maximum probability is the class categories of x.
In practical applications, weight decaying is added usually in above-mentioned cost function to solve the parameter of Softmax recurrence
Numerical problem caused by redundancy, then cost function become:
There are this weight attenuation term, cost function to reform into stringent convex function, guarantees to obtain unique solution, have
Effect solves parameter redundancy issue.
Partial derivative is asked to J (θ):
The renewal process of θ can be obtained according to gradient descent method:
Wherein, α is Learning Step, therefore can be obtained:
Go out to approach the θ of optimal solution by gradient descent method iteration, then all θ values generation is returned to and assumes function hθ(x) it in, asks
The probability for taking all station location markers of numerous characteristic quantity set expression under event of failure, obtains trained Softmax and returns mould
Type.
Preferably, in step 104, using the corresponding Softmax regression models of the optimized parameter according to the real-time of power grid
Fault data predicts fault type and position, to determine power distribution network one-way earth fault route selection.
Preferably, wherein the method further includes:
The fault signature data of sample in test set are corresponded to using the optimized parameter corresponding Softmax regression models
Fault type and abort situation predicted;
Prediction accuracy is calculated, and judges whether prediction accuracy is more than or equal to default accuracy threshold value;Wherein,
If prediction accuracy is more than or equal to default accuracy threshold value, directly utilizes and have determined that optimized parameter is corresponding
Softmax regression models determine Single-phase Earth-fault Selection in Distribution Systems;
If prediction accuracy is less than default accuracy threshold value, increase iterations according to iteration step length is preset, and return
Step 3 carries out route selection using Softmax regression models and training set and recognizes the determining optimized parameter of training, to determine optimized parameter pair
The Softmax regression models answered.
In embodiments of the present invention, the sample data in test set is substituted into and assumes that function calculates its generic
Probability, the maximum classification of select probability verify accuracy rate as prediction result.The circuit that will likely be broken down is compiled
Number, it is denoted as 0~n, regression training is carried out to Softmax models using the tranining database for each numbering corresponding fault signature,
It is basic number of iterations with 100 times, obtains the corresponding Softmax regression models of optimized parameter, trained model is utilized into test
Data are tested.If accuracy rate is less than 95%, increases by 100 iterations, continue with Softmax regression models and instruction
Practice collection progress route selection identification training and determine optimized parameter, and determine the corresponding Softmax regression models of optimized parameter, until accurate
Until rate is higher than 95%.
The embodiment illustrated the present invention in detail below
Fig. 3 is the power distribution network typical overhead gauze frame structure chart according to embodiment of the present invention.As shown in figure 3, being one
Typical 10kV distribution network structure structures, neutral grounding mode be directly grounded, fault ground type is metallic earthing.
Each node location shown in the line, i.e., 646,645,632,633,634,611,684,652,671,680,692 and 675 are saved
12 recording type fault detectors of point place placement, collection voltages, electric current, phase, temporal information carry out between fault detector pair
When, the synchronism of retention time record, and setting singlephase earth fault and same on 0~9 circuit shown in The Scarlet Letter in figure 3 respectively
Step acquisition fault data.
It synthesizes to obtain the variation delta f of frequency by collected fault datai, the variation delta V of voltagei, the change of frequency
Rate (Δ f/ Δs t)i, change rate (the Δ V/ Δs t) of voltagei, change rate (the Δ I/ Δs t) of electric currenti, the change rate of phase angleThere is change rate (the Δ P/ Δs t) of poweri, change rate (the Δ Q/ Δs t) of inactivityi, frequency with power change rate
(Δf/ΔP)i, current harmonics distortion CTHD, voltage harmonic distortion VTHD, power factorAnd 3,5 times of voltage and current
Harmonic component.Then, data are handled by rough set theory, determines that fault signature data include:The variable quantity of voltage
ΔVi, change rate (the Δ I/ Δs t) of electric currenti, change rate (the Δ V/ Δs t) of voltageiAnd the change rate of phase angleTotally 4
A higher characteristic quantity of the degree of correlation.
After determining characteristic quantity, cost function is trained using gradient descent method, obtains optimized parameter, determine optimal
The corresponding Softmax regression models of parameter.Then, accuracy is verified using the sample data in test set.
In an embodiment of the present invention, it is 95% to preset accuracy threshold value, and prediction result is divided into 10 classes, corresponds in Fig. 3
Shown respective feed connection, respectively 0~9.It is verified data separating point diagram by 500 iteration as shown in figure 4,1000 iteration
As shown in figure 5, being overlapped as can be seen from Figure, in Fig. 4 more, it is overlapped less in Fig. 5, increasing for iterations can be by different characteristic number
Classify according to corresponding abort situation, overlapping, that is, characteristic is unable to judge accurately abort situation classification, is kept completely separate then
It can accurate judgement.Wherein, the judging nicety rate of Fig. 4 is 97%, is more than default accuracy threshold value 95%, then can directly utilize
Determine that the corresponding Softmax regression models of optimized parameter determine Single-phase Earth-fault Selection in Distribution Systems.
After determining the corresponding Softmax regression models of optimized parameter, returned using the corresponding Softmax of the optimized parameter
Return model to be predicted fault type and position according to the real time fail data of power grid, determines that power distribution network one-way earth fault selects
Line.
Fig. 6 is the system for determining Single-phase Earth-fault Selection in Distribution Systems based on Softmax according to embodiment of the present invention
600 structural schematic diagram.As shown in fig. 6, embodiments of the present invention offer determines power distribution network single-phase earthing based on Softmax
The system 600 of failure line selection includes:Fault data acquiring unit 601, Database unit 602,603 and of model determination unit
Failure line selection determination unit 604.Preferably, in the fault data acquiring unit 601, the multiple pre- of circuit in power distribution network is obtained
If the fault data of node is each preset in node, wherein the fault data includes:Three-phase voltage data, three-phase electricity fluxion
According to, the absolute time information that occurs of the phase data of voltage, the phase information of electric current and failure.
Preferably, wherein acquiring the fault data at each default node using recording type fault detector.
Preferably, in the Database unit 602, the fault data is handled, obtains fault signature number
According to, and failure identification and fault signature database are established according to the fault signature data, fault type and abort situation mark,
The wherein described fault signature data include multiple characteristic quantities.
Preferably, wherein the fault data acquiring unit, handles the fault data, fault signature number is obtained
According to, including:
The fault data is handled, synthesis obtains failure volume, wherein the failure volume includes:Frequency
Variable quantity, the variable quantity of voltage, the change rate of frequency, the change rate of voltage, the change rate of electric current, phase angle change rate, have
The change rate of power, the change rate of inactivity, frequency is distorted with change rate, the current harmonics of power, voltage harmonic is distorted, power
The each harmonic component of factor and voltage and current;
The failure volume is handled according to rough set theory, the low failure volume of the degree of correlation is got rid of, obtains
Take fault signature data.
Preferably, in the model determination unit 603, the failure identification and the data in fault signature database are pressed
Be divided into training set and test set according to fault type, and according to default iterations using Softmax regression models and training set into
Row route selection identification training determines optimized parameter, to determine the corresponding Softmax regression models of optimized parameter.
Preferably, wherein the model determination unit, Softmax regression models and training are utilized according to default iterations
Collection carries out route selection identification training and determines optimized parameter, to determine the corresponding Softmax regression models of optimized parameter, including:
Determine that the cost function of Softmax regression algorithms, cost function are:
Wherein, the sample of Softmax regression models comes from k class, shares m, then the training set being made of these samples
For { (x(1),y(1)),(x(2),y(2)),…,(x(m),y(m)), whereinLabel:y(t)∈ { 1,2 ..., k }, for
Given input x, it is assumed that function estimates probability value p (y=j | x) for each classification j, for estimating each point of x
The probability that class result occurs;
Route selection identification training is carried out using gradient descent method to the cost function according to default iterations, is determined optimal
Parameter obtains the corresponding Softmax regression models of optimized parameter.
Preferably, in the failure line selection determination unit 604, with the corresponding Softmax regression models of the optimized parameter
Fault type and position are predicted according to the real time fail data of power grid, to determine power distribution network one-way earth fault route selection.
Preferably, wherein the system also includes:Accuracy validation unit,
The fault signature data of sample in test set are corresponded to using the optimized parameter corresponding Softmax regression models
Fault type and abort situation predicted;Prediction accuracy is calculated, and it is default to judge whether prediction accuracy is more than or equal to
Accuracy threshold value;Wherein, it if prediction accuracy is more than or equal to default accuracy threshold value, directly utilizes and has determined that optimized parameter pair
The Softmax regression models answered determine Single-phase Earth-fault Selection in Distribution Systems;If prediction accuracy is less than default accuracy threshold value,
Iterations are then increased according to default iteration step length, and returns to model determination unit and utilizes Softmax regression models and training set
It carries out route selection identification training and determines optimized parameter, to determine the corresponding Softmax regression models of optimized parameter.
The system 600 and this hair that Single-phase Earth-fault Selection in Distribution Systems is determined based on Softmax of the embodiment of the present invention
Another bright embodiment determines that the method 100 of Single-phase Earth-fault Selection in Distribution Systems is corresponding based on Softmax, herein not
It repeats again.
The present invention is described by reference to a small amount of embodiment.However, it is known in those skilled in the art, as
Defined by subsidiary Patent right requirement, in addition to the present invention other embodiments disclosed above are equally fallen the present invention's
In range.
Normally, all terms used in the claims are all solved according to them in the common meaning of technical field
It releases, unless in addition clearly being defined wherein.All references " one/described/be somebody's turn to do [device, component etc.] " are all opened ground
It is construed at least one of described device, component etc. example, unless otherwise expressly specified.Any method disclosed herein
Step need not all be run with disclosed accurate sequence, unless explicitly stated otherwise.
Claims (10)
1. a kind of method determining Single-phase Earth-fault Selection in Distribution Systems based on Softmax, which is characterized in that the method packet
It includes:
Step 1, the fault data for each presetting node in power distribution network in multiple default nodes of circuit is obtained, wherein the failure
Data include:Three-phase voltage data, three-phase current data, the phase data of voltage, the phase information of electric current and failure occur
Absolute time information;
Step 2, the fault data is handled, obtains fault signature data, and according to the fault signature data, failure
Type and abort situation mark establish failure identification and fault signature database, wherein the fault signature data include multiple spies
Sign amount;
Step 3, the data in the failure identification and fault signature database are divided into training set and test according to fault type
Collection, and carry out route selection using Softmax regression models and training set according to default iterations and recognize the determining optimized parameter of training,
To determine the corresponding Softmax regression models of optimized parameter;
Step 4, using the corresponding Softmax regression models of the optimized parameter according to the real time fail data of power grid to failure classes
Type and position are predicted, to determine power distribution network one-way earth fault route selection.
2. according to the method described in claim 1, it is characterized in that, acquiring each default node using recording type fault detector
The fault data at place.
3. according to the method described in claim 1, it is characterized in that, described handle the fault data, failure is obtained
Characteristic, including:
The fault data is handled, synthesis obtains failure volume, wherein the failure volume includes:The change of frequency
Change amount, the variable quantity of voltage, the change rate of frequency, the change rate of voltage, the change rate of electric current, phase angle change rate, have power
Change rate, the change rate of inactivity, frequency be distorted with change rate, the current harmonics of power, voltage harmonic distortion, power factor
And each harmonic component of voltage and current;
The failure volume is handled according to rough set theory, gets rid of the low failure volume of the degree of correlation, obtains event
Hinder characteristic.
4. according to the method described in claim 1, it is characterized in that, the basis presets iterations using Softmax recurrence
Model and training set carry out route selection identification training and determine optimized parameter, to determine the corresponding Softmax regression models of optimized parameter,
Including:
Determine that the cost function of Softmax regression algorithms, cost function are:
Wherein, the sample of Softmax regression models comes from k class, shares m, then is { (x by the training set that these samples form(1),y(1)),(x(2),y(2)),…,(x(m),y(m)), whereinLabel:y(t)∈ { 1,2 ..., k }, for what is given
Input x, it is assumed that function estimates probability value p (y=j | x) for each classification j, for estimating each classification results of x
The probability of appearance;
Route selection identification training is carried out using gradient descent method to the cost function according to default iterations, determines optimal ginseng
Number obtains the corresponding Softmax regression models of optimized parameter.
5. according to the method described in claim 1, it is characterized in that, the method further includes:
Utilize the event corresponding to the fault signature data of sample in test set of the corresponding Softmax regression models of the optimized parameter
Barrier type and abort situation are predicted;
Prediction accuracy is calculated, and judges whether prediction accuracy is more than or equal to default accuracy threshold value;Wherein,
If prediction accuracy is more than or equal to default accuracy threshold value, directly utilizes and have determined that the corresponding Softmax of optimized parameter
Regression model determines Single-phase Earth-fault Selection in Distribution Systems;
If prediction accuracy is less than default accuracy threshold value, increase iterations, and return to step 3 according to iteration step length is preset
Route selection is carried out using Softmax regression models and training set and recognizes the determining optimized parameter of training, to determine that optimized parameter is corresponding
Softmax regression models.
6. a kind of system determining Single-phase Earth-fault Selection in Distribution Systems based on Softmax, which is characterized in that the system packet
It includes:
Fault data acquiring unit, for obtaining the number of faults for each presetting node in power distribution network in multiple default nodes of circuit
According to wherein the fault data includes:Three-phase voltage data, three-phase current data, the phase data of voltage, the phase of electric current letter
The absolute time information that breath and failure occur;
Database unit obtains fault signature data, and according to the failure for handling the fault data
Characteristic, fault type and abort situation mark establish failure identification and fault signature database, wherein the fault signature
Data include multiple characteristic quantities;
Model determination unit, for the data in the failure identification and fault signature database to be divided into instruction according to fault type
Practice collection and test set, and true using Softmax regression models and training set progress route selection identification training according to default iterations
Optimized parameter is determined, to determine the corresponding Softmax regression models of optimized parameter;
Failure line selection determination unit, for the real-time event with the corresponding Softmax regression models of the optimized parameter according to power grid
Barrier data predict fault type and position, to determine power distribution network one-way earth fault route selection.
7. system according to claim 6, which is characterized in that acquire each default node using recording type fault detector
The fault data at place.
8. system according to claim 6, which is characterized in that the fault data acquiring unit, to the fault data
It is handled, obtains fault signature data, including:
The fault data is handled, synthesis obtains failure volume, wherein the failure volume includes:The change of frequency
Change amount, the variable quantity of voltage, the change rate of frequency, the change rate of voltage, the change rate of electric current, phase angle change rate, have power
Change rate, the change rate of inactivity, frequency be distorted with change rate, the current harmonics of power, voltage harmonic distortion, power factor
And each harmonic component of voltage and current;
The failure volume is handled according to rough set theory, gets rid of the low failure volume of the degree of correlation, obtains event
Hinder characteristic.
9. system according to claim 6, which is characterized in that the model determination unit, according to default iterations profit
Route selection identification training is carried out with Softmax regression models and training set and determines optimized parameter, to determine that optimized parameter is corresponding
Softmax regression models, including:
Determine that the cost function of Softmax regression algorithms, cost function are:
Wherein, the sample of Softmax regression models comes from k class, shares m, then is { (x by the training set that these samples form(1),y(1)),(x(2),y(2)),…,(x(m),y(m)), whereinLabel:y(t)∈ { 1,2 ..., k }, for what is given
Input x, it is assumed that function estimates probability value p (y=j | x) for each classification j, for estimating each classification results of x
The probability of appearance;
Route selection identification training is carried out using gradient descent method to the cost function according to default iterations, determines optimal ginseng
Number obtains the corresponding Softmax regression models of optimized parameter.
10. system according to claim 6, which is characterized in that the system also includes:Accuracy validation unit,
For being corresponded to the fault signature data of sample in test set using the corresponding Softmax regression models of the optimized parameter
Fault type and abort situation predicted;
For calculating prediction accuracy, and judge whether prediction accuracy is more than or equal to default accuracy threshold value;Wherein,
If prediction accuracy is more than or equal to default accuracy threshold value, directly utilizes and have determined that the corresponding Softmax of optimized parameter
Regression model determines Single-phase Earth-fault Selection in Distribution Systems;
If prediction accuracy is less than default accuracy threshold value, increase iterations according to iteration step length is preset, and return to model
Determination unit carries out route selection using Softmax regression models and training set and recognizes the determining optimized parameter of training, with the optimal ginseng of determination
The corresponding Softmax regression models of number.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201810432276.5A CN108562821B (en) | 2018-05-08 | 2018-05-08 | Method and system for determining single-phase earth fault line selection of power distribution network based on Softmax |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201810432276.5A CN108562821B (en) | 2018-05-08 | 2018-05-08 | Method and system for determining single-phase earth fault line selection of power distribution network based on Softmax |
Publications (2)
Publication Number | Publication Date |
---|---|
CN108562821A true CN108562821A (en) | 2018-09-21 |
CN108562821B CN108562821B (en) | 2021-09-28 |
Family
ID=63538006
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201810432276.5A Active CN108562821B (en) | 2018-05-08 | 2018-05-08 | Method and system for determining single-phase earth fault line selection of power distribution network based on Softmax |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN108562821B (en) |
Cited By (6)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN110659405A (en) * | 2019-09-25 | 2020-01-07 | 南京源堡科技研究院有限公司 | Network information acquisition method based on cloud environment |
CN111160241A (en) * | 2019-12-27 | 2020-05-15 | 华中科技大学 | Power distribution network fault classification method, system and medium based on deep learning |
CN111598166A (en) * | 2020-05-18 | 2020-08-28 | 国网山东省电力公司电力科学研究院 | Single-phase earth fault classification method and system based on principal component analysis and Softmax function |
CN111796161A (en) * | 2020-05-27 | 2020-10-20 | 山西浩然机电设备工程有限公司 | Fault detection system and fault detection method for overhead cable |
CN112924813A (en) * | 2021-01-28 | 2021-06-08 | 国网浙江省电力有限公司绍兴供电公司 | Power distribution network short-circuit fault monitoring method and device based on electrical data |
CN116227538A (en) * | 2023-04-26 | 2023-06-06 | 国网山西省电力公司晋城供电公司 | Clustering and deep learning-based low-current ground fault line selection method and equipment |
Citations (9)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN103760464A (en) * | 2014-01-07 | 2014-04-30 | 河南理工大学 | Small current grounding system fault line selecting method based on analytic graph solving and SVM |
EP2752674A1 (en) * | 2013-01-03 | 2014-07-09 | ABB Technology AG | A detection method of a ground fault in an electric power distribution network |
CN104297635A (en) * | 2014-10-14 | 2015-01-21 | 河南理工大学 | Fault line selection method for distribution network on basis of atom sparse decomposition and extreme learning machine |
CN105759167A (en) * | 2016-01-28 | 2016-07-13 | 江苏省电力公司南京供电公司 | Wavelet neural network-based distribution network single-phase short circuit line selection method |
CN106291234A (en) * | 2016-07-29 | 2017-01-04 | 武汉大学 | A kind of transmission line of electricity internal fault external fault based on convolutional neural networks judges and fault phase-selecting method |
CN106951900A (en) * | 2017-04-13 | 2017-07-14 | 杭州申昊科技股份有限公司 | A kind of automatic identifying method of arrester meter reading |
CN107766879A (en) * | 2017-09-30 | 2018-03-06 | 中国南方电网有限责任公司 | The MLP electric network fault cause diagnosis methods of feature based information extraction |
CN107909118A (en) * | 2017-12-11 | 2018-04-13 | 北京映翰通网络技术股份有限公司 | A kind of power distribution network operating mode recording sorting technique based on deep neural network |
CN107947156A (en) * | 2017-11-24 | 2018-04-20 | 国网辽宁省电力有限公司 | Based on the electric network fault critical clearing time method of discrimination for improving Softmax recurrence |
-
2018
- 2018-05-08 CN CN201810432276.5A patent/CN108562821B/en active Active
Patent Citations (9)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
EP2752674A1 (en) * | 2013-01-03 | 2014-07-09 | ABB Technology AG | A detection method of a ground fault in an electric power distribution network |
CN103760464A (en) * | 2014-01-07 | 2014-04-30 | 河南理工大学 | Small current grounding system fault line selecting method based on analytic graph solving and SVM |
CN104297635A (en) * | 2014-10-14 | 2015-01-21 | 河南理工大学 | Fault line selection method for distribution network on basis of atom sparse decomposition and extreme learning machine |
CN105759167A (en) * | 2016-01-28 | 2016-07-13 | 江苏省电力公司南京供电公司 | Wavelet neural network-based distribution network single-phase short circuit line selection method |
CN106291234A (en) * | 2016-07-29 | 2017-01-04 | 武汉大学 | A kind of transmission line of electricity internal fault external fault based on convolutional neural networks judges and fault phase-selecting method |
CN106951900A (en) * | 2017-04-13 | 2017-07-14 | 杭州申昊科技股份有限公司 | A kind of automatic identifying method of arrester meter reading |
CN107766879A (en) * | 2017-09-30 | 2018-03-06 | 中国南方电网有限责任公司 | The MLP electric network fault cause diagnosis methods of feature based information extraction |
CN107947156A (en) * | 2017-11-24 | 2018-04-20 | 国网辽宁省电力有限公司 | Based on the electric network fault critical clearing time method of discrimination for improving Softmax recurrence |
CN107909118A (en) * | 2017-12-11 | 2018-04-13 | 北京映翰通网络技术股份有限公司 | A kind of power distribution network operating mode recording sorting technique based on deep neural network |
Non-Patent Citations (1)
Title |
---|
林莘 等: "超高压输电线路的潜供电弧频谱特性", 《高电压技术》 * |
Cited By (9)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN110659405A (en) * | 2019-09-25 | 2020-01-07 | 南京源堡科技研究院有限公司 | Network information acquisition method based on cloud environment |
CN111160241A (en) * | 2019-12-27 | 2020-05-15 | 华中科技大学 | Power distribution network fault classification method, system and medium based on deep learning |
CN111160241B (en) * | 2019-12-27 | 2022-08-12 | 华中科技大学 | Power distribution network fault classification method, system and medium based on deep learning |
CN111598166A (en) * | 2020-05-18 | 2020-08-28 | 国网山东省电力公司电力科学研究院 | Single-phase earth fault classification method and system based on principal component analysis and Softmax function |
CN111598166B (en) * | 2020-05-18 | 2023-10-17 | 国网山东省电力公司电力科学研究院 | Single-phase earth fault classification method and system based on principal component analysis and Softmax function |
CN111796161A (en) * | 2020-05-27 | 2020-10-20 | 山西浩然机电设备工程有限公司 | Fault detection system and fault detection method for overhead cable |
CN111796161B (en) * | 2020-05-27 | 2023-06-13 | 山西浩然机电设备工程有限公司 | Fault detection system and fault detection method for overhead cable |
CN112924813A (en) * | 2021-01-28 | 2021-06-08 | 国网浙江省电力有限公司绍兴供电公司 | Power distribution network short-circuit fault monitoring method and device based on electrical data |
CN116227538A (en) * | 2023-04-26 | 2023-06-06 | 国网山西省电力公司晋城供电公司 | Clustering and deep learning-based low-current ground fault line selection method and equipment |
Also Published As
Publication number | Publication date |
---|---|
CN108562821B (en) | 2021-09-28 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN108562821A (en) | A kind of method and system determining Single-phase Earth-fault Selection in Distribution Systems based on Softmax | |
CN110082640B (en) | Distribution network single-phase earth fault identification method based on long-time memory network | |
Gracia et al. | Best ANN structures for fault location in single-and double-circuit transmission lines | |
Jiao et al. | A new method to improve fault location accuracy in transmission line based on fuzzy multi-sensor data fusion | |
CN101493491B (en) | Substation ground network defect synthesis diagnosis method and diagnosis system thereof | |
CN101477172B (en) | Analogue circuit fault diagnosis method based on neural network | |
CN106777984A (en) | A kind of method that photovoltaic array Working state analysis and fault diagnosis are realized based on density clustering algorithm | |
CN109635928A (en) | A kind of voltage sag reason recognition methods based on deep learning Model Fusion | |
CN105676085A (en) | Extra-high voltage GIS partial discharge detection method based on multi-sensor information fusion | |
CN106127047B (en) | A kind of electric system malicious data detection method based on Jensen-Shannon distance | |
CN104578053B (en) | Power system transient stability Forecasting Methodology based on disturbed voltage trace bunch feature | |
Chaitanya et al. | An intelligent fault detection and classification scheme for distribution lines integrated with distributed generators | |
Kim et al. | PMU-based event localization technique for wide-area power system | |
CN104407273B (en) | Electric energy quality disturbance source positioning method considering monitoring reliability | |
CN103439631A (en) | Method and system for detecting corrosion state of grounding grid | |
CN104599193A (en) | Rule base based single-phase ground fault positioning method of distribution network | |
CN105574669A (en) | Space-time union data clustering analysis based abnormal status detection method for power transmission and transformation device | |
CN109828168A (en) | Converter method for diagnosing faults based on Density Estimator | |
CN107462810A (en) | A kind of fault section location method suitable for active power distribution network | |
CN107121617A (en) | A kind of direct current transmission line fault distance-finding method of use filter branches electric current and K k-nearest neighbors | |
Aboshady et al. | Fault detection and classification scheme for PV system using array power and cross-strings differential currents | |
CN104332984B (en) | A kind of node voltage based on noise like influences each other the on-line identification method of the factor | |
CN109270383A (en) | A kind of non-intrusion type charging pile automatic testing method | |
CN106646138B (en) | Distribution net work earthing fault localization method based on the conversion of more sample frequency wavelet character energy | |
CN106646106A (en) | Power grid fault detection method based on change point detection technology |
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 |