CN113780308A - GIS partial discharge mode identification method and system based on kernel principal component analysis and neural network - Google Patents
GIS partial discharge mode identification method and system based on kernel principal component analysis and neural network Download PDFInfo
- Publication number
- CN113780308A CN113780308A CN202110995882.XA CN202110995882A CN113780308A CN 113780308 A CN113780308 A CN 113780308A CN 202110995882 A CN202110995882 A CN 202110995882A CN 113780308 A CN113780308 A CN 113780308A
- Authority
- CN
- China
- Prior art keywords
- neural network
- partial discharge
- principal component
- component analysis
- gis
- 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
- 238000013528 artificial neural network Methods 0.000 title claims abstract description 134
- 230000036961 partial effect Effects 0.000 title claims abstract description 130
- 238000000513 principal component analysis Methods 0.000 title claims abstract description 65
- 238000000034 method Methods 0.000 title claims abstract description 23
- 239000013598 vector Substances 0.000 claims abstract description 70
- 230000007547 defect Effects 0.000 claims abstract description 31
- 230000009467 reduction Effects 0.000 claims abstract description 30
- 238000012549 training Methods 0.000 claims abstract description 22
- 238000012847 principal component analysis method Methods 0.000 claims abstract description 20
- 230000009466 transformation Effects 0.000 claims abstract description 18
- 102100025142 Beta-microseminoprotein Human genes 0.000 claims abstract 4
- 101000576812 Homo sapiens Beta-microseminoprotein Proteins 0.000 claims abstract 4
- 238000012567 pattern recognition method Methods 0.000 claims description 34
- 210000002569 neuron Anatomy 0.000 claims description 29
- 238000012545 processing Methods 0.000 claims description 16
- 238000003909 pattern recognition Methods 0.000 claims description 14
- 239000002184 metal Substances 0.000 claims description 11
- 239000002923 metal particle Substances 0.000 claims description 9
- 208000028659 discharge Diseases 0.000 description 139
- 230000006870 function Effects 0.000 description 43
- 239000011159 matrix material Substances 0.000 description 13
- 238000012423 maintenance Methods 0.000 description 8
- 238000003745 diagnosis Methods 0.000 description 7
- 230000004913 activation Effects 0.000 description 6
- 238000004364 calculation method Methods 0.000 description 5
- 238000001514 detection method Methods 0.000 description 5
- 238000004422 calculation algorithm Methods 0.000 description 4
- 238000009413 insulation Methods 0.000 description 4
- 238000013507 mapping Methods 0.000 description 4
- 238000005259 measurement Methods 0.000 description 4
- 238000003062 neural network model Methods 0.000 description 4
- 230000009286 beneficial effect Effects 0.000 description 3
- 238000011156 evaluation Methods 0.000 description 3
- 230000008569 process Effects 0.000 description 3
- 230000002829 reductive effect Effects 0.000 description 3
- 239000000725 suspension Substances 0.000 description 3
- 230000015556 catabolic process Effects 0.000 description 2
- 238000010276 construction Methods 0.000 description 2
- 238000000354 decomposition reaction Methods 0.000 description 2
- 238000006731 degradation reaction Methods 0.000 description 2
- 230000004807 localization Effects 0.000 description 2
- 230000007246 mechanism Effects 0.000 description 2
- 238000012544 monitoring process Methods 0.000 description 2
- 238000010606 normalization Methods 0.000 description 2
- 230000010363 phase shift Effects 0.000 description 2
- 238000007619 statistical method Methods 0.000 description 2
- NAWXUBYGYWOOIX-SFHVURJKSA-N (2s)-2-[[4-[2-(2,4-diaminoquinazolin-6-yl)ethyl]benzoyl]amino]-4-methylidenepentanedioic acid Chemical compound C1=CC2=NC(N)=NC(N)=C2C=C1CCC1=CC=C(C(=O)N[C@@H](CC(=C)C(O)=O)C(O)=O)C=C1 NAWXUBYGYWOOIX-SFHVURJKSA-N 0.000 description 1
- 230000009471 action Effects 0.000 description 1
- 238000013473 artificial intelligence Methods 0.000 description 1
- 210000004556 brain Anatomy 0.000 description 1
- 230000003925 brain function Effects 0.000 description 1
- 230000008094 contradictory effect Effects 0.000 description 1
- 238000013461 design Methods 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 230000000694 effects Effects 0.000 description 1
- 238000005538 encapsulation Methods 0.000 description 1
- 238000000605 extraction Methods 0.000 description 1
- 238000001914 filtration Methods 0.000 description 1
- 238000011478 gradient descent method Methods 0.000 description 1
- 230000005484 gravity Effects 0.000 description 1
- 230000010365 information processing Effects 0.000 description 1
- 238000007689 inspection Methods 0.000 description 1
- 230000000670 limiting effect Effects 0.000 description 1
- 238000012886 linear function Methods 0.000 description 1
- 239000000463 material Substances 0.000 description 1
- 238000013178 mathematical model Methods 0.000 description 1
- 239000011859 microparticle Substances 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 238000001208 nuclear magnetic resonance pulse sequence Methods 0.000 description 1
- 230000002265 prevention Effects 0.000 description 1
- 238000011112 process operation Methods 0.000 description 1
- 230000002441 reversible effect Effects 0.000 description 1
- 230000000630 rising effect Effects 0.000 description 1
- 238000013024 troubleshooting Methods 0.000 description 1
- 230000016776 visual perception Effects 0.000 description 1
Images
Classifications
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01R—MEASURING ELECTRIC VARIABLES; MEASURING MAGNETIC VARIABLES
- G01R31/00—Arrangements for testing electric properties; Arrangements for locating electric faults; Arrangements for electrical testing characterised by what is being tested not provided for elsewhere
- G01R31/12—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
- G01R31/1254—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 gas-insulated power appliances or vacuum gaps
-
- 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
- G01R31/327—Testing of circuit interrupters, switches or circuit-breakers
- G01R31/3271—Testing of circuit interrupters, switches or circuit-breakers of high voltage or medium voltage devices
- G01R31/3275—Fault detection or status indication
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
- G06F18/213—Feature extraction, e.g. by transforming the feature space; Summarisation; Mappings, e.g. subspace methods
- G06F18/2135—Feature extraction, e.g. by transforming the feature space; Summarisation; Mappings, e.g. subspace methods based on approximation criteria, e.g. principal component analysis
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
- G06N3/084—Backpropagation, e.g. using gradient descent
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Data Mining & Analysis (AREA)
- General Engineering & Computer Science (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Life Sciences & Earth Sciences (AREA)
- Artificial Intelligence (AREA)
- Evolutionary Computation (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Evolutionary Biology (AREA)
- Bioinformatics & Computational Biology (AREA)
- Health & Medical Sciences (AREA)
- Biomedical Technology (AREA)
- Biophysics (AREA)
- Computational Linguistics (AREA)
- General Health & Medical Sciences (AREA)
- Molecular Biology (AREA)
- Computing Systems (AREA)
- Mathematical Physics (AREA)
- Software Systems (AREA)
- Testing Relating To Insulation (AREA)
Abstract
The invention discloses a GIS partial discharge mode identification method based on kernel principal component analysis and a neural network, which comprises the following steps: (1) collecting partial discharge signals of typical defects in GIS equipment by using an ultrahigh frequency sensor; (2) after the amplitude of the partial discharge signal is normalized, drawing a three-dimensional PRPS map of the partial discharge signal; (3) carrying out Gabor transformation on the three-dimensional PRPS map to obtain a transformed subgraph, and extracting texture feature vectors and shape feature vectors of the transformed subgraph to form original feature vectors; (4) performing dimensionality reduction on the original feature vector by adopting a kernel principal component analysis method to obtain a feature vector training set; (5) constructing a multilayer BP neural network, and training the multilayer BP neural network by using a feature vector training set; (6) and (3) identifying GIS partial discharge signals acquired and actually detected by the ultrahigh frequency sensor by adopting a trained multilayer BP neural network, and outputting an identification result. In addition, the invention also discloses a GIS partial discharge mode identification system.
Description
Technical Field
The invention relates to a partial discharge diagnosis method, in particular to a partial discharge pattern recognition method.
Background
Gas Insulated Switchgear (GIS) is an important electrical device in an electric power system, and has the advantages of small floor area, high operation reliability, flexible configuration, convenience in maintenance and the like. GIS can effectively alleviate the consumption of power grid construction to land resource, and with the development of urban power grid construction, the number of GIS transformer substations increases continuously, and its application in electric power system is more and more extensive.
However, due to the totally-enclosed characteristic of the GIS, fault location and maintenance are difficult, maintenance work is complicated, the average power failure maintenance time after an accident is longer than that of conventional equipment, and a large amount of manpower and material resources are required to be input, so that time and labor are wasted. In addition, the encapsulation of various electrical elements also makes GIS troubleshooting often relate to non-fault element, and the power failure scope is great. Therefore, the state evaluation and the maintenance of the GIS equipment are well done, the potential or generated faults of the insulating part are found in time, and the timely fault prevention or treatment is carried out, so that the GIS equipment state evaluation and maintenance method has important significance for the stable operation of a transformer substation and even a power grid.
Partial discharge is the main manifestation form of GIS insulation degradation, and partial discharge detection is one of the current online monitoring modes commonly used for the insulation state of power equipment, and when partial discharge occurs, early warning and action are immediately taken to effectively prevent serious accidents. Different insulation degradation mechanisms can cause different types of discharge forms, and further damage to GIS insulation in different degrees, so that pattern recognition of GIS partial discharge is an important part in GIS state evaluation.
Based on the above, the invention is expected to obtain a GIS partial discharge mode identification method and system based on kernel principal component analysis and neural network, which can perform Gabor transformation on the three-dimensional PRPS map of the partial discharge signal, extract the transformed subgraph characteristic quantity, reduce the dimension by using kernel principal component analysis method, realize the identification of the type of the partial discharge defect of the GIS fault based on multilayer BP neural network, and improve the intelligent level of power system fault diagnosis.
Disclosure of Invention
One of the purposes of the invention is to provide a GIS partial discharge pattern recognition method based on kernel principal component analysis and neural network, which can perform Gabor transformation on a three-dimensional PRPS map of a partial discharge signal, extract transformed subgraph characteristic quantities, reduce dimensions by a kernel principal component analysis method, realize the partial discharge defect type recognition of GIS faults based on multilayer BP neural network, and improve the intelligent level of power system fault diagnosis.
Based on the above purpose, the invention provides a GIS partial discharge pattern recognition method based on kernel principal component analysis and neural network, which comprises the following steps:
(1) collecting partial discharge signals of typical defects in GIS equipment by using an ultrahigh frequency sensor;
(2) after the amplitude of the partial discharge signal is normalized, drawing a three-dimensional PRPS map of the partial discharge signal;
(3) carrying out Gabor transformation on the three-dimensional PRPS map to obtain a transformed subgraph, and extracting texture feature vectors and shape feature vectors of the transformed subgraph to form original feature vectors;
(4) performing dimensionality reduction on the original feature vector by adopting a kernel principal component analysis method to obtain a feature vector training set;
(5) constructing a multilayer BP neural network, and training the multilayer BP neural network by using a feature vector training set;
(6) and (3) identifying GIS partial discharge signals acquired and actually detected by the ultrahigh frequency sensor by adopting a trained multilayer BP neural network, and outputting an identification result.
In the technical scheme of the invention, the ultrahigh frequency sensor can be used for collecting partial discharge signals of typical defects in GIS equipment, the collected partial discharge signals are processed to further draw a three-dimensional map of partial discharge Pulse Sequence (PRPS), then Gabor transformation which is commonly applied to the field of image processing is used for filtering the three-dimensional PRPS map, effective decomposition of different scales and directions is realized by using transformation coefficients, and a transformation subgraph can be obtained; and extracting the characteristic quantity of the transformed subgraph obtained by decomposition to form an original characteristic vector.
Because the GIS partial discharge data feature space dimension is high, the subsequent calculation amount is large, and the recognition rate can also be reduced, the invention uses a Kernel Principal Component Analysis (KPCA) to perform dimension reduction processing on the original feature vector to obtain a feature vector training set.
In the face of mass data and rapidly increased operation and maintenance work requirements, in order to improve the efficiency of discharge fault identification and diagnosis and improve the intelligent level of operation, inspection and maintenance, an artificial intelligence algorithm is needed. Therefore, the invention constructs a multilayer BP neural network, trains the multilayer BP neural network by using a feature vector training set, can identify the actually detected GIS partial discharge signal by adopting the trained multilayer BP neural network, outputs an identification result and realizes the identification of the GIS partial discharge defect type.
In the present invention, in the step (6), the measured partial discharge signal of the GIS device may be collected again by using an ultrahigh frequency sensor. And then, carrying out normalization processing on the amplitude of the actually measured partial discharge signal, and drawing a three-dimensional PRPS (pulse repetition phase shift protection) map of the partial discharge signal.
And (3) after the three-dimensional PRPS map drawn by the actually measured partial discharge signal is subjected to Gabor transformation, obtaining a transformation subgraph, and extracting the characteristic vector of the transformation subgraph. Correspondingly, the obtained feature vector is subjected to dimensionality reduction by adopting a kernel principal component analysis method, the feature vector obtained after dimensionality reduction can be used as input quantity to be input into a trained multilayer BP neural network to obtain output quantity, and a recognition result can be obtained based on the output quantity, so that the recognition of the GIS partial discharge defect is realized.
Further, in the method for identifying the GIS partial discharge mode based on the nuclear principal component analysis and the neural network, the typical defects comprise floating potential discharge, metal needle point discharge and metal particle discharge.
In the above technical solution of the present invention, typical defects inside the GIS device may include floating potential discharge, metal needle point discharge, and metal particle discharge. Aiming at the internal partial discharge defects of 3 typical GIS devices including suspension potential discharge, metal needle point discharge and metal particle discharge, the invention adopts an ultrahigh frequency sensor and uses an ultrahigh frequency detection method to detect the partial discharge signals.
The Ultra High Frequency (UHF) detection method can realize live detection or on-line monitoring under the condition that equipment is not powered off, has stronger anti-interference capability and can effectively inhibit some background noises with lower frequency. When the partial discharge occurs in a small range, a pulse current with duration only in ns level is generated, the pulse current has an extremely steep rising edge, and electromagnetic waves with the frequency of up to several GHz are excited to radiate to the periphery.
For different defects in the GIS equipment, in order to avoid errors caused by different amplitude orders of the discharge signals, in the step (2) of the GIS partial discharge mode identification method, the amplitude of the partial discharge signals needs to be normalized.
In some embodiments, 60 sets of signals may be collected for each typical defect, each set of signals may include 50 power frequency cycles, phase windows 0-360 ° may be windowed, and each cycle may be divided into 100 phase windows; subsequently, an amplitude-phase-period three-dimensional PRPS map of the partial discharge signal can be drawn.
Further, in the method for identifying a GIS partial discharge pattern based on kernel principal component analysis and neural network of the present invention, in step (3), a gaussian function is used as a window function when Gabor transformation is performed.
In the above technical solution of the present invention, in step (3), the Gabor transform is a short-time Fourier transform, and compared with the conventional Fourier transform, the Gabor wavelet has good time-frequency localization characteristics and is very sensitive to the edge of an image, and thus is often applied to texture recognition. According to the Heisenberg inaccurate measurement principle, any measurement accuracy cannot be obtained in both time domain and frequency domain, the lower bound is the minimum area which can be reached by a time domain window, and a Gaussian function is in the limit value.
Therefore, in the above technical solution, selecting a gaussian function as the window function can make the Gabor transform balance the contradiction between the time domain resolution and the frequency domain resolution, thereby obtaining the higher values of the two at the same time.
In step (3), the two-dimensional Gabor transform is performed on the three-dimensional PRPS map obtained by processing, which may include Gabor transforms in three directions of θ being 0 °, 45 ° and 90 °, and each three-dimensional PRPS map may be converted into 3 transformed subgraphs, so as to obtain an exploded view of partial discharge images of defect models in the GIS device.
And (3) extracting Tamura texture features and a gray-gradient co-occurrence matrix of the transformed subgraph by using a statistical method, and describing the texture features of the image to obtain 6 Tamura texture feature vectors of roughness, contrast, directionality, linearity, regularity and roughness. Meanwhile, according to the joint frequency distribution of two pixels in the image, which is further referred to by the gray level co-occurrence matrix, 15 statistics helpful for reflecting the image texture can be obtained: autocorrelation, contrast, correlation, cluster saliency, cluster shadowing, variability, energy, entropy, uniformity, maximum probability, sum of squares, sum-mean, sum-variance, sum-entropy, and variance of differences. Thus, a total of 21 texture feature vectors describing the texture can be obtained for each transformed subgraph.
Aiming at the aspect of shape characteristics, the Hu invariant moment of the partial discharge atlas can be simultaneously obtained to obtain 7 characteristic quantities and 6 Zernike moment characteristic quantities; thus, a total of 13 shape feature vectors describing the shape can be obtained for each transformed subgraph. Each three-dimensional PRPS map can be converted into 3 transformed subgraphs, so that a total of 102 feature quantities can be obtained for each three-dimensional PRPS map of partial discharge signals.
It should be noted that Tamura et al propose the expression of texture features based on the human study of psychology of visual perception of texture. The six components of Tamura texture feature may correspond to six attributes of the texture feature from a psychological perspective, roughness (coarseness), contrast (contrast), directionality (directionality), linearity (linerikeness), regularity (regularity), and roughness (roughness).
Accordingly, a supplementary description can be made of Zernike moments, which are orthogonalization functions based on Zernike polynomials, the set of orthogonal polynomials utilized being 1 perfect orthogonal set within a unit circle. When calculating the Zernike moments for 1 image, the pixel coordinates are mapped into a unit circle with the centroid (also called the center of gravity) of the image as the origin. The Zernike moments are complex moments that are generally characterized by the modes of the Zernike moments to describe the shape of an object. The shape features of 1 target object can be well represented by 1 group of very small Zernike moment feature vectors, the low-order moment feature vectors describe the overall shape of 1 image target, and the high-order moment feature vectors describe the details of the image target. The Zernike moments have the basic idea of being very similar to the fourier transform, and their principle is to spread the waveforms or signals to obtain a perfect orthogonal set. But differs from the fourier transform in that the perfect orthogonal set of Zernike moments is within the unit circle. When the Zernike moments are calculated, the boundary information of the images does not need to be taken into consideration, and when the images with complex shapes and less definite boundary information are faced, the Zernike moments are more suitable to be selected, and the Zernike moments have rotation invariance, which is one of great advantages.
Further, in the GIS partial discharge pattern recognition method based on kernel principal component analysis and neural network according to the present invention, in step (4), a gaussian function is used as a kernel function of the kernel principal component analysis.
Further, in the method for identifying a GIS partial discharge pattern based on kernel principal component analysis and neural network of the present invention, in step (4), when the kernel principal component analysis method is used to perform dimensionality reduction on the original feature vector, the first 11 principal components with the highest contribution rate are selected as the feature vectors after dimensionality reduction.
Further, in the GIS partial discharge pattern recognition method based on kernel principal component analysis and neural network according to the present invention, in step (4), a polynomial kernel function is used as a kernel function of the kernel principal component analysis.
Further, in the method for identifying a GIS partial discharge pattern based on kernel principal component analysis and neural network of the present invention, in step (4), when the kernel principal component analysis method is used to perform dimensionality reduction on the original feature vector, the first 5 principal components with the highest contribution rate are selected as the feature vectors after dimensionality reduction.
Further, in the GIS partial discharge pattern recognition method based on kernel principal component analysis and neural network of the present invention, the multilayer BP neural network has 3 hidden layers.
In the above technical solution of the present invention, in step (5), a 5-layer BP neural network may be constructed, parameters such as the number of neurons in the hidden layer of the BP neural network, an activation function, and the like are determined, and a multi-layer BP neural network is trained using a feature vector training set.
It should be noted that a BP neural network (back propagation neural network) is a multilayer feedforward neural network based on error back propagation, and since the BP neural network successfully solves the problem of adjusting weights of the multilayer feedforward neural network for solving nonlinear continuous functions, many neural network models in practical application of an artificial neural network all adopt the BP neural network and its variation forms, which has become one of the most widely applied neural network models at present.
In the above technical solution of the present invention, the 5-layer BP neural network may include 1 input layer, 3 hidden layers, and 1 output layer, wherein the connection mode of the neurons between each layer may be full connection, and various neurons in a layer are not connected.
Further, in the GIS partial discharge pattern recognition method based on kernel principal component analysis and neural network of the present invention, the multilayer BP neural network has a 5-layer structure, and further has an input layer and an output layer, the connection mode between the neurons in each layer is full connection, and the neurons in the layer are not connected.
Accordingly, another objective of the present invention is to provide a GIS partial discharge pattern recognition system based on kernel principal component analysis and neural network, which can be used to implement the GIS partial discharge pattern recognition method of the present invention, and can effectively recognize the type of the partial discharge defect of the GIS fault, improve the intelligent level of the power system fault diagnosis, and have good application prospects.
Based on the above object, the present invention further provides a GIS partial discharge pattern recognition system based on kernel principal component analysis and neural network, which includes:
the system comprises an ultrahigh frequency sensor, a data acquisition module and a data processing module, wherein the ultrahigh frequency sensor is used for acquiring partial discharge signals of typical defects in the GIS equipment as sample data and acquiring actually measured partial discharge signals of the GIS equipment;
a recognition module performing steps (2) - (6) of the GIS partial discharge pattern recognition method based on kernel principal component analysis and neural network according to any one of claims 1-9.
Compared with the prior art, the GIS partial discharge mode identification method and system based on kernel principal component analysis and neural network have the following advantages and beneficial effects:
the GIS partial discharge mode identification method based on kernel principal component analysis and neural network can perform Gabor transformation on the three-dimensional PRPS map of the partial discharge signal, extract the transformed subgraph characteristic quantity, reduce the dimension by the kernel principal component analysis method, realize the identification of the GIS fault partial discharge defect type based on the multilayer BP neural network, and improve the intelligent level of the power system fault diagnosis.
Correspondingly, the GIS partial discharge pattern recognition system based on kernel principal component analysis and neural network can be used for implementing the GIS partial discharge pattern recognition method, and the GIS partial discharge pattern recognition system based on kernel principal component analysis and neural network has the advantages and beneficial effects.
Drawings
Fig. 1 schematically shows a flowchart of steps of a GIS partial discharge pattern recognition method based on kernel principal component analysis and neural network according to an embodiment of the present invention.
Fig. 2 schematically shows a neuron model of a BP neural network according to an embodiment of the GIS partial discharge pattern recognition method based on kernel principal component analysis and the neural network according to the present invention.
Fig. 3 schematically shows a network structure of a BP neural network according to an embodiment of the GIS partial discharge pattern recognition method based on kernel principal component analysis and neural network according to the present invention.
Fig. 4 schematically shows a flowchart of steps of a GIS partial discharge pattern recognition method based on kernel principal component analysis and neural network according to another embodiment of the present invention.
Fig. 5 schematically shows a neuron model of a BP neural network in another embodiment of the GIS partial discharge pattern recognition method based on kernel principal component analysis and the neural network according to the present invention.
Detailed Description
The method and system for identifying GIS partial discharge patterns based on kernel principal component analysis and neural network according to the present invention will be further explained and explained with reference to the drawings and specific embodiments of the specification, which, however, should not be construed as unduly limiting the technical solutions of the present invention.
Fig. 1 schematically shows a flowchart of steps of a GIS partial discharge pattern recognition method based on kernel principal component analysis and neural network according to an embodiment of the present invention.
As shown in fig. 1, in this embodiment, the method for identifying GIS partial discharge patterns based on kernel principal component analysis and neural network according to the present invention may include the following steps (1) to (6):
(1) and acquiring partial discharge signals of typical defects in the GIS equipment by adopting an ultrahigh frequency sensor.
Referring to fig. 1, in the present embodiment, typical defects inside the GIS device may include: floating potential discharge, metal needle point discharge and metal particle discharge. Aiming at the typical defects in 3 GIS devices, namely, suspension potential discharge, metal needle point discharge and metal particle discharge, the invention adopts an ultrahigh frequency sensor and can acquire partial discharge signals of the typical defects in the three GIS devices by using an ultrahigh frequency detection method.
(2) And after the amplitude of the partial discharge signal is normalized, drawing a three-dimensional PRPS map of the partial discharge signal.
In the step (2) of the present invention, for different types of defects inside the GIS device, in order to avoid errors caused by different magnitude orders of the discharge signals, the amplitude of the local discharge signal needs to be further normalized.
In some embodiments, for 3 typical defects of floating potential discharge, metal needle point discharge and metal particle discharge, 60 sets of signals may be collected for each typical defect, each set of signals may include 50 power frequency periods, and phase windows are opened for 0-360 degrees, and each period may be divided into 100 phase windows; subsequently, an amplitude-phase-period three-dimensional PRPS map of the partial discharge signal can be drawn.
(3) And carrying out Gabor transformation on the three-dimensional PRPS map to obtain a transformed subgraph, and extracting the texture feature vector and the shape feature vector of the transformed subgraph to form an original feature vector.
In the step (3), the Gabor transform is a short-time Fourier transform, and compared with the conventional Fourier transform, the Gabor wavelet has good time-frequency localization characteristics and is very sensitive to the edge of the image, so that the Gabor transform is often applied to texture recognition. According to the Heisenberg inaccurate measurement principle, any measurement accuracy cannot be obtained in both time domain and frequency domain, the lower bound is the minimum area which can be reached by a time domain window, and a Gaussian function is in the limit value. Therefore, in the above technical solution, when Gabor transform is performed, a gaussian function is used as a window function, so that the Gabor transform balances contradictions between time domain and frequency domain resolutions, thereby obtaining higher values of the two.
In step (3), the two-dimensional Gabor transform is performed on the three-dimensional PRPS map obtained by processing, which may include Gabor transforms in three directions of θ being 0 °, 45 ° and 90 °, and each three-dimensional PRPS map may be converted into 3 transformed subgraphs, so as to obtain an exploded view of partial discharge images of defect models in the GIS device.
And (3) extracting Tamura texture features and a gray-gradient co-occurrence matrix of the transformed subgraph by using a statistical method, and describing the texture features of the image to obtain 6 Tamura texture feature vectors of roughness, contrast, directionality, linearity, regularity and roughness. Meanwhile, according to the joint frequency distribution of two pixels in the image, which is further referred to by the gray level co-occurrence matrix, 15 statistics helpful for reflecting the image texture can be obtained: autocorrelation, contrast, correlation, cluster saliency, cluster shadowing, variability, energy, entropy, uniformity, maximum probability, sum of squares, sum-mean, sum-variance, sum-entropy, and variance of differences. Thus, a total of 21 texture feature vectors describing the texture can be obtained for each transformed subgraph.
Aiming at the aspect of shape characteristics, the Hu invariant moment of the partial discharge atlas can be simultaneously obtained to obtain 7 characteristic quantities and 6 Zernike moment characteristic quantities; thus, a total of 13 shape feature vectors describing the shape can be obtained for each transformed subgraph. Each three-dimensional PRPS map can be converted into 3 transformed subgraphs, so that a total of 102 feature quantities can be obtained for each three-dimensional PRPS map of partial discharge signals.
(4) And (3) carrying out dimensionality reduction on the original feature vectors by adopting a kernel principal component analysis method, selecting the first 11 principal components with the highest contribution rate as the feature vectors subjected to dimensionality reduction, and obtaining a feature vector training set.
Because GIS partial discharge data feature space dimension is higher, not only has the possibility of including redundant characteristic vector, influences the recognition effect, still can increase the task volume for follow-up identification process, brings great burden. Therefore, it is necessary to perform dimension reduction on the original feature vector, and in the step (4), the invention performs dimension reduction on the original feature vector by using a kernel principal component analysis method, and the kernel principal component analysis method can realize nonlinear dimension reduction of data and is suitable for processing linear inseparable data sets.
The kernel principal component analysis method can utilize a kernel function (kernel) idea to map original data from a low dimension to a high dimension space in a nonlinear mode, and then perform feature extraction processing after linear data are obtained, so that the purpose of reducing data dimension is achieved on the premise of keeping original data features.
The specific implementation process of the kernel principal component analysis method can be expressed as follows:
for example, for a set of data x1,x2，…，xR∈RrThere is:
a non-linear function phi can be introduced and the set of data is mapped into feature space F to obtain phi (x)1),Ф(x2),…,Ф(xR) And can also satisfy the following formula (2):
thus, the covariance matrix C in the feature space F can be obtained as:
in the above formula (3), j is 1,2, …, and R (R represents the number of samples); phi (x)j) Representing the original data xjMapping in a feature space F; phi (x)j)TThe expression phi (x)j) The transposed matrix of (2).
Assuming that λ is an eigenvalue of the covariance matrix C shown in the above equation (3) and v is an eigenvector of the covariance matrix C, the following equation (4) may exist:
λv＝Cv (4)
where λ ≧ 0, the parameter a is presenti(i ═ 1,2 …, R) is such that the following formula (5) holds:
in the above formula (5), i is 1,2, …, R; phi (x)i) Representing the original data xiMapping in a feature space F; a isiIs a constant coefficient.
The above formula (4) is related to phi (x)k) By inner product, the following can be obtained:
the comprehensive formula can be obtained:
in the above formula (7), k is 1,2, …, R, Φ (x)k) Denotes xkMapping in a feature space F;
the kernel matrix K is defined as dimension R x R, which can be expressed as the following formula (8):
Kij＝(Φ(xi)Φ(xj)) (8)
for phi (x) in the eigenvector VkThe projection on is:
in the above equation (9), Φ (x) represents a mapping of original data x in the feature space F;
the eigenvalues λ representing the kernel matrix KkThe ith element of the corresponding feature vector.
gk(x) Replacing the inner product with a kernel function corresponding to the kth principal component of Φ (x) to obtain:
in the above formula (10), VkThe kth element of the eigenvector V of the covariance matrix representing the mapped data in the eigenspace F; k (x, x)i) Representing a quantity resulting from the calculation.
According to the principal component contribution rate, the single contribution rate w of each principal component can be calculatedi：
In the above formula (11), λiThe ith eigenvalue of the covariance matrix of the sample; lambda [ alpha ]tThe t-th eigenvalue of the sample covariance matrix is represented.
Accordingly, the individual contribution rates of each principal component are accumulated to obtain an accumulated contribution rate.
In step (4) of the GIS partial discharge pattern recognition method of the present invention, a gaussian function or a polynomial kernel function may be used as a kernel function for kernel principal component analysis.
As can be seen from fig. 1, in the present embodiment, the gaussian function is used as a kernel function for kernel principal component analysis, and when the kernel principal component analysis method is used to perform the dimensionality reduction processing on the original feature quantity, the first 11 principal components with the highest contribution rate may be selected as the feature vectors after dimensionality reduction.
Correspondingly, in some other embodiments, a polynomial kernel function may also be used as a kernel function of the kernel principal component analysis, and when the polynomial kernel function is used as the kernel function of the kernel principal component analysis, when the kernel principal component analysis is used to perform the dimension reduction processing on the original feature quantity, the first 5 principal components with the highest contribution rate may be selected as the feature vector after the dimension reduction.
When a polynomial kernel function is used as a kernel function of kernel principal component analysis, the step flow chart of the GIS partial discharge pattern recognition method based on kernel principal component analysis and neural network according to the present invention can refer to fig. 4. The neuron model of the BP neural network employed at this time can be referred to fig. 5.
(5) And constructing a BP neural network with 3 hidden layers, determining parameters of the network, and training the BP neural network by using a feature vector training set.
Neural Networks (NN) refer to a mathematical model established on the basis of simulating human brain structure and some working mechanisms, and the simple computing units are neurons, and the Network formed by connecting the neurons can reflect many basic characteristics of human brain functions, such as parallel, distributed storage and processing, self-organization, self-adaption and self-learning capabilities, and is particularly suitable for processing inaccurate and fuzzy information processing problems needing to consider many factors and conditions simultaneously. Three-layer and above-three-layer neural networks have the approximant capability of self-adaption to any nonlinear function, and are widely applied to the design work of a partial discharge pattern recognition pattern classifier.
Fig. 2 schematically shows a neuron model of a neural network.
As shown in FIG. 2, x1，x2，…，xqIs an input to the neural network; w is a variable weight, which mayComprising w1，w2，…，wq(ii) a b is the bias of the neural network; f is an activation function; the output of the neuron is then a ═ f (wx + b).
Correspondingly, the BP neural network is a multilayer feedforward neural network based on error back propagation, and because the BP neural network successfully solves the problem of multilayer feedforward neural network weight adjustment for solving nonlinear continuous functions, more neural network models in the practical application of the artificial neural network adopt the BP neural network and the variation form thereof, and the BP neural network becomes one of the most widely applied neural network models.
It should be noted that, in the embodiment shown in fig. 1, the BP neural network according to the present invention may be configured as a 5-layer structure, which has 1 input layer, 3 hidden layers and 1 output layer, and the network structure of the BP neural network with the 5-layer structure may refer to fig. 3. Fig. 3 schematically shows a network structure of a BP neural network according to an embodiment of the GIS partial discharge pattern recognition method based on kernel principal component analysis and neural network according to the present invention.
As shown in fig. 3, in the present embodiment, the multi-layer BP neural network has a 5-layer structure, which has 1 input layer, 3 hidden layers, and 1 output layer, and the connection between the neurons in each layer is full connection, and the neurons in the layer are not connected.
Because the feature vector training set after dimensionality reduction has 11 feature vectors, the number of neurons in the input layer of the BP neural network is set to be 11. The more the number of layers of the multilayer BP neural network is, the higher the classification precision is; however, if the number of layers of the BP neural network is too large, not only the network is more complicated, but also the classification accuracy of the network is reduced, so that the number of layers of the hidden layer is set to 3.
In addition, the number of nodes of the hidden layer of the BP neural network is also one of the most important factors influencing the network performance. The less the number of nodes of a hidden layer in the BP neural network is, the less information obtained by the network is, the lower the classification precision is, and the slower the network convergence speed is; the more the number of nodes of the hidden layer in the BP neural network is, the higher the classification precision is. However, it should be noted that too many nodes in the hidden layer may also cause the topology structure of the BP neural network to be too complex, which results in too long learning time, over-training of the network, poor fault tolerance, and increased recognition error. Therefore, the number and the dimension of the feature vectors after the dimension reduction can be selected to be 10 for the number of neurons in the hidden layer of each layer and 1 for the number of neurons in the output layer.
In the present embodiment, the output of the BP neural network having a 5-layer structure is a type of discharge defect, and can represent a metal microparticle discharge, a floating potential discharge, and a metal needle point discharge by 1,2, and 3, respectively.
It should be noted that, a typical BP neural network adopts a gradient descent algorithm, before prediction is performed by using the BP neural network, a multi-layer BP neural network needs to be trained first, and a weight threshold value of the multi-layer BP neural network is adjusted through training so that the BP neural network has associative memory and prediction capabilities. Therefore, in step (5) of the GIS partial discharge pattern recognition method of the present invention, a multi-layer BP neural network needs to be trained using a feature vector training set.
In the present invention, the process of training the multi-layer BP neural network using the feature vector training set may include the following steps:
a) network initialization: setting initial values for connection weight values between the input layer and the hidden layer, connection weight values between the hidden layers and the output layer, and threshold values of each neuron of the hidden layers and the output layer, and giving parameters such as learning rate of the network, training stopping conditions and the like.
b) Hidden layer output calculation: and taking the 11 reduced eigenvectors as input quantities of the BP neural network, and calculating hidden layer output a ═ f (wX + b) according to the input vector X, the connection weight w between the input layer and the hidden layer bias b.
c) Output layer output calculation: and calculating the output of the BP neural network according to the output a of the hidden layer, the connection weight w between the hidden layer and the output layer and the bias b of the output layer. The output of BP neural network can be converted into numbers 1,2 and 3, and the numbers 1,2 and 3 represent metal particle discharge, suspension potential discharge and metal needle point discharge respectively.
d) And (3) error calculation: and calculating the prediction error of the network according to the output and the expected output of the BP neural network, and using the mean square error as an index for measuring the network performance.
e) Updating parameters: according to the error of the network prediction, continuously correcting and updating the parameters of the network by using a BP algorithm according to a steepest gradient descent method: and the weight w and the bias b until the total error of the BP neural network reaches the minimum or meets the expectation. The BP neural network is a reverse-push learning algorithm of a multilayer network, and in a reverse learning stage, if an expected output value cannot be obtained in an output layer, errors are calculated recursively layer by layer, and a weight value and a threshold value of each layer are modified.
It should be noted that, in this embodiment, the parameter selection of the BP neural network may be as shown in table 1 below, where table 1 lists the parameter selection of the BP neural network with 3 hidden layers.
Table 1.
| Type of parameter | Value of parameter |
| Hidden layer 1 activation function f | Hyperbolic tangent |
| Implicit layer 2 activation function f | Hyperbolic tangent |
| Hidden layer 3 activation function f | Hyperbolic tangent |
| Output layer activation function | Hyperbolic tangent |
| Hidden layer1 number of neurons | 10 |
| Hidden layer 2 neuron number | 10 |
| Hidden layer 3 neuron number | 10 |
(6) And (3) identifying GIS partial discharge signals acquired and actually detected by the ultrahigh frequency sensor by adopting a trained multilayer BP neural network, and outputting an identification result.
In the present invention, in the step (6), an ultrahigh frequency sensor may be adopted to collect a measured partial discharge signal of the GIS device. And then, carrying out normalization processing on the amplitude of the actually measured partial discharge signal, and drawing a three-dimensional PRPS (pulse repetition phase shift protection) map of the partial discharge signal.
And obtaining a transformation subgraph after Gabor transformation of a three-dimensional PRPS atlas drawn by the actually measured partial discharge signal, and extracting texture characteristic vectors and shape characteristic vectors of the transformation subgraph to form original characteristic vectors. Correspondingly, the original characteristic vector is subjected to dimensionality reduction by adopting a kernel principal component analysis method for forming the original characteristic vector, the characteristic vector obtained after dimensionality reduction is input into a trained multilayer BP neural network as an input quantity to obtain one of output quantities 1,2 and 3, and the output quantities represent metal particle discharge, floating potential discharge and metal needle point discharge respectively, so that the GIS partial discharge defect is identified.
Fig. 4 schematically shows a flowchart of steps of a GIS partial discharge pattern recognition method based on kernel principal component analysis and neural network according to another embodiment of the present invention.
Fig. 5 schematically shows a neuron model of a BP neural network in another embodiment of the GIS partial discharge pattern recognition method based on kernel principal component analysis and the neural network according to the present invention.
As shown in fig. 4 and 5, in the GIS partial discharge pattern recognition method according to the present invention, in step (4), unlike the above-described embodiments shown in fig. 1 and 3, which use a gaussian function as a kernel function for kernel principal component analysis, the embodiments shown in fig. 4 and 5 use a polynomial kernel function as a kernel function for kernel principal component analysis.
As shown in fig. 4, when a polynomial kernel function is used as a kernel function of the kernel principal component analysis, the flow of steps of the GIS partial discharge pattern recognition method according to the present invention is practically unchanged, and it also needs to go through steps (1) to (6). In this embodiment, when the polynomial kernel function is used as the kernel function for kernel principal component analysis, the first 5 principal components with the highest contribution rate may be selected as the feature vector after dimensionality reduction, and the rest of the process operations are substantially the same as those in fig. 1 to 3, and are not described here again.
As shown in fig. 5, in the present embodiment, when the kernel principal component analysis method is used to perform dimensionality reduction on the original feature quantity, the BP neural network used still has a 5-layer structure, which has 1 input layer, 3 hidden layers, and 1 output layer, the connection mode between the neurons in each layer is full connection, and the neurons in each layer are not connected.
Since the number of the feature vectors after the dimensionality reduction is only 5, the number of the neurons of the input layer of the BP neural network is set to be 5. Correspondingly, the number of the hidden layer neurons of the BP neural network is set to be 10, and the number of the output layer neurons is set to be 1.
Of course, in order to reasonably implement the GIS partial discharge pattern recognition method based on the kernel principal component analysis and the neural network, the invention also provides a GIS partial discharge pattern recognition system based on the kernel principal component analysis and the neural network. The GIS partial discharge pattern recognition system can be used for implementing the GIS partial discharge pattern recognition method.
In the present invention, the GIS partial discharge pattern recognition system of the present invention may include: the system comprises an ultrahigh frequency sensor and an identification module. The ultrahigh frequency sensor can be used for collecting partial discharge signals of typical defects in the GIS equipment as sample data and collecting actually measured partial discharge signals of the GIS equipment.
Accordingly, the identification module according to the present invention may perform steps (2) - (6) of the GIS partial discharge pattern identification method based on kernel principal component analysis and neural network according to the present invention.
In conclusion, the GIS partial discharge mode identification method based on the kernel principal component analysis and the neural network can perform Gabor transformation on the three-dimensional PRPS map of the partial discharge signal, extract the transformed subgraph characteristic quantity, reduce the dimension by the kernel principal component analysis method, realize the identification of the type of the partial discharge defect of the GIS fault based on the multilayer BP neural network, and improve the intelligent level of the fault diagnosis of the power system.
Correspondingly, the GIS partial discharge pattern recognition system based on kernel principal component analysis and neural network can be used for implementing the GIS partial discharge pattern recognition method, and the GIS partial discharge pattern recognition system based on kernel principal component analysis and neural network has the advantages and beneficial effects.
It should be noted that the prior art in the protection scope of the present invention is not limited to the examples given in the present application, and all the prior art which is not inconsistent with the technical scheme of the present invention, including but not limited to the prior patent documents, the prior publications and the like, can be included in the protection scope of the present invention.
In addition, the combination of the features in the present application is not limited to the combination described in the claims of the present application or the combination described in the embodiments, and all the features described in the present application may be freely combined or combined in any manner unless contradictory to each other.
It should also be noted that the above-mentioned embodiments are only specific embodiments of the present invention. It is apparent that the present invention is not limited to the above embodiments and similar changes or modifications can be easily made by those skilled in the art from the disclosure of the present invention and shall fall within the scope of the present invention.
Claims (10)
1. A GIS partial discharge mode identification method based on kernel principal component analysis and neural network is characterized by comprising the following steps:
(1) collecting partial discharge signals of typical defects in GIS equipment by using an ultrahigh frequency sensor;
(2) after the amplitude of the partial discharge signal is normalized, drawing a three-dimensional PRPS map of the partial discharge signal;
(3) carrying out Gabor transformation on the three-dimensional PRPS map to obtain a transformed subgraph, and extracting texture feature vectors and shape feature vectors of the transformed subgraph to form original feature vectors;
(4) performing dimensionality reduction on the original feature vector by adopting a kernel principal component analysis method to obtain a feature vector training set;
(5) constructing a multilayer BP neural network, and training the multilayer BP neural network by using a feature vector training set;
(6) and (3) identifying GIS partial discharge signals acquired and actually detected by the ultrahigh frequency sensor by adopting a trained multilayer BP neural network, and outputting an identification result.
2. The GIS partial discharge pattern recognition method based on kernel principal component analysis and neural network as claimed in claim 1, wherein the typical defects include floating potential discharge, metal needle point discharge and metal particle discharge.
3. The method for GIS partial discharge pattern recognition based on kernel principal component analysis and neural network as claimed in claim 1, wherein in step (3), a Gaussian function is used as the window function when Gabor transform is performed.
4. The GIS partial discharge pattern recognition method based on kernel principal component analysis and neural network as claimed in claim 1, wherein in step (4), a Gaussian function is used as a kernel function of the kernel principal component analysis.
5. The GIS partial discharge pattern recognition method based on kernel principal component analysis and neural network as claimed in claim 4, wherein in step (4), when the kernel principal component analysis method is used to perform dimensionality reduction processing on the original feature vectors, the first 11 principal components with the highest contribution rate are selected as the feature vectors after dimensionality reduction.
6. The method for GIS partial discharge pattern recognition based on kernel principal component analysis and neural network as claimed in claim 1, wherein in step (4), a polynomial kernel function is adopted as the kernel function of the kernel principal component analysis.
7. The GIS partial discharge pattern recognition method based on kernel principal component analysis and neural network as claimed in claim 6, wherein in step (4), when the kernel principal component analysis method is used to perform dimensionality reduction processing on the original feature vectors, the first 5 principal components with the highest contribution rate are selected as the feature vectors after dimensionality reduction.
8. The GIS partial discharge pattern recognition method based on kernel principal component analysis and neural network as claimed in claim 1, wherein the multi-layer BP neural network has 3 hidden layers.
9. The GIS partial discharge pattern recognition method based on kernel principal component analysis and neural network as claimed in claim 8, wherein the multi-layer BP neural network is a 5-layer structure, and further has an input layer and an output layer, the connection mode between neurons in each layer is full connection, and there is no connection between neurons in each layer.
10. A GIS partial discharge pattern recognition system based on kernel principal component analysis and neural network is characterized by comprising:
the system comprises an ultrahigh frequency sensor, a data acquisition module and a data processing module, wherein the ultrahigh frequency sensor is used for acquiring partial discharge signals of typical defects in the GIS equipment as sample data and acquiring actually measured partial discharge signals of the GIS equipment;
a recognition module performing steps (2) - (6) of the GIS partial discharge pattern recognition method based on kernel principal component analysis and neural network according to any one of claims 1-9.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202110995882.XA CN113780308B (en) | 2021-08-27 | GIS partial discharge pattern recognition method and system based on kernel principal component analysis and neural network |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202110995882.XA CN113780308B (en) | 2021-08-27 | GIS partial discharge pattern recognition method and system based on kernel principal component analysis and neural network |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN113780308A true CN113780308A (en) | 2021-12-10 |
| CN113780308B CN113780308B (en) | 2024-09-27 |
Family
ID=
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN117630611A (en) * | 2024-01-22 | 2024-03-01 | 南京卓煊电力科技有限公司 | Full-bandwidth high-frequency partial discharge PRPD spectrogram capturing and generating method and system |
Citations (19)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US7443154B1 (en) * | 2003-10-04 | 2008-10-28 | Seektech, Inc. | Multi-sensor mapping omnidirectional sonde and line locator |
| CN101853304A (en) * | 2010-06-08 | 2010-10-06 | 河海大学 | Remote sensing image retrieval method based on feature selection and semi-supervised learning |
| CN108052975A (en) * | 2017-12-12 | 2018-05-18 | 浙江大学宁波理工学院 | It is a kind of that real-time working condition Forecasting Methodology is run based on the vehicle of core pivot and neutral net |
| CN108919067A (en) * | 2018-05-28 | 2018-11-30 | 黔南民族师范学院 | A kind of recognition methods for GIS partial discharge mode |
| CN108921230A (en) * | 2018-07-25 | 2018-11-30 | 浙江浙能嘉华发电有限公司 | Method for diagnosing faults based on class mean value core pivot element analysis and BP neural network |
| CN108957251A (en) * | 2018-05-18 | 2018-12-07 | 深圳供电局有限公司 | Cable joint partial discharge mode identification method |
| CN109102012A (en) * | 2018-07-30 | 2018-12-28 | 上海交通大学 | A kind of defect identification method and system of local discharge signal |
| CN109116203A (en) * | 2018-10-31 | 2019-01-01 | 红相股份有限公司 | Power equipment partial discharges fault diagnostic method based on convolutional neural networks |
| CN109342909A (en) * | 2018-12-14 | 2019-02-15 | 中国测试技术研究院电子研究所 | A kind of cable accessory Partial Discharge Pattern Recognition Method based on SLLE |
| CN109683066A (en) * | 2018-11-08 | 2019-04-26 | 山东康威通信技术股份有限公司 | Power cable typical defect local discharge signal recognition methods |
| CN109871746A (en) * | 2018-12-30 | 2019-06-11 | 国网辽宁省电力有限公司电力科学研究院 | A kind of mode identification method based on the diagnosis of GIS insulation defect |
| CN109932627A (en) * | 2019-04-30 | 2019-06-25 | 国网山东省电力公司烟台供电公司 | A kind of localization method and system of GIS pressure test shelf depreciation |
| CN110309010A (en) * | 2019-06-05 | 2019-10-08 | 国网江苏省电力有限公司电力科学研究院 | Partial discharge network training method and device for the explanation of power equipment phase |
| CN111220882A (en) * | 2019-12-31 | 2020-06-02 | 国网辽宁省电力有限公司电力科学研究院 | Method for extracting characteristic quantity of GIS optical partial discharge spectrum |
| CN112070104A (en) * | 2020-07-10 | 2020-12-11 | 国网浙江平阳县供电有限责任公司 | Main transformer partial discharge identification method |
| CN112287953A (en) * | 2019-07-24 | 2021-01-29 | 国网山东省电力公司济南供电公司 | Method and system for GIS insulation defect category identification |
| CN112305379A (en) * | 2019-07-24 | 2021-02-02 | 国网山东省电力公司济南供电公司 | Mode identification method and system for GIS insulation defect |
| CN112462137A (en) * | 2020-09-18 | 2021-03-09 | 国网辽宁省电力有限公司电力科学研究院 | Equipment fault feature extraction method based on wavelet packet and Hilbert envelope spectrum analysis |
| CN113065522A (en) * | 2021-04-26 | 2021-07-02 | 哈尔滨理工大学 | Transformer partial discharge type identification method based on deep neural network |
Patent Citations (19)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US7443154B1 (en) * | 2003-10-04 | 2008-10-28 | Seektech, Inc. | Multi-sensor mapping omnidirectional sonde and line locator |
| CN101853304A (en) * | 2010-06-08 | 2010-10-06 | 河海大学 | Remote sensing image retrieval method based on feature selection and semi-supervised learning |
| CN108052975A (en) * | 2017-12-12 | 2018-05-18 | 浙江大学宁波理工学院 | It is a kind of that real-time working condition Forecasting Methodology is run based on the vehicle of core pivot and neutral net |
| CN108957251A (en) * | 2018-05-18 | 2018-12-07 | 深圳供电局有限公司 | Cable joint partial discharge mode identification method |
| CN108919067A (en) * | 2018-05-28 | 2018-11-30 | 黔南民族师范学院 | A kind of recognition methods for GIS partial discharge mode |
| CN108921230A (en) * | 2018-07-25 | 2018-11-30 | 浙江浙能嘉华发电有限公司 | Method for diagnosing faults based on class mean value core pivot element analysis and BP neural network |
| CN109102012A (en) * | 2018-07-30 | 2018-12-28 | 上海交通大学 | A kind of defect identification method and system of local discharge signal |
| CN109116203A (en) * | 2018-10-31 | 2019-01-01 | 红相股份有限公司 | Power equipment partial discharges fault diagnostic method based on convolutional neural networks |
| CN109683066A (en) * | 2018-11-08 | 2019-04-26 | 山东康威通信技术股份有限公司 | Power cable typical defect local discharge signal recognition methods |
| CN109342909A (en) * | 2018-12-14 | 2019-02-15 | 中国测试技术研究院电子研究所 | A kind of cable accessory Partial Discharge Pattern Recognition Method based on SLLE |
| CN109871746A (en) * | 2018-12-30 | 2019-06-11 | 国网辽宁省电力有限公司电力科学研究院 | A kind of mode identification method based on the diagnosis of GIS insulation defect |
| CN109932627A (en) * | 2019-04-30 | 2019-06-25 | 国网山东省电力公司烟台供电公司 | A kind of localization method and system of GIS pressure test shelf depreciation |
| CN110309010A (en) * | 2019-06-05 | 2019-10-08 | 国网江苏省电力有限公司电力科学研究院 | Partial discharge network training method and device for the explanation of power equipment phase |
| CN112287953A (en) * | 2019-07-24 | 2021-01-29 | 国网山东省电力公司济南供电公司 | Method and system for GIS insulation defect category identification |
| CN112305379A (en) * | 2019-07-24 | 2021-02-02 | 国网山东省电力公司济南供电公司 | Mode identification method and system for GIS insulation defect |
| CN111220882A (en) * | 2019-12-31 | 2020-06-02 | 国网辽宁省电力有限公司电力科学研究院 | Method for extracting characteristic quantity of GIS optical partial discharge spectrum |
| CN112070104A (en) * | 2020-07-10 | 2020-12-11 | 国网浙江平阳县供电有限责任公司 | Main transformer partial discharge identification method |
| CN112462137A (en) * | 2020-09-18 | 2021-03-09 | 国网辽宁省电力有限公司电力科学研究院 | Equipment fault feature extraction method based on wavelet packet and Hilbert envelope spectrum analysis |
| CN113065522A (en) * | 2021-04-26 | 2021-07-02 | 哈尔滨理工大学 | Transformer partial discharge type identification method based on deep neural network |
Non-Patent Citations (3)
| Title |
|---|
| NAN HE: "Gas-insulated Switchgear Discharge Pattern Recognition Based on KPCA and Multilayer Neural Network", INTERNATIONAL CONFERENCE ON COMMUNICATION, IMAGE AND SIGNAL PROCESSING, vol. 3, pages 208 - 213 * |
| 李宾宾;张健;柯艳国;田宇;杨为;朱胜龙;: "基于暂态地电压的开关柜局放特征提取及类型识别方法研究", 高压电器, no. 03, pages 43 - 50 * |
| 董晨晔;钱勇;钟勇;许永鹏;盛戈;江秀臣;: "基于EEMD和二维Gabor变换的GIS局放特征融合识别方法", 电气自动化, no. 05, pages 25 - 28 * |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN117630611A (en) * | 2024-01-22 | 2024-03-01 | 南京卓煊电力科技有限公司 | Full-bandwidth high-frequency partial discharge PRPD spectrogram capturing and generating method and system |
| CN117630611B (en) * | 2024-01-22 | 2024-04-12 | 南京卓煊电力科技有限公司 | Full-bandwidth high-frequency partial discharge PRPD spectrogram capturing and generating method and system |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN114755745B (en) | Hail weather identification and classification method based on multi-channel depth residual shrinkage network | |
| Raymond et al. | Partial discharge classifications: Review of recent progress | |
| Sahoo et al. | Trends in partial discharge pattern classification: a survey | |
| CN109884419B (en) | Smart power grid power quality online fault diagnosis method | |
| Liu et al. | Deep learning and recognition of radar jamming based on CNN | |
| Wu et al. | A hybrid support vector regression approach for rainfall forecasting using particle swarm optimization and projection pursuit technology | |
| CN109389171B (en) | Medical image classification method based on multi-granularity convolution noise reduction automatic encoder technology | |
| CN111044570A (en) | Defect identification and early warning method and device for power equipment and computer equipment | |
| CN108919059A (en) | A kind of electric network failure diagnosis method, apparatus, equipment and readable storage medium storing program for executing | |
| CN104299232B (en) | SAR image segmentation method based on self-adaptive window directionlet domain and improved FCM | |
| CN109919241A (en) | Hyperspectral unknown class target detection method based on probability model and deep learning | |
| CN111950406A (en) | Finger vein identification method, device and storage medium | |
| CN107944353A (en) | SAR image change detection based on profile ripple BSPP networks | |
| Baumann et al. | Impulse test fault diagnosis on power transformers using Kohonen's self-organizing neural network | |
| CN116154972B (en) | Distributed power grid power quality monitoring method and system | |
| CN113447771A (en) | Partial discharge pattern recognition method based on SIFT-LDA characteristics | |
| CN103218623B (en) | The radar target feature extraction method differentiating projection is kept based on self-adaptation neighbour | |
| CN114237046A (en) | Partial discharge pattern recognition method based on SIFT data feature extraction algorithm and BP neural network model | |
| CN112464876A (en) | Fault diagnosis method and device for power equipment, computer equipment and storage medium | |
| CN104102900A (en) | Vehicle identification system | |
| CN113935413A (en) | Distribution network wave recording file waveform identification method based on convolutional neural network | |
| CN113514743A (en) | Construction method of GIS partial discharge pattern recognition system based on multi-dimensional features | |
| Raymond et al. | Novel data augmentation for improved insulation fault diagnosis under nonideal condition | |
| CN113780308A (en) | GIS partial discharge mode identification method and system based on kernel principal component analysis and neural network | |
| CN103345739B (en) | A kind of high-resolution remote sensing image building area index calculation method based on texture |
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 |