CN109085468A - A kind of recognition methods of cable local discharge insulation defect - Google Patents
A kind of recognition methods of cable local discharge insulation defect Download PDFInfo
- Publication number
- CN109085468A CN109085468A CN201810846572.XA CN201810846572A CN109085468A CN 109085468 A CN109085468 A CN 109085468A CN 201810846572 A CN201810846572 A CN 201810846572A CN 109085468 A CN109085468 A CN 109085468A
- Authority
- CN
- China
- Prior art keywords
- insulation defect
- local discharge
- recognition methods
- classifier
- cable local
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
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
- G01R31/12—Testing dielectric strength or breakdown voltage ; Testing or monitoring effectiveness or level of insulation, e.g. of a cable or of an apparatus, for example using partial discharge measurements; Electrostatic testing
- G01R31/1227—Testing dielectric strength or breakdown voltage ; Testing or monitoring effectiveness or level of insulation, e.g. of a cable or of an apparatus, for example using partial discharge measurements; Electrostatic testing of components, parts or materials
- G01R31/1263—Testing dielectric strength or breakdown voltage ; Testing or monitoring effectiveness or level of insulation, e.g. of a cable or of an apparatus, for example using partial discharge measurements; Electrostatic testing of components, parts or materials of solid or fluid materials, e.g. insulation films, bulk material; of semiconductors or LV electronic components or parts; of cable, line or wire insulation
- G01R31/1272—Testing dielectric strength or breakdown voltage ; Testing or monitoring effectiveness or level of insulation, e.g. of a cable or of an apparatus, for example using partial discharge measurements; Electrostatic testing of components, parts or materials of solid or fluid materials, e.g. insulation films, bulk material; of semiconductors or LV electronic components or parts; of cable, line or wire insulation of cable, line or wire insulation, e.g. using partial discharge measurements
Landscapes
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Testing Relating To Insulation (AREA)
Abstract
The invention discloses a kind of recognition methods of cable local discharge insulation defect, comprising steps of (1) obtains several insulation defect model;(2) apply voltage to insulation defect model, to acquire local discharge signal, formedSignal graph, whereinOperating frequency phase is characterized, Q characterizes discharge capacity, n characterizationPlane is divided into the shelf depreciation number occurred in each minizone in several minizones;(3) use two dimension Littlewood-Paley experience wavelet transformation willSignal graph decomposes, and obtains experience wavelet coefficient subgraph;(4) Tamura feature, moment characteristics and the entropy feature for extracting experience wavelet coefficient subgraph, obtain characteristic vector space;(5) dimension-reduction treatment is carried out to characteristic vector space, to select validity feature parameter;(6) validity feature parameter input classifier is trained and is tested;(7) local discharge signal to be identified is inputted by the classifier of training and test, classifier exports recognition result.
Description
Technical field
The present invention relates to a kind of recognition methods more particularly to a kind of identification sides that the defect for cable is identified
Method.
Background technique
Power cable line is component part important in urban distribution network, and safe and reliable operation has urban distribution network
It is significant.But under the influence of rugged environment and itself local defect, it may occur that cable insulation aging is tight
Weight, and then cause route, equipment fault.Shelf depreciation is to reflect the important indicator of power cable insulation performance, different defect institutes
The local discharge signal feature of generation is different, by analyzing it, can effectively judge the insulation ag(e)ing of power cable
Situation prevents failure further expansion.
Therefore, the type and feature for accurately grasping the insulation defect of cable, have the safe and reliable operation of power equipment
Important meaning.However, for the current prior art, the type and feature of to master cable insulation defect, it is necessary first to
Pattern-recognition is carried out to the local discharge signal of acquisition, and how to extract effective characteristic parameter in the process becomes research
Emphasis and difficult point.Feature extraction common at present includes temporal analysis and statistical analysis method, wherein temporal analysis
Result it is more serious by electromagnetic interference influence, discrimination is lower.And statistical analysis method may be because of the discharge time originals such as less
Because there is invalid information, and when sample number is less, discrimination is lower.
Based on this, it is expected that obtaining a kind of recognition methods for cable insulation defect, overcome above-mentioned of the existing technology
Defect effectively extracts the characteristic parameter of local discharge signal, to identify the insulation ag(e)ing situation of cable.
Summary of the invention
The object of the present invention is to provide a kind of recognition methods of cable local discharge insulation defect, can effectively identify
The characteristic parameter for effectively extracting local discharge signal, to identify the insulation ag(e)ing situation of cable, and then realizes splendid electricity
The insulation defect recognition effect of cable shelf depreciation improves the detection system intelligence water detected to cable local discharge
It is flat.
Based on above-mentioned purpose, the invention proposes a kind of recognition methods of cable local discharge insulation defect comprising step
It is rapid:
(1) the several insulation defect model of cable is obtained;
(2) apply voltage to various insulation defect models, to acquire its local discharge signal, formed respectiveSignal graph, whereinOperating frequency phase is characterized, Q characterizes discharge capacity, n characterizationIt is small that plane is divided into several
The shelf depreciation number occurred in each minizone in section;
(3) use two dimension Littlewood-Paley experience wavelet transformation willSignal graph decomposes, and obtains eachThe experience wavelet coefficient subgraph of signal graph;
(4) Tamura feature, moment characteristics and the entropy feature for extracting experience wavelet coefficient subgraph, obtain characteristic vector space;
(5) dimension-reduction treatment is carried out to characteristic vector space, to choose several validity features ginseng out of characteristic vector space
Amount;
(6) several validity feature parameters input classifier is trained and is tested;
(7) local discharge signal to be identified is inputted by the classifier of training and test, classifier exports identification knot
Fruit.
In technical solutions according to the invention, each insulation defect model is being establishedAfter signal graph, benefit
It is right with two-dimentional Littlewood-Paley experience wavelet transformation (hereinafter referred to as 2D-LPEWT)Signal graph is filtered
Wave.2D-LPEWT is extension of the one-dimensional experience wavelet transformation on two-dimensional surface, direct by construction orthogonal wavelet filter group
The natural mode of vibration for extracting original signal, does not need successive ignition, and then reduce the influence of modal overlap.Therefore, 2D- is utilized
Each insulation defect model may be implemented in the wavelet coefficient of LPEWTEffective decomposition of signal graph, obtained experience are small
The effective information in 2D-LPEWT can farthest be retained in wave system number subgraph.
In order to preferably illustrate the technical solution of this case, shift step of 2D-LPEWT is illustrated at this, transformation step
It is rapid as described below:
It willSignal graph is inputted as input picture, and maximum band number N is arranged, in some embodiments,
4 are set by maximum band number N;
The pseudo- pole of input picture is subjected to Fast Fourier Transform, and frequency spectrum is homogenized by following formula, is obtained
Wherein,The frequency spectrum of input signal after indicating equalization, ω indicate frequency, NθIndicate the big of discrete angular
It is small, FpIt (f) is the two-dimension fourier transform of pseudo- pole, f is frequency domain symbol, and i is metering symbol, θiRepresent i-th of angle;
To the frequency spectrum of the input signal after equalizationIt carries out Fourier border detection and obtains the set on Fourier boundary
Ω, whereinN indicates that metering symbol, N indicate that maximum band number, ω indicate frequency, and according to base
Respective filter group B is established in the Fourier transform of Littlewood-Paley small echoεlρ, that is, wavelet frame is established,
The Fourier transform of Littlewood-Paley small echo is given by:
If n ≠ N-1,
If n=N-1,
Wherein, N is maximum band number, and n is metering symbol, n=1 ..., N-1.F2(φ1) (ω) expression experience scale letter
Number, F2(ψn) (ω) expression experience wave function, as n=N-1, F2(ψn) (ω)=F2(ψN-1)(ω).The discussion restricted area of ω
Between be [0, π], and arrange ω0=0, ωN=π.γ indicates constant parameter, ensures that two continuous transitional regions are not overlapped.β
It is an arbitrary Ck([0,1]) function, meets characteristic:
Establishing filter group BεlρAfterwards, input picture is filtered by 2D-LPEWT, i.e., using following formula to defeated
Enter image to be filtered:
Wherein,Indicate the detail coefficients of 2D-LPEWT,Indicate approximation coefficient, F2(f) (ω) table
Show two-dimension fourier transform,Indicate two dimensional inverse fourier transform,The complex conjugate of expression experience wave function,The complex conjugate of expression experience scaling function;
By finally obtaining filter group BεlρAnd detail coefficientsFinally construct the small of 2D-LPEWT
Wave frame and detail coefficients.
Hereby it is achieved that cable local discharge insulation defectSignal graph effectively decomposes, and is obtained by decomposing
The experience wavelet coefficient subgraph obtained can effectively extract feature such as Tamura feature, moment characteristics and entropy feature, to construct
Obtain characteristic vector space.Inventor is selected out of characteristic vector space by carrying out dimension-reduction treatment to characteristic vector space
Take several validity feature parameters, calculated with Simplified analysis, at the same can to avoid due to multicollinearity existing between feature and
Cause great error.Several validity feature parameters input classifier is trained and is tested, finally by office to be identified
Discharge signal input in portion's is by the classifier of training and test, classifier exports recognition result, to realize to cable insulation
The recognition effect of defect.
Further, in the recognition methods of cable local discharge insulation defect of the present invention, the insulation defect
Model includes at least insulative air gap discharging model, high-voltage corona discharge model, suspension electrode discharging model and surface-discharge mould
Type.
Further, in the recognition methods of cable local discharge insulation defect of the present invention, the Tamara is special
Sign textural characteristics include roughness, contrast, directionality, line similarity, rule degree and rough degree.
In the above scheme, roughness can be by calculating the best scale S in entire imagebestAverage value come
It arrives, following formula can be passed through:
Wherein, FcrsIndicate roughness, m indicates that the primitive number under image abscissa, n indicate the primitive under image ordinate
Number, i indicate that i-th of primitive in abscissa m, j indicate i-th of primitive in ordinate n.
Contrast is led to by being counted to the pixel intensity distribution situation in whole image or region
Following formula is crossed to be calculated:
Wherein, FconIndicate contrast, α4=μ4/σ4, μ4Indicate four squares, σ2It is variance.
Directionality indicates the gradient vector at each pixel, i.e., both horizontally and vertically on variable quantity, figure can be passed through
It calculates as convolution, is obtained by following formula:
| Δ G |=(| ΔH|+|ΔV|)/2
θ=arctan (ΔV/ΔH)+π/2
Wherein, | Δ G | it is gradient vector, θ is angle vector, ΔHAnd ΔVRespectively both horizontally and vertically on variation
Amount.
Line similarity is used to the texel for describing to be made of line.Direction co-occurrence matrix element PDd(i, j) is defined as
Occurs the relative frequency of two adjacent cells along edge direction spacing distance d on image, line similarity measurements are obtained by following formula
:
Wherein, FlinIndicate line similarity, PDdIndicate n × n local direction co-occurrence matrix.
Rule degree is used to describe the regularity of repeat pattern, by as the roughness of independent characteristic, contrast, directionality,
Measurement of the variation summation of each in line similarity as rule degree is obtained by following formula:
Freg=1-r (σcrs+σcon+σdir+σlin)
Wherein, FregIndicate that rule degree, r indicate normalization factor, σcrs、σcon、σdirAnd σlinRespectively indicate roughness
Deviation, contrast deviation, directivity deviation and line similarity deviation.
Rough degree carrys out approximate calculation as measurement by using roughness and contrast and obtains rough degree, passes through following formula
It obtains:
Frgh=Fcrs+Fcon
Wherein, FrghIndicate rough degree, FconIndicate contrast.
Further, in the recognition methods of cable local discharge insulation defect of the present invention, the entropy feature packet
Include comentropy, approximate entropy and Sample Entropy.
In above scheme, comentropy can be indicated using following formula:
In formula: H (X) indicates comentropy, p (xi) represent certain system X state xi(i=1,2 ..., n) occur probability, 0
≤p(xi)≤1 andInformation entropy is bigger, and the degree of disorder is higher, and valuable information is fewer, and vice versa.
Approximate entropy is for generating the probability of new model, physics sheet in measure time sequence complexity and metric signal
Matter is to measure the logarithm conditional probability mean value that new model occurs in signal sequence when dimension variation.
Complexity and rule of the Sample Entropy using the method for sample statistics come assessment result, for time of measuring sequence data
Rule property.
Further, in the recognition methods of cable local discharge insulation defect of the present invention, the moment characteristics packet
Include Hu not bending moment and Zernike square.
Further, in the recognition methods of cable local discharge insulation defect of the present invention, in step (5),
Dimensionality reduction is carried out to characteristic vector space using factor analysis.
Further, in the recognition methods of cable local discharge insulation defect of the present invention, the classifier is
K-nearest neighbor classifier, decision tree classifier or support vector machine classifier.
Further, in the recognition methods of cable local discharge insulation defect of the present invention, the classifier is
Support vector machine classifier.
Further, in the recognition methods of cable local discharge insulation defect of the present invention, the step (4)
It include: that the maximum two groups of S of its experience wavelet coefficient subgraph standard deviation are chosen for each insulation defect modeli1And Si2, right
Si1And Si2Extract respectively: (a) 6 Tamara textural characteristics are defined as Ti1...Ti6And Ti7...Ti12;(b) Hu not bending moment and
Zernike square, is defined as Hi1、Zi1And Hi2、Zi2;And (c) 3 entropy features, it is defined as Ei1...Ei3And Ei4...Ei6, amount to
22 feature vectors, to constitute described eigenvector space.
Further, in the recognition methods of cable local discharge insulation defect of the present invention, in step (5),
The specific factor variance for calculating all feature vectors in characteristic vector space, to all specific factor variances of acquisition according to big
It is small to be ranked up, it chooses specific factor variance size and comes preceding 13 corresponding feature vectors as validity feature vector.
The recognition methods of cable local discharge insulation defect of the present invention has the following beneficial effects:
The recognition methods of the cable local discharge insulation defect, which can be identified effectively, effectively extracts local discharge signal
Characteristic parameter, to identify the insulation ag(e)ing situation of cable, and then realize the insulation defect of splendid cable local discharge
Recognition effect improves the detection system intelligent level detected to cable local discharge.
Detailed description of the invention
Fig. 1 is the stream of the recognition methods of cable local discharge insulation defect of the present invention in one embodiment
Journey schematic diagram.
Fig. 2 shows the recognition methods of cable local discharge insulation defect of the present invention in one embodiment
Using insulative air gap discharging model it is obtainedSignal graph.
Fig. 3 shows the recognition methods of cable local discharge insulation defect of the present invention in one embodiment
Use high-voltage corona discharge model it is obtainedSignal graph.
Fig. 4 shows the recognition methods of cable local discharge insulation defect of the present invention in one embodiment
Using surface-discharge model it is obtainedSignal graph.
Fig. 5 shows the recognition methods of cable local discharge insulation defect of the present invention in one embodiment
Using suspension electrode discharging model it is obtainedSignal graph.
Fig. 6 shows the recognition methods of cable local discharge insulation defect of the present invention in one embodiment
To show that discharging model is obtainedSignal graph carries out Fourier boundary detected by 2D-LPEWT.
Fig. 7 to Figure 10 respectively illustrates the recognition methods of cable local discharge insulation defect of the present invention in one kind
Under embodiment by taking surface-discharge model as an example when, it is obtained to itsSignal graph obtains after carrying out 2D-LPEWT
Different angle experience wavelet coefficient subgraph, i.e., as N=4,4 detail coefficients being obtained by wavelet transformation.
Specific embodiment
It below will according to specific embodiment and Figure of description is to cable local discharge insulation defect of the present invention
Recognition methods is described further, but the explanation does not constitute the improper restriction to technical solution of the present invention.
Different type can be preferably recognized in order to verify the recognition methods for the cable local discharge insulation defect for using this case
Cable local discharge insulation defect thus have chosen the cable local discharge insulation defect of four seed types, construct this four
The cable local discharge insulation defect model of seed type, and by the cable local discharge insulation defect model of four seed type
Tag is connected into test macro, and slowly boosting is until observe obvious discharge pulse, in pressure process, observation office
It puts instrument to show as a result, when electric discharge phenomena take place and continually and steadily discharge, stops pressure process, record start electric discharge electricity
Pressure;To every kind of cable insulation defect model, 140 data samples are recorded using PD meter, each sample includes continuous 50
The discharge information of a power frequency period.
The manufacturing process of above-mentioned four kinds of cable local discharges insulation defect model is as described below:
Insulative air gap discharging model: before installing silastic terminal pipe, with guarded blade utility knife by the crosslinked polyethylene of direct current cables
(hereinafter referred to as XLPE) insulating surface makes dent (depth 5mm, wide 5mm), forms defect, and when installing terminal connector, lacks
It falls into and does not nearby smear high pressure silicone grease, in order to avoid influence defect effect.
High-voltage corona discharge model: after silastic terminal pipe is installed, the steel needle that a root long degree is 4cm is connect terminal
Head penetrates, and XLPE insulating inner is inserted into, and contact test product direct current cables wire stylet, thus simulated high-pressure corona discharge.
Suspension electrode discharging model: before installing silastic terminal pipe, it is 3cm that one piece of area is reserved in XLPE insulation2
Interior semi-conductive layer, independently of the interior semi-conductive layer of cable body, to simulate suspended discharge.
Surface-discharge model: before silastic terminal pipe is installed, interior semi-conductive layer end is fabricated to equilateral triangle shape
Standing shape, side length 1cm, so that template surface is discharged.
It should be pointed out that above-mentioned cable local discharge insulation defect is only the schematic solution of recognition methods work to this case
Explanation is released, the improper restriction to this case technical solution is not constituted, those skilled in the art can be according to embodiment
Concrete condition selects several cable local discharge insulation defect, and to the several cable local discharge insulation defect into
Row identification, and it is not limited to four kinds of above-mentioned cable local discharge insulation defects.
Then cable local discharge insulation defect is known using the recognition methods of cable local discharge insulation defect
Not, which can refer to Fig. 1, and Fig. 1 is the recognition methods of cable local discharge insulation defect of the present invention one
Flow diagram under kind embodiment.
As shown in Figure 1, in present embodiment, recognition methods comprising steps of
Step 100: obtaining the several insulation defect model of cable;
Step 200: applying voltage to various insulation defect models, to acquire its local discharge signal, formed respectiveSignal graph, whereinOperating frequency phase is characterized, Q characterizes discharge capacity, n characterizationIt is small that plane is divided into several
The shelf depreciation number occurred in each minizone in section;
Step 300: using two dimension Littlewood-Paley experience wavelet transformation willSignal graph decomposes, and obtains
To eachThe experience wavelet coefficient subgraph of signal graph;
Step 400: extracting Tamura feature, moment characteristics and the entropy feature of experience wavelet coefficient subgraph, obtain feature vector
Space;
Step 500: dimension-reduction treatment being carried out to characteristic vector space, to choose several out of characteristic vector space effectively
Characteristic parameter;
Step 600: several validity feature parameters input classifier is trained and is tested;
Step 700: local discharge signal to be identified is inputted by the classifier of training and test, classifier is exported
Recognition result.
It should be pointed out that in step 400, Tamara feature texture feature include roughness, contrast, directionality,
Line similarity, rule degree and rough degree;Entropy feature includes comentropy, approximate entropy and Sample Entropy;Moment characteristics include Hu not bending moment and
Zernike square chooses the maximum two groups of S of its experience wavelet coefficient subgraph standard deviation for each insulation defect modeli1And
Si2, to Si1And Si2Extract respectively: (a) 6 Tamara textural characteristics are defined as Ti1...Ti6And Ti7...Ti12;(b) Hu is constant
Square and Zernike square, are defined as Hi1、Zi1And Hi2、Zi2;And (c) 3 entropy features, it is defined as Ei1...Ei3With
Ei4...Ei6, amount to 22 feature vectors, to constitute described eigenvector space.
In the present embodiment, in step 500, dimensionality reduction is carried out to characteristic vector space using factor analysis.
In addition, in step 500, when taking several validity feature parameters, calculate in characteristic vector space all features to
The specific factor variance of amount is ranked up all specific factor variances of acquisition according to size, and it is big to choose specific factor variance
It is small to come preceding 13 corresponding feature vectors as validity feature vector
It is obtained by step 200Signal graph concrete condition can be referring to figs. 2 to Fig. 5, wherein Fig. 2 is shown
The use insulative air gap of the recognition methods of cable local discharge insulation defect of the present invention in one embodiment discharges
Model is obtainedSignal graph.Fig. 3 shows the recognition methods of cable local discharge insulation defect of the present invention
Use high-voltage corona discharge model in one embodiment is obtainedSignal graph.Fig. 4 shows the present invention
The use surface-discharge model of the recognition methods of the cable local discharge insulation defect in one embodiment obtains
'sSignal graph.Fig. 5 shows the recognition methods of cable local discharge insulation defect of the present invention in a kind of reality
It applies obtained using suspension electrode discharging model under modeSignal graph.
In order to preferably illustrate the 2D-LPEWT in present embodiment, shift step of 2D-LPEWT is said at this
Bright, shift step is as described below:
It willSignal graph is inputted as input picture, and maximum band number N is arranged, in the present embodiment, will
Maximum band number N is set as 4;
The pseudo- pole of input picture is subjected to Fast Fourier Transform, and frequency spectrum is homogenized by following formula, is obtained
Wherein,The frequency spectrum of input signal after indicating equalization, ω indicate frequency, NθIndicate the big of discrete angular
It is small, FpIt (f) is the two-dimension fourier transform of pseudo- pole, f is frequency domain symbol, and i is metering symbol, θiRepresent i-th of angle;
To the frequency spectrum of the input signal after equalizationIt carries out Fourier border detection and obtains the set on Fourier boundary
Ω, whereinN indicates that metering symbol, N indicate that maximum band number, ω indicate frequency, and according to base
Respective filter group B is established in the Fourier transform of Littlewood-Paley small echoεlρ, that is, wavelet frame is established,
The Fourier transform of Littlewood-Paley small echo is given by:
If n ≠ N-1,
If n=N-1,
Wherein, N is maximum band number, and n is metering symbol, n=1 ..., N-1.F2(φ1) (ω) expression experience scale letter
Number, F2(ψn) (ω) expression experience wave function, as n=N-1, F2(ψn) (ω)=F2(ψN-1)(ω).The discussion restricted area of ω
Between be [0, π], and arrange ω0=0, ωN=π.γ indicates constant parameter, ensures that two continuous transitional regions are not overlapped.β
It is an arbitrary Ck([0,1]) function, meets characteristic:
Establishing filter group BεlρAfterwards, input picture is filtered by 2D-LPEWT, i.e., using following formula to defeated
Enter image to be filtered:
Wherein,Indicate the detail coefficients of 2D-LPEWT,Indicate approximation coefficient, F2(f) (ω) table
Show two-dimension fourier transform,Indicate two dimensional inverse fourier transform,The complex conjugate of expression experience wave function,The complex conjugate of expression experience scaling function;
By finally obtaining filter group BεlρAnd detail coefficientsFinally construct the small of 2D-LPEWT
Wave frame and detail coefficients.
In order to make it easy to understand, by taking surface-discharge defect model as an example, it will be shown in Fig. 4Signal graph carries out 2D-
LPEWT, detected Fourier boundary is as shown in fig. 6, and 2D-LPEWT experience wavelet coefficient subgraph such as Fig. 7 obtained
To shown in Figure 10.
Fig. 6 shows the recognition methods of cable local discharge insulation defect of the present invention in one embodiment
To show that discharging model is obtainedSignal graph carries out Fourier boundary detected by 2D-LPEWT.
Fig. 7 to Figure 10 respectively illustrates the recognition methods of cable local discharge insulation defect of the present invention in one kind
Under embodiment by taking surface-discharge model as an example when, it is obtained to itsSignal graph obtains after carrying out 2D-LPEWT
Different angle experience wavelet coefficient subgraph, i.e., as N=4,4 detail coefficients being obtained by wavelet transformation.
Can be seen that in conjunction with Fig. 4 and Fig. 7 to Figure 10 the experience wavelet coefficient subgraph that is obtained after 2D-LPEWT from
Different angle refines original image spectrum, and the independent event for including in each experience wavelet coefficient subgraph is portrayed, such as is put
Electric phase, impulse magnitude with it is consistent shown in Fig. 4, and each experience wavelet coefficient subgraph scale is more accurate.It is possible thereby to see
Out, the experience wavelet coefficient subgraph that the 2D-LPEWT processing carried out using the recognition methods of this case is obtained is remaining original image
Effective information on the basis of, more conducively extraction TuPu method.
It is obtained in view of each cable local discharge defect modelSignal graph can pass through 2D-LPEWT
Processing obtains four groups of experience wavelet coefficient subgraphs, if carrying out feature extraction, handles that data volume is larger, and recognition speed is slower,
It is thus preferred that can be got the bid using each cable local discharge defect model experience wavelet coefficient subgraph obtained is chosen
The quasi- maximum two groups of subgraphs of difference carry out Tamura feature, moment characteristics and entropy feature, to obtain characteristic vector space, experience is small
The calculating of the standard deviation of wave system number subgraph can use after image is converted grayscale image by matlab and seek pixel and mean value, into
And it is acquired using standard deviation formula.
By taking some cable local discharge insulation defect model as an example, experience wavelet coefficient subgraph standard obtained
The maximum two groups of subgraphs of difference are denoted as S respectivelyi1And Si2, Si1And Si2Tamara textural characteristics T is extracted respectivelyi1To Ti12, moment characteristics
Hi1、Hi2、Zi1、Zi2And entropy feature Ei1To Ei6, wherein Ti1To Ti6It is corresponding in turn to Si1Roughness, contrast, directionality,
Line similarity, rule degree and rough degree, Ti7To Ti12It is corresponding in turn to Si2Roughness, contrast, directionality, line similarity, rule
Then degree and rough degree.And Hi1And Zi1It is corresponding to indicate Si1Hu not bending moment and Zernike square, Hi2And Zi2It is corresponding to indicate
Si2Hu not bending moment and Zernike square.Ei1To Ei3Correspondence successively indicates Si1Comentropy, approximate entropy and Sample Entropy, Ei4Extremely
Ei6Correspondence successively indicates Si2Comentropy, approximate entropy and Sample Entropy.Above-mentioned to amount to 22 feature vectors, constitutive characteristic vector is empty
Between.
Table 1 lists subgraph Si1And Si2Each feature vector and each feature vector corresponding to pass through Factor minute
The specific factor variance (being indicated with SVar) of analysis method estimation.
Table 1.
It calculates for simplifying the analysis, while can be great to avoid being caused due to multicollinearity existing between feature
Error is ranked up, using spy using factor analysis to dimension-reduction treatment is carried out according to the specific factor variance size of estimation
Different factor variance size come first 13 corresponding feature vectors (i.e. feature vector serial number 8,20,17,19,22,10,3,2,
6,14,5,4 and 7 feature vector) it is identified, best recognition result can be reached.Therefore, above-mentioned first 13 are taken
Corresponding feature vector obtains different cable local discharge insulation defect moulds as validity feature vector, using factor analysis
The characteristic value of selected validity feature vector, is listed in table 2 in type.
Table 2.
Feature vector serial number | A | B | C | D |
8 | -0.323 | -0.199 | 0.982 | -0.29 |
20 | 0.066 | 0.745 | -0.198 | -0.143 |
17 | 0.544 | -0.095 | -0.892 | -0.53 |
19 | 0.04 | 0.072 | 0.427 | 0.046 |
22 | 1.605 | 0.77 | 0.769 | 0.994 |
10 | 0.408 | 0.379 | 0.492 | 0.383 |
3 | 0.531 | 0.294 | 0.052 | 0.358 |
2 | 1.397 | 0.836 | 0.305 | 0.983 |
6 | 1.108 | 0.86 | 0.493 | 0.545 |
14 | 1.207 | 0.762 | 0.283 | 1.078 |
5 | 0.507 | 0.079 | 7.686 | 0.353 |
4 | 0.297 | 0.182 | 0.61 | 0.616 |
7 | 2.316 | 1.622 | 0.775 | 1.623 |
Note: the A in table 2 indicates the characteristic value of the validity feature vector of insulative air gap discharging model;B indicates high-voltage corona
The characteristic value of the validity feature vector of discharging model;C indicates the characteristic value of the validity feature vector of surface-discharge model;D is indicated
The characteristic value of the validity feature vector of suspension electrode discharging model.
The validity feature parameter input classifier finally chosen is trained and tests, and shelf depreciation to be identified is believed
Number input by training and test classifier in, classifier export recognition result.Wherein, classifier is classified using k-nearest neighbor
Device (hereinafter referred to as KNN classifier), decision tree classifier or support vector machine classifier (hereinafter referred to as SVM classifier).
In addition, in order to verify the recognition accuracy for the recognition methods for using this case, by insulative air gap discharging model, high pressure
Corona discharge model, the collected local discharge signal of surface-discharge model and suspension electrode discharging model institute are using original
Signal graph directly extract feature and through general wavelet transformation (2 Dimensions wavelet transform, hereinafter referred to as
Feature is extracted after 2DWT), is subsequently input and is trained and tests in classifier, and by the recognition result of final output and this case
The recognition result exported compares.
When test, the number of training of every kind of cable local discharge insulation defect model is 140,70,28,14 respectively
Under the conditions of, it is put into remaining all samples as test sample, it is as shown in table 3 to obtain recognition result.
Table 3.
As can be seen from Table 3, under different number of training, different classifier performances has very big difference.KNN classification
Although device discrimination be not up to it is very high, number of training on its recognition result influence it is smaller, it is as a result more stable;And it determines
Plan Tree Classifier reaches identical discrimination to the more demanding of training samples number, needs more to train number;SVM points
Class device is then more suitable for higher-dimension, nonlinear pattern classification, and when using the recognition methods of this case, discrimination is more than
80%.
In addition, by table 3 can with it is further seen that, when the timing of number of training one, what different feature extracting methods obtained
Signal, there are notable differences for the recognition effect of classifier.When the recognition methods for extracting feature using original signal figure is identified
When, discrimination only up to reach 82.14%;When the recognition methods for extracting feature using 2DWT is identified, although in KNN
Higher discrimination is reached under classifier, but has been not particularly suited for other two kinds of classifiers the effect is unsatisfactory;And use this case
Recognition methods when being identified, due to farthest remaining the effective information in signal graph, even in training sample
When only 14,89.29% discrimination also can achieve, and use the accuracy of identification of SVM classifier significantly under this method
Increase, discrimination can reach 94.64%, hence it is evident that higher than the discrimination of other two kinds of recognition methods, therefore have good
Practicability.
It can be seen that cable local discharge insulation defect of the present invention in conjunction with table 1 to table 3 and Fig. 1 to Figure 10
Recognition methods can effectively identify the characteristic parameter for effectively extracting local discharge signal, to identify the insulation ag(e)ing shape of cable
Condition, and then the insulation defect recognition effect of splendid cable local discharge is realized, it improves and cable local discharge is examined
The detection system intelligent level of survey.
It should be noted that in protection scope of the present invention prior art part be not limited to present specification to
Embodiment out, all prior arts not contradicted with the solution of the present invention, including but not limited to first patent document,
First public publication, formerly openly use etc., it can all be included in protection scope of the present invention.
In addition, it should also be noted that, institute in the combination of each technical characteristic and unlimited this case claim in this case
Combination documented by the combination or specific embodiment of record, all technical characteristics documented by this case can be with
Any mode is freely combined or is combined, unless generating contradiction between each other.
The above embodiment is a preferred embodiment of the present invention, but embodiments of the present invention are not by above-described embodiment
Limitation, it is other it is any without departing from the spirit and principles of the present invention made by change, modification, substitution, combination, letter
Change, should be equivalent substitute mode, be included within the scope of the present invention.
Claims (10)
1. a kind of recognition methods of cable local discharge insulation defect, which is characterized in that comprising steps of
(1) the several insulation defect model of cable is obtained;
(2) apply voltage to various insulation defect models, to acquire its local discharge signal, formed respectiveSignal
Figure, whereinOperating frequency phase is characterized, Q characterizes discharge capacity, n characterizationPlane is divided into each of several minizones
The shelf depreciation number occurred in minizone;
(3) use two dimension Littlewood-Paley experience wavelet transformation willSignal graph decomposes, and obtains each
The experience wavelet coefficient subgraph of signal graph;
(4) Tamura feature, moment characteristics and the entropy feature for extracting experience wavelet coefficient subgraph, obtain characteristic vector space;
(5) dimension-reduction treatment is carried out to characteristic vector space, to choose several validity feature parameters out of characteristic vector space;
(6) several validity feature parameters input classifier is trained and is tested;
(7) local discharge signal to be identified is inputted by the classifier of training and test, classifier exports recognition result.
2. the recognition methods of cable local discharge insulation defect as described in claim 1, which is characterized in that the insulation defect
Model includes at least insulative air gap discharging model, high-voltage corona discharge model, suspension electrode discharging model and surface-discharge model.
3. the recognition methods of cable local discharge insulation defect as described in claim 1, which is characterized in that the Tamara is special
Sign textural characteristics include roughness, contrast, directionality, line similarity, rule degree and rough degree.
4. the recognition methods of cable local discharge insulation defect as claimed in claim 3, which is characterized in that the entropy feature packet
Include comentropy, approximate entropy and Sample Entropy.
5. the recognition methods of cable local discharge insulation defect as claimed in claim 4, which is characterized in that the moment characteristics packet
Include Hu not bending moment and Zernike square.
6. the recognition methods of cable local discharge insulation defect as described in claim 1, which is characterized in that in step (5),
Dimensionality reduction is carried out to characteristic vector space using factor analysis.
7. the recognition methods of cable local discharge insulation defect as described in claim 1, which is characterized in that the classifier is
K-nearest neighbor classifier, decision tree classifier or support vector machine classifier.
8. the recognition methods of cable local discharge insulation defect as claimed in claim 7, which is characterized in that the classifier is
Support vector machine classifier.
9. the recognition methods of cable local discharge insulation defect as claimed in claim 5, which is characterized in that the step (4)
It include: that the maximum two groups of S of its experience wavelet coefficient subgraph standard deviation are chosen for each insulation defect modeli1And Si2, right
Si1And Si2Extract respectively: (a) 6 Tamara textural characteristics are defined as Ti1...Ti6And Ti7...Ti12;(b) Hu not bending moment and
Zernike square, is defined as Hi1、Zi1And Hi2、Zi2;And (c) 3 entropy features, it is defined as Ei1...Ei3And Ei4...Ei6, amount to
22 feature vectors, to constitute described eigenvector space.
10. the recognition methods of cable local discharge insulation defect as claimed in claim 9, which is characterized in that in step (5)
In, calculate the specific factor variance of all feature vectors in characteristic vector space, to all specific factor variances of acquisition according to
Size is ranked up, and is chosen specific factor variance size and is come preceding 13 corresponding feature vectors as validity feature vector.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201810846572.XA CN109085468A (en) | 2018-07-27 | 2018-07-27 | A kind of recognition methods of cable local discharge insulation defect |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201810846572.XA CN109085468A (en) | 2018-07-27 | 2018-07-27 | A kind of recognition methods of cable local discharge insulation defect |
Publications (1)
Publication Number | Publication Date |
---|---|
CN109085468A true CN109085468A (en) | 2018-12-25 |
Family
ID=64831217
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201810846572.XA Pending CN109085468A (en) | 2018-07-27 | 2018-07-27 | A kind of recognition methods of cable local discharge insulation defect |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN109085468A (en) |
Cited By (10)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN110632479A (en) * | 2019-11-03 | 2019-12-31 | 西南交通大学 | EPR cable insulation performance evaluation method based on discharge capacity normalized transformation |
CN110646717A (en) * | 2019-11-03 | 2020-01-03 | 西南交通大学 | Insulation aging simulation and evaluation method for ethylene propylene rubber cable |
CN110766264A (en) * | 2019-07-31 | 2020-02-07 | 贵州电网有限责任公司 | Transmission cable health state monitoring and evaluating method |
CN111209713A (en) * | 2020-01-03 | 2020-05-29 | 长江存储科技有限责任公司 | Wafer data processing method and device |
CN111707915A (en) * | 2020-07-10 | 2020-09-25 | 中车唐山机车车辆有限公司 | Online insulation evaluation method and device based on partial discharge of traction motor |
CN112578244A (en) * | 2020-12-08 | 2021-03-30 | 广西电网有限责任公司电力科学研究院 | Method for evaluating internal defect discharge of GIS (gas insulated switchgear) by utilizing ultraviolet light |
CN112730984A (en) * | 2021-03-31 | 2021-04-30 | 国网江西省电力有限公司供电服务管理中心 | Low-voltage distribution network phase identification method based on intelligent electric meter |
CN115546533A (en) * | 2022-08-31 | 2022-12-30 | 日照东方电机有限公司 | Method and system for insulating stator winding of permanent magnet motor |
CN116502051A (en) * | 2023-06-26 | 2023-07-28 | 广东电网有限责任公司珠海供电局 | Submarine cable local defect identification method and device |
CN117557569A (en) * | 2024-01-12 | 2024-02-13 | 吉林交通职业技术学院 | Road pavement construction quality detection method and system |
Citations (11)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
JP3989786B2 (en) * | 2002-07-24 | 2007-10-10 | 株式会社メガチップス | Compression encoding apparatus, compression encoding method and program |
CN101650404A (en) * | 2009-09-18 | 2010-02-17 | 国网电力科学研究院武汉南瑞有限责任公司 | Method for detecting and identifying constructional deficiencies of high-voltage and supervoltage cables and cable sets |
CN201477169U (en) * | 2009-08-28 | 2010-05-19 | 上海市电力公司电缆输配电公司 | Insulation defect detection device for high voltage cable accessory |
CN102004211A (en) * | 2009-08-28 | 2011-04-06 | 上海市电力公司电缆输配电公司 | Method for detecting insulation defect of high-voltage cable accessory |
CN103091610A (en) * | 2013-01-11 | 2013-05-08 | 辽宁省电力有限公司朝阳供电公司 | Method for transformer fault graphic early warning |
CN103698674A (en) * | 2014-01-02 | 2014-04-02 | 国家电网公司 | Flow cable partial discharge online monitoring system |
CN104502820A (en) * | 2014-12-25 | 2015-04-08 | 国家电网公司 | Two-dimensional wavelet transformation based transformer partial discharge spectrum denoising method |
CN104808107A (en) * | 2015-04-16 | 2015-07-29 | 国家电网公司 | XLPE cable partial discharge defect type identification method |
CN107861029A (en) * | 2017-09-27 | 2018-03-30 | 国网山东省电力公司莱芜供电公司 | A kind of shelf depreciation stage recognition methods based on radar spectrogram |
CN108120907A (en) * | 2018-01-10 | 2018-06-05 | 清华大学 | The partial discharge diagnostic method of feature extraction under a kind of tremendously low frequency voltage based on power frequency |
CN109102508A (en) * | 2018-08-31 | 2018-12-28 | 上海交通大学 | The recognition methods of ac cable insulation defect based on shelf depreciation characteristics of image |
-
2018
- 2018-07-27 CN CN201810846572.XA patent/CN109085468A/en active Pending
Patent Citations (11)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
JP3989786B2 (en) * | 2002-07-24 | 2007-10-10 | 株式会社メガチップス | Compression encoding apparatus, compression encoding method and program |
CN201477169U (en) * | 2009-08-28 | 2010-05-19 | 上海市电力公司电缆输配电公司 | Insulation defect detection device for high voltage cable accessory |
CN102004211A (en) * | 2009-08-28 | 2011-04-06 | 上海市电力公司电缆输配电公司 | Method for detecting insulation defect of high-voltage cable accessory |
CN101650404A (en) * | 2009-09-18 | 2010-02-17 | 国网电力科学研究院武汉南瑞有限责任公司 | Method for detecting and identifying constructional deficiencies of high-voltage and supervoltage cables and cable sets |
CN103091610A (en) * | 2013-01-11 | 2013-05-08 | 辽宁省电力有限公司朝阳供电公司 | Method for transformer fault graphic early warning |
CN103698674A (en) * | 2014-01-02 | 2014-04-02 | 国家电网公司 | Flow cable partial discharge online monitoring system |
CN104502820A (en) * | 2014-12-25 | 2015-04-08 | 国家电网公司 | Two-dimensional wavelet transformation based transformer partial discharge spectrum denoising method |
CN104808107A (en) * | 2015-04-16 | 2015-07-29 | 国家电网公司 | XLPE cable partial discharge defect type identification method |
CN107861029A (en) * | 2017-09-27 | 2018-03-30 | 国网山东省电力公司莱芜供电公司 | A kind of shelf depreciation stage recognition methods based on radar spectrogram |
CN108120907A (en) * | 2018-01-10 | 2018-06-05 | 清华大学 | The partial discharge diagnostic method of feature extraction under a kind of tremendously low frequency voltage based on power frequency |
CN109102508A (en) * | 2018-08-31 | 2018-12-28 | 上海交通大学 | The recognition methods of ac cable insulation defect based on shelf depreciation characteristics of image |
Non-Patent Citations (1)
Title |
---|
秦雪等: "基于2D-LPEWT 的特征提取方法在电缆局部放电分析中的应用", 《电工技术学报》 * |
Cited By (15)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN110766264A (en) * | 2019-07-31 | 2020-02-07 | 贵州电网有限责任公司 | Transmission cable health state monitoring and evaluating method |
CN110646717A (en) * | 2019-11-03 | 2020-01-03 | 西南交通大学 | Insulation aging simulation and evaluation method for ethylene propylene rubber cable |
CN110632479B (en) * | 2019-11-03 | 2020-09-29 | 西南交通大学 | EPR cable insulation performance evaluation method based on discharge capacity normalized transformation |
CN110632479A (en) * | 2019-11-03 | 2019-12-31 | 西南交通大学 | EPR cable insulation performance evaluation method based on discharge capacity normalized transformation |
CN111209713A (en) * | 2020-01-03 | 2020-05-29 | 长江存储科技有限责任公司 | Wafer data processing method and device |
CN111209713B (en) * | 2020-01-03 | 2023-08-18 | 长江存储科技有限责任公司 | Wafer data processing method and device |
CN111707915B (en) * | 2020-07-10 | 2022-10-14 | 中车唐山机车车辆有限公司 | Online insulation evaluation method and device based on partial discharge of traction motor |
CN111707915A (en) * | 2020-07-10 | 2020-09-25 | 中车唐山机车车辆有限公司 | Online insulation evaluation method and device based on partial discharge of traction motor |
CN112578244A (en) * | 2020-12-08 | 2021-03-30 | 广西电网有限责任公司电力科学研究院 | Method for evaluating internal defect discharge of GIS (gas insulated switchgear) by utilizing ultraviolet light |
CN112730984A (en) * | 2021-03-31 | 2021-04-30 | 国网江西省电力有限公司供电服务管理中心 | Low-voltage distribution network phase identification method based on intelligent electric meter |
CN115546533A (en) * | 2022-08-31 | 2022-12-30 | 日照东方电机有限公司 | Method and system for insulating stator winding of permanent magnet motor |
CN116502051A (en) * | 2023-06-26 | 2023-07-28 | 广东电网有限责任公司珠海供电局 | Submarine cable local defect identification method and device |
CN116502051B (en) * | 2023-06-26 | 2023-09-26 | 广东电网有限责任公司珠海供电局 | Submarine cable local defect identification method and device |
CN117557569A (en) * | 2024-01-12 | 2024-02-13 | 吉林交通职业技术学院 | Road pavement construction quality detection method and system |
CN117557569B (en) * | 2024-01-12 | 2024-04-02 | 吉林交通职业技术学院 | Road pavement construction quality detection method and system |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN109085468A (en) | A kind of recognition methods of cable local discharge insulation defect | |
Guo et al. | Deep-learning-based fault classification using Hilbert–Huang transform and convolutional neural network in power distribution systems | |
CN107451557B (en) | Power transmission line short-circuit fault diagnosis method based on empirical wavelet transform and local energy | |
CN109102508B (en) | Method for identifying insulation defects of alternating current cable based on partial discharge image characteristics | |
CN108957251B (en) | Cable joint partial discharge mode identification method | |
CN109164362B (en) | Method and system for identifying partial discharge defect fault of direct current cable | |
WO2015070513A1 (en) | Pattern recognition method for partial discharge of three-phase in one enclosure type ultrahigh voltage gis | |
CN107358259B (en) | Covering ice for insulator detection method based on GLOH description son and GVF-Snake model | |
Chang et al. | Separation of corona using wavelet packet transform and neural network for detection of partial discharge in gas-insulated substations | |
CN104155585A (en) | GIS partial discharge type identification method based on GK fuzzy clustering | |
Wang et al. | Optimal features selected by NSGA-II for partial discharge pulses separation based on time-frequency representation and matrix decomposition | |
Zaro et al. | Power quality detection and classification using S-transform and rule-based decision tree | |
CN108957261A (en) | A kind of direct current cables typical defect Recognition of Partial Discharge | |
Koley et al. | Wavelet-aided SVM tool for impulse fault identification in transformers | |
CN114237046B (en) | Partial discharge pattern recognition method based on SIFT data feature extraction algorithm and BP neural network model | |
CN111220882A (en) | Method for extracting characteristic quantity of GIS optical partial discharge spectrum | |
CN110161388A (en) | A kind of the fault type recognition method and its system of high-tension apparatus | |
CN109829412A (en) | The Partial Discharge Pattern Recognition Method of fractal characteristic is decomposed based on dynamic mode | |
Bajwa et al. | An investigation into partial discharge pulse extraction methods | |
Zhao et al. | Interpretation of transformer winding deformation fault by the spectral clustering of FRA signature | |
CN115728612A (en) | Transformer discharge fault diagnosis method and device | |
Purkait et al. | Impulse fault classification in transformers by fractal analysis | |
Gao et al. | PD pattern recognition for stator bar models with six kinds of characteristic vectors using BP network | |
CN115079052A (en) | Transformer fault diagnosis method and system | |
Chan et al. | Automatic blind equalization and thresholding for partial discharge measurement in power transformer |
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 | ||
WD01 | Invention patent application deemed withdrawn after publication |
Application publication date: 20181225 |
|
WD01 | Invention patent application deemed withdrawn after publication |