CN112036450B - High-voltage cable partial discharge mode identification method and system based on transfer learning - Google Patents
High-voltage cable partial discharge mode identification method and system based on transfer learning Download PDFInfo
- Publication number
- CN112036450B CN112036450B CN202010807457.9A CN202010807457A CN112036450B CN 112036450 B CN112036450 B CN 112036450B CN 202010807457 A CN202010807457 A CN 202010807457A CN 112036450 B CN112036450 B CN 112036450B
- Authority
- CN
- China
- Prior art keywords
- target
- data set
- partial discharge
- insulation defect
- feature
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active
Links
- 238000000034 method Methods 0.000 title claims abstract description 27
- 238000013526 transfer learning Methods 0.000 title claims abstract description 25
- 230000007547 defect Effects 0.000 claims abstract description 157
- 238000009413 insulation Methods 0.000 claims abstract description 140
- 238000012549 training Methods 0.000 claims abstract description 68
- 238000012216 screening Methods 0.000 claims abstract description 46
- 238000012360 testing method Methods 0.000 claims abstract description 20
- 238000013527 convolutional neural network Methods 0.000 claims abstract description 9
- 238000004364 calculation method Methods 0.000 claims description 12
- 239000004065 semiconductor Substances 0.000 claims description 10
- 238000013508 migration Methods 0.000 claims description 9
- 230000005012 migration Effects 0.000 claims description 9
- 238000007621 cluster analysis Methods 0.000 claims description 7
- 238000004422 calculation algorithm Methods 0.000 claims description 6
- 238000004590 computer program Methods 0.000 claims description 6
- 238000010276 construction Methods 0.000 claims description 3
- 238000003909 pattern recognition Methods 0.000 claims description 3
- 238000012567 pattern recognition method Methods 0.000 claims 2
- 230000009286 beneficial effect Effects 0.000 description 8
- 238000004458 analytical method Methods 0.000 description 4
- 238000002474 experimental method Methods 0.000 description 3
- 230000008569 process Effects 0.000 description 3
- 238000013528 artificial neural network Methods 0.000 description 2
- 230000015556 catabolic process Effects 0.000 description 2
- 240000005572 Syzygium cordatum Species 0.000 description 1
- 235000006650 Syzygium cordatum Nutrition 0.000 description 1
- 230000009471 action Effects 0.000 description 1
- 230000002950 deficient Effects 0.000 description 1
- 238000006731 degradation reaction Methods 0.000 description 1
- 238000010586 diagram Methods 0.000 description 1
- 238000001035 drying Methods 0.000 description 1
- 238000000605 extraction Methods 0.000 description 1
- 230000006870 function Effects 0.000 description 1
- 230000009545 invasion Effects 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 238000011176 pooling Methods 0.000 description 1
- 238000001228 spectrum Methods 0.000 description 1
Classifications
-
- 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
- G06F18/214—Generating training patterns; Bootstrap methods, e.g. bagging or boosting
-
- 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
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
- G06F18/241—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches
- G06F18/2413—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches based on distances to training or reference patterns
- G06F18/24133—Distances to prototypes
-
- 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
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/045—Combinations of networks
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Data Mining & Analysis (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Evolutionary Computation (AREA)
- General Engineering & Computer Science (AREA)
- Artificial Intelligence (AREA)
- Life Sciences & Earth Sciences (AREA)
- Bioinformatics & Computational Biology (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Evolutionary Biology (AREA)
- General Health & Medical Sciences (AREA)
- Computing Systems (AREA)
- Mathematical Physics (AREA)
- Software Systems (AREA)
- Molecular Biology (AREA)
- Computational Linguistics (AREA)
- Biophysics (AREA)
- Biomedical Technology (AREA)
- Health & Medical Sciences (AREA)
- Testing Relating To Insulation (AREA)
Abstract
The invention relates to a high-voltage cable partial discharge mode identification method and a high-voltage cable partial discharge mode identification system based on transfer learning, wherein the method comprises the steps of obtaining a reference PD characteristic data set corresponding to a known reference insulation defect, and screening out partial discharge core characteristic information; migrating the target PD to the target insulation defect, and testing the target PD to construct a target PD characteristic data set; screening the target PD feature data set according to the reference PD feature data set to generate a target training data set; and establishing and initializing a learning model based on the convolutional neural network, training by using a target training data set, and identifying a partial discharge mode of the target insulation defect by using the trained learning model. According to the invention, the corresponding partial discharge core feature information is extracted by referring to the insulation defect, and is transferred to the target insulation defect, so that the target insulation defect partial discharge mode can be accurately identified under the condition of less feature data by utilizing the learned learning model, and the accuracy of identifying the partial discharge mode is improved.
Description
Technical Field
The invention relates to the technical field of high-voltage cables, in particular to a high-voltage cable partial discharge mode identification method and system based on transfer learning.
Background
The application of high-voltage cables in power systems is becoming more and more widespread, and the high-voltage cables become an integral part of the power systems. Because the cable is operated under high voltage and high current for a long time, under the action of factors such as external damage, equipment defects, water tree invasion and the like, the generated defects of different types can lead the high-voltage cable to generate partial discharge PD (Partial Discharge), the partial discharge of the high-voltage cable can cause the generation and expansion of the electric tree in the cable insulation, the insulation degradation is accelerated, the breakdown is finally caused, and great hidden trouble is brought to the safe operation of the power system. Therefore, in order to find the potential safety hazard generated by partial discharge in advance and avoid sudden accidents of the high-voltage cable, an effective method is needed to identify the partial discharge mode.
As high voltage cables are increasingly used in power systems, various new insulation defects are constantly exposed. When a new type of insulation defect is exposed, a laboratory often cannot perform a large number of partial discharge test experiments in time, which results in very little data for the new type of insulation defect, so that the new type of insulation defect cannot be accurately identified.
Disclosure of Invention
The invention aims to solve the technical problem of providing a high-voltage cable partial discharge mode identification method and system based on transfer learning aiming at the defects of the prior art.
The technical scheme for solving the technical problems is as follows: a high-voltage cable partial discharge mode identification method based on transfer learning comprises the following steps:
acquiring a reference PD characteristic data set corresponding to a known reference insulation defect of a high-voltage cable, and screening out partial discharge core characteristic information of the reference insulation defect;
migrating the partial discharge core characteristic information of the reference insulation defect to a target insulation defect, testing the target insulation defect, and constructing a target PD characteristic data set corresponding to the target insulation defect;
screening the target PD feature data set according to the reference PD feature data set to generate a target training data set;
and establishing and initializing a learning model based on a convolutional neural network, training by using the target training data set, and identifying a partial discharge mode of the target insulation defect by using the trained learning model.
The beneficial effects of the invention are as follows: according to the high-voltage cable partial discharge mode identification method based on transfer learning, through analyzing and screening the known reference insulation defects, extracting the corresponding partial discharge core characteristic information, transferring the partial discharge core characteristic information to the target insulation defects, constructing the target PD characteristic data set, screening the target training data set from the target PD characteristic data set, learning through a learning model, accurately identifying the partial discharge mode of the target insulation defects under the condition of less characteristic data by utilizing the learned learning model, and improving the accuracy of partial discharge mode identification.
Based on the technical scheme, the invention can also be improved as follows:
further: before acquiring the reference PD signature data set corresponding to the known reference insulation defect of the high voltage cable, the method further includes:
and searching for an insulation defect similar to the type of the target insulation defect from the known typical insulation defects of the high-voltage cable according to the defect position, whether the outer semiconductor is contacted or not and the damage depth, and forming the reference insulation defect.
The beneficial effects of the above-mentioned further scheme are: according to the defect position, whether the outer semiconductor is contacted or not and the damage depth, the insulation defect similar to the type of the target insulation defect can be searched from the known typical insulation defects of the high-voltage cable, so that the insulation defect which is close to the target insulation defect can be primarily screened, the target PD characteristic data set can be conveniently screened according to the reference PD characteristic data set in the follow-up process, the calculated amount is reduced, the calculation time is shortened, and the calculation efficiency is improved.
Further: the step of acquiring the PD characteristic data set of the reference insulation defect and screening out the partial discharge core characteristic information of the reference insulation defect specifically comprises the following steps:
testing the reference insulation defect to obtain a reference PD characteristic data set;
classifying the reference PD feature data set by using a cluster analysis algorithm, and screening partial discharge core feature information used for representing the partial discharge characteristics from the reference PD feature data set according to a preset partial discharge core parameter type.
The beneficial effects of the above-mentioned further scheme are: through testing the reference insulation defects, a corresponding reference PD feature data set can be accurately obtained, and then classification is carried out by combining a clustering analysis method, so that partial discharge core feature information can be screened out from the reference PD feature data set according to a preset partial discharge core parameter type, and the partial discharge core feature information can be conveniently and subsequently subjected to migration learning.
Further: the step of screening the target PD feature data set according to the reference PD feature data set to generate a target training data set specifically comprises the following steps:
calculating a reference PD feature value according to the reference PD feature data set, and calculating a target PD feature value according to the target PD feature data set;
calculating the Euclidean distance between the reference PD feature value and the target PD feature value, extracting feature data corresponding to the PD feature value with the highest matching degree with the target PD feature value according to the Euclidean distance, wherein the calculation formula of the Euclidean distance is as follows:
wherein dist (X, Y) represents Euclidean distance of two groups of characteristic data, n is the number of characteristic variables, and X i 、y i Respectively representing corresponding characteristic variables in the two sets of characteristic data;
and setting the label number of the extracted characteristic data as the label number corresponding to the target insulation defect, and generating a target training data set.
The beneficial effects of the above-mentioned further scheme are: through calculating the Euclidean distance between the reference PD characteristic value and the target PD characteristic value, the matching degree between the corresponding characteristic data between the two groups of characteristics can be reflected through the Euclidean distance, so that the characteristic data with the highest matching degree can be selected to generate a more accurate training data set, and subsequent learning and training are facilitated.
Further: the training is performed by using the target training data set, and the partial discharge mode for identifying the target insulation defect by using the trained learning model specifically comprises the following steps:
training a first classifier in the learning model by using the target training data set, and extracting network weights of all layers in the first classifier;
and migrating the network weight to a second classifier in the learning model, identifying the target insulation defect by using the second classifier, and outputting an identification result.
The beneficial effects of the above-mentioned further scheme are: the first classifier is trained through the target training data set, so that the network weights of all layers in the first classifier can be accurately obtained after a sufficient number of training, and the network weights are transferred to the second classifier, and the accurate identification of the target insulation defects is completed.
The invention also provides a high-voltage cable partial discharge mode identification system based on transfer learning, which comprises the following steps:
the acquisition and screening module is used for acquiring a reference PD characteristic data set corresponding to a known reference insulation defect of the high-voltage cable and screening out partial discharge core characteristic information of the reference insulation defect;
the migration construction module is used for migrating the partial discharge core characteristic information of the reference insulation defect to a target insulation defect, testing the target insulation defect and constructing a target PD characteristic data set corresponding to the target insulation defect;
the feature screening module is used for screening the target PD feature data set according to the reference PD feature data set to generate a target training data set;
and the training recognition module is used for establishing and initializing a learning model based on the convolutional neural network, training by using the target training data set, and recognizing a partial discharge mode of the target insulation defect by using the trained learning model.
According to the high-voltage cable partial discharge mode identification system based on transfer learning, through analyzing and screening the known reference insulation defects, extracting the corresponding partial discharge core characteristic information, transferring the partial discharge core characteristic information to the target insulation defects, constructing the target PD characteristic data set, screening the target training data set from the target PD characteristic data set, learning through a learning model, accurately identifying the partial discharge mode of the target insulation defects under the condition of less characteristic data by utilizing the learned learning model, and improving the accuracy of partial discharge mode identification.
Based on the technical scheme, the invention can also be improved as follows:
further: the acquisition and screening module acquires a reference PD feature data set corresponding to the reference insulation defect, and screens out the partial discharge core feature information of the reference insulation defect, wherein the specific implementation is as follows:
testing the reference insulation defect to obtain a reference PD characteristic data set;
classifying the reference PD feature data set by using a cluster analysis algorithm, and screening partial discharge core feature information used for representing the partial discharge characteristics from the reference PD feature data set according to a preset partial discharge core parameter type.
The beneficial effects of the above-mentioned further scheme are: through testing the reference insulation defects, a corresponding reference PD feature data set can be accurately obtained, and then classification is carried out by combining a clustering analysis method, so that partial discharge core feature information can be screened out from the reference PD feature data set according to a preset partial discharge core parameter type, and the partial discharge core feature information can be conveniently and subsequently subjected to migration learning.
Further: the feature screening module is specifically used for:
calculating a reference PD feature value according to the reference PD feature data set, and calculating a target PD feature value according to the target PD feature data set;
calculating the Euclidean distance between the reference PD feature value and the target PD feature value, extracting feature data corresponding to the PD feature value with the highest matching degree with the target PD feature value according to the Euclidean distance, wherein the calculation formula of the Euclidean distance is as follows:
wherein dist (X, Y) represents Euclidean distance of two groups of characteristic data, n is the number of characteristic variables, and X i 、y i Respectively representing corresponding characteristic variables in the two sets of characteristic data;
and setting the label number of the extracted characteristic data as the label number corresponding to the target insulation defect, and generating a target training data set.
The beneficial effects of the above-mentioned further scheme are: through calculating the Euclidean distance between the reference PD characteristic value and the target PD characteristic value, the matching degree between the corresponding characteristic data between the two groups of characteristics can be reflected through the Euclidean distance, so that the characteristic data with the highest matching degree can be selected to generate a more accurate training data set, and subsequent learning and training are facilitated.
The invention also provides a computer readable storage medium storing a computer program which, when executed by a processor, implements the method.
The invention also provides high-voltage cable partial discharge mode identification equipment based on transfer learning, which comprises the storage medium and a processor, wherein the processor realizes the steps of the method when executing the computer program on the storage medium.
Drawings
FIG. 1 is a schematic flow chart of a high-voltage cable partial discharge mode identification method based on transfer learning;
fig. 2 is a schematic structural diagram of a high-voltage cable partial discharge mode identification system based on transfer learning according to the present invention.
Detailed Description
The principles and features of the present invention are described below with reference to the drawings, the examples are illustrated for the purpose of illustrating the invention and are not to be construed as limiting the scope of the invention.
As shown in fig. 1, a high-voltage cable partial discharge mode identification method based on transfer learning includes the following steps:
s11: acquiring a reference PD characteristic data set corresponding to a known reference insulation defect of a high-voltage cable, and screening out partial discharge core characteristic information of the reference insulation defect;
s12: migrating the partial discharge core characteristic information of the reference insulation defect to a target insulation defect, testing the target insulation defect, and constructing a target PD characteristic data set corresponding to the target insulation defect;
s13: screening the target PD feature data set according to the reference PD feature data set to generate a target training data set;
s14: and establishing and initializing a learning model based on a convolutional neural network, training by using the target training data set, and identifying a partial discharge mode of the target insulation defect by using the trained learning model.
The beneficial effects of the invention are as follows: according to the high-voltage cable partial discharge mode identification method based on transfer learning, through analyzing and screening the known reference insulation defects, extracting the corresponding partial discharge core characteristic information, transferring the partial discharge core characteristic information to the target insulation defects, constructing the target PD characteristic data set, screening the target training data set from the target PD characteristic data set, learning through a learning model, accurately identifying the partial discharge mode of the target insulation defects under the condition of less characteristic data by utilizing the learned learning model, and improving the accuracy of partial discharge mode identification.
Optionally, in one or more embodiments of the present invention, before acquiring the reference PD signature data set corresponding to the known reference insulation defect of the high voltage cable, the method further includes:
s10: and searching for an insulation defect similar to the type of the target insulation defect from the known typical insulation defects of the high-voltage cable according to the defect position, whether the outer semiconductor is contacted or not and the damage depth, and forming the reference insulation defect.
According to the defect position, whether the outer semiconductor is contacted or not and the damage depth, the insulation defect similar to the type of the target insulation defect can be searched from the known typical insulation defects of the high-voltage cable, so that the insulation defect which is close to the target insulation defect can be primarily screened, the target PD characteristic data set can be conveniently screened according to the reference PD characteristic data set in the follow-up process, the calculated amount is reduced, the calculation time is shortened, and the calculation efficiency is improved.
Because the neural network classifier is generally quite robust, the neural network classifier is simply screened according to preset similarity judgment standards (such as whether defect positions are the same, whether outer semiconductors are contacted and whether defect depths are the same (or the levels of the defect depths)) and the function is only to reduce the time for calculating the Euclidean distance subsequently. Such as insulation hole defects, semiconductor contact spike defects, semiconductor non-contact spike defects, cable jacket breakage defects, etc., may be counted as similar defects.
In one or more embodiments of the present invention, the acquiring the PD feature data set of the reference insulation defect and screening out the partial discharge core feature information of the reference insulation defect specifically includes the following steps:
s21: testing the reference insulation defect to obtain a reference PD characteristic data set;
s22: classifying the reference PD feature data set by using a cluster analysis algorithm, and screening partial discharge core feature information used for representing the partial discharge characteristics from the reference PD feature data set according to a preset partial discharge core parameter type.
Through testing the reference insulation defects, a corresponding reference PD feature data set can be accurately obtained, and then classification is carried out by combining a clustering analysis method, so that partial discharge core feature information can be screened out from the reference PD feature data set according to a preset partial discharge core parameter type, and the partial discharge core feature information can be conveniently and subsequently subjected to migration learning.
In practice, the instantaneous pulse signal is obtained by testing the reference insulation characteristic, and the PD characteristic data set is generated in consideration of frequency, amplitude and phase spectrum. For example, the discharge amount, peak voltage, average voltage, root mean square value, standard deviation, etc. may be selected from the amplitude characteristics, the pulse width, rise time, fall time, equivalent time length, equivalent width, etc. may be selected from the frequency characteristics, and the phase angle may be selected from the phase map characteristics, as well as other characteristics such as pulse acuity, bias, peak value, crest factor, shape factor, defect location, etc. And then, carrying out de-drying and PD pulse extraction on the data obtained by the tests, and constructing PD characteristic parameters to obtain a PD characteristic data set.
Since the partial discharge signal curve contains a large number of features, training time is seriously increased by using all the features for training a model, and thus features with the greatest influence on partial discharge, such as peak voltage, pulse width, rise and fall time, crest factor, shape factor, and the like, need to be screened therefrom by using cluster analysis. Since a large number of experiments have been performed on known defects, the PD characteristic data is generally very abundant, and it is difficult to perform a large number of detailed experiments on the target insulation defects in a short time, and the PD characteristic data is generally deficient. In the PD signature dataset, the label for each sample is the corresponding defect type.
In one or more embodiments of the present invention, the screening the target PD feature data set according to the reference PD feature data set, and generating the target training data set specifically includes the following steps:
s31: calculating a reference PD feature value according to the reference PD feature data set, and calculating a target PD feature value according to the target PD feature data set;
s32: calculating the Euclidean distance between the reference PD feature value and the target PD feature value, extracting feature data corresponding to the PD feature value with the highest matching degree with the target PD feature value according to the Euclidean distance, wherein the calculation formula of the Euclidean distance is as follows:
wherein dist (X, Y) represents Euclidean distance of two groups of characteristic data, n is the number of characteristic variables, and X i 、y i Respectively representing corresponding characteristic variables in the two sets of characteristic data;
s33: and setting the label number of the extracted characteristic data as the label number corresponding to the target insulation defect, and generating a target training data set.
Through calculating the Euclidean distance between the reference PD characteristic value and the target PD characteristic value, the matching degree between the corresponding characteristic data between the two groups of characteristics can be reflected through the Euclidean distance, so that the characteristic data with the highest matching degree can be selected to generate a more accurate training data set, and subsequent learning and training are facilitated.
In one or more embodiments of the present invention, the training using the target training data set and the identifying the partial discharge pattern of the target insulation defect using the trained learning model specifically includes the following steps:
s41: training a first classifier in the learning model by using the target training data set, and extracting network weights of all layers in the first classifier;
in practice, convolutional neural network-based learning models typically include an input layer, at least one convolutional layer, a pooling layer, a fully-connected layer, and an output layer. And training the first classifier to extract the network weight of each layer in the first classifier.
S42: and migrating the network weight to a second classifier in the learning model, identifying the target insulation defect by using the second classifier, and outputting an identification result.
The first classifier is trained through the target training data set, so that the network weights of all layers in the first classifier can be accurately obtained after a sufficient number of training, and the network weights are transferred to the second classifier, and the accurate identification of the target insulation defects is completed. In this embodiment, the first classifier and the second classifier may both select an SVM classifier or a CNN classifier.
As shown in fig. 2, the present invention further provides a high-voltage cable partial discharge mode identification system based on transfer learning, which includes:
the acquisition and screening module is used for acquiring a reference PD characteristic data set corresponding to a known reference insulation defect of the high-voltage cable and screening out partial discharge core characteristic information of the reference insulation defect;
the migration construction module is used for migrating the partial discharge core characteristic information of the reference insulation defect to a target insulation defect, testing the target insulation defect and constructing a target PD characteristic data set corresponding to the target insulation defect;
the feature screening module is used for screening the target PD feature data set according to the reference PD feature data set to generate a target training data set;
and the training recognition module is used for establishing and initializing a learning model based on the convolutional neural network, training by using the target training data set, and recognizing a partial discharge mode of the target insulation defect by using the trained learning model.
According to the high-voltage cable partial discharge mode identification system based on transfer learning, through analyzing and screening the known reference insulation defects, extracting the corresponding partial discharge core characteristic information, transferring the partial discharge core characteristic information to the target insulation defects, constructing the target PD characteristic data set, screening the target training data set from the target PD characteristic data set, learning through a learning model, accurately identifying the partial discharge mode of the target insulation defects under the condition of less characteristic data by utilizing the learned learning model, and improving the accuracy of partial discharge mode identification.
Optionally, in one or more embodiments of the present invention, the high voltage cable partial discharge pattern recognition system based on migration learning further includes a matching module for searching for an insulation defect similar to a target insulation defect type from typical insulation defects known to the high voltage cable according to a defect position, whether an outer semiconductor is contacted, and a breakage depth, and forming the reference insulation defect.
According to the defect position, whether the outer semiconductor is contacted or not and the damage depth, the insulation defect similar to the type of the target insulation defect can be searched from the known typical insulation defects of the high-voltage cable, so that the insulation defect which is close to the target insulation defect can be primarily screened, the target PD characteristic data set can be conveniently screened according to the reference PD characteristic data set in the follow-up process, the calculated amount is reduced, the calculation time is shortened, and the calculation efficiency is improved.
In one or more embodiments of the present invention, the acquiring and screening module acquires a reference PD feature data set corresponding to the reference insulation defect, and screens out partial discharge core feature information of the reference insulation defect, where the specific implementation is as follows:
testing the reference insulation defect to obtain a reference PD characteristic data set;
classifying the reference PD feature data set by using a cluster analysis algorithm, and screening partial discharge core feature information used for representing the partial discharge characteristics from the reference PD feature data set according to a preset partial discharge core parameter type.
Through testing the reference insulation defects, a corresponding reference PD feature data set can be accurately obtained, and then classification is carried out by combining a clustering analysis method, so that partial discharge core feature information can be screened out from the reference PD feature data set according to a preset partial discharge core parameter type, and the partial discharge core feature information can be conveniently and subsequently subjected to migration learning.
In one or more embodiments of the present invention, the feature screening module is specifically configured to:
calculating a reference PD feature value according to the reference PD feature data set, and calculating a target PD feature value according to the target PD feature data set;
calculating the Euclidean distance between the reference PD feature value and the target PD feature value, extracting feature data corresponding to the PD feature value with the highest matching degree with the target PD feature value according to the Euclidean distance, wherein the calculation formula of the Euclidean distance is as follows:
wherein dist (X, Y) represents Euclidean distance of two groups of characteristic data, n is the number of characteristic variables, and X i 、y i Respectively representing corresponding characteristic variables in the two sets of characteristic data;
and setting the label number of the extracted characteristic data as the label number corresponding to the target insulation defect, and generating a target training data set.
Through calculating the Euclidean distance between the reference PD characteristic value and the target PD characteristic value, the matching degree between the corresponding characteristic data between the two groups of characteristics can be reflected through the Euclidean distance, so that the characteristic data with the highest matching degree can be selected to generate a more accurate training data set, and subsequent learning and training are facilitated.
In one or more embodiments of the present invention, the training recognition module is specifically configured to:
training a first classifier in the learning model by using the target training data set, and extracting network weights of all layers in the first classifier;
and migrating the network weight to a second classifier in the learning model, identifying the target insulation defect by using the second classifier, and outputting an identification result.
The invention also provides a computer readable storage medium storing a computer program which, when executed by a processor, implements the method.
The invention also provides high-voltage cable partial discharge mode identification equipment based on transfer learning, which comprises the storage medium and a processor, wherein the processor realizes the steps of the method when executing the computer program on the storage medium.
The foregoing description of the preferred embodiments of the invention is not intended to limit the invention to the precise form disclosed, and any such modifications, equivalents, and alternatives falling within the spirit and scope of the invention are intended to be included within the scope of the invention.
Claims (10)
1. The high-voltage cable partial discharge mode identification method based on transfer learning is characterized by comprising the following steps of:
acquiring a reference PD characteristic data set corresponding to a known reference insulation defect of a high-voltage cable, and screening out partial discharge core characteristic information of the reference insulation defect;
migrating the partial discharge core characteristic information of the reference insulation defect to a target insulation defect, testing the target insulation defect, and constructing a target PD characteristic data set corresponding to the target insulation defect;
screening the target PD feature data set according to the reference PD feature data set to generate a target training data set;
and establishing and initializing a learning model based on a convolutional neural network, training by using the target training data set, and identifying a partial discharge mode of the target insulation defect by using the trained learning model.
2. The method for identifying a partial discharge pattern of a high-voltage cable based on transfer learning according to claim 1, wherein before obtaining a reference PD feature data set corresponding to a known reference insulation defect of the high-voltage cable, the method further comprises:
and searching for an insulation defect similar to the type of the target insulation defect from the known typical insulation defects of the high-voltage cable according to the defect position, whether the outer semiconductor is contacted or not and the damage depth, and forming the reference insulation defect.
3. The high-voltage cable partial discharge mode identification method based on transfer learning according to claim 1, wherein the steps of obtaining the PD feature data set of the reference insulation defect and screening out the partial discharge core feature information of the reference insulation defect specifically comprise the following steps:
testing the reference insulation defect to obtain a reference PD characteristic data set;
classifying the reference PD feature data set by using a cluster analysis algorithm, and screening partial discharge core feature information used for representing the partial discharge characteristics from the reference PD feature data set according to a preset partial discharge core parameter type.
4. The high-voltage cable partial discharge pattern recognition method based on transfer learning according to claim 3, wherein the step of screening the target PD feature data set according to the reference PD feature data set to generate a target training data set specifically comprises the steps of:
calculating a reference PD feature value according to the reference PD feature data set, and calculating a target PD feature value according to the target PD feature data set;
calculating the Euclidean distance between the reference PD feature value and the target PD feature value, extracting feature data corresponding to the PD feature value with the highest matching degree with the target PD feature value according to the Euclidean distance, wherein the calculation formula of the Euclidean distance is as follows:
wherein dist (X, Y) represents two groupsEuclidean distance of characteristic data, n is the number of characteristic variables, x i 、y i Respectively representing corresponding characteristic variables in the two sets of characteristic data;
and setting the label number of the extracted characteristic data as the label number corresponding to the target insulation defect, and generating a target training data set.
5. The high-voltage cable partial discharge pattern recognition method based on transfer learning according to any one of claims 1 to 4, wherein the training using the target training data set and the learning model after training to recognize the partial discharge pattern of the target insulation defect specifically comprises the steps of:
training a first classifier in the learning model by using the target training data set, and extracting network weights of all layers in the first classifier;
and migrating the network weight to a second classifier in the learning model, identifying the target insulation defect by using the second classifier, and outputting an identification result.
6. High-voltage cable partial discharge mode identification system based on migration learning, characterized by comprising:
the acquisition and screening module is used for acquiring a reference PD characteristic data set corresponding to a known reference insulation defect of the high-voltage cable and screening out partial discharge core characteristic information of the reference insulation defect;
the migration construction module is used for migrating the partial discharge core characteristic information of the reference insulation defect to a target insulation defect, testing the target insulation defect and constructing a target PD characteristic data set corresponding to the target insulation defect;
the feature screening module is used for screening the target PD feature data set according to the reference PD feature data set to generate a target training data set;
and the training recognition module is used for establishing and initializing a learning model based on the convolutional neural network, training by using the target training data set, and recognizing a partial discharge mode of the target insulation defect by using the trained learning model.
7. The high-voltage cable partial discharge mode identification system based on transfer learning according to claim 6, wherein the obtaining and screening module obtains a reference PD feature data set corresponding to the reference insulation defect, and screens out the partial discharge core feature information of the reference insulation defect, which is specifically implemented as follows:
testing the reference insulation defect to obtain a reference PD characteristic data set;
classifying the reference PD feature data set by using a cluster analysis algorithm, and screening partial discharge core feature information used for representing the partial discharge characteristics from the reference PD feature data set according to a preset partial discharge core parameter type.
8. The high-voltage cable partial discharge pattern recognition system based on transfer learning of claim 7, wherein the feature screening module is specifically configured to:
calculating a reference PD feature value according to the reference PD feature data set, and calculating a target PD feature value according to the target PD feature data set;
calculating the Euclidean distance between the reference PD feature value and the target PD feature value, extracting feature data corresponding to the PD feature value with the highest matching degree with the target PD feature value according to the Euclidean distance, wherein the calculation formula of the Euclidean distance is as follows:
wherein dist (X, Y) represents Euclidean distance of two groups of characteristic data, n is the number of characteristic variables, and X i 、y i Respectively representing corresponding characteristic variables in the two sets of characteristic data;
and setting the label number of the extracted characteristic data as the label number corresponding to the target insulation defect, and generating a target training data set.
9. A computer readable storage medium, characterized in that a computer program is stored, which, when being executed by a processor, implements the method of any of claims 1 to 5.
10. A high voltage cable partial discharge pattern recognition device based on transfer learning, comprising the storage medium of claim 9 and a processor implementing the steps of the method of any of claims 1 to 5 when executing a computer program on the storage medium.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202010807457.9A CN112036450B (en) | 2020-08-12 | 2020-08-12 | High-voltage cable partial discharge mode identification method and system based on transfer learning |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202010807457.9A CN112036450B (en) | 2020-08-12 | 2020-08-12 | High-voltage cable partial discharge mode identification method and system based on transfer learning |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN112036450A CN112036450A (en) | 2020-12-04 |
| CN112036450B true CN112036450B (en) | 2024-02-23 |
Family
ID=73578614
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202010807457.9A Active CN112036450B (en) | 2020-08-12 | 2020-08-12 | High-voltage cable partial discharge mode identification method and system based on transfer learning |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN112036450B (en) |
Families Citing this family (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN113533910A (en) * | 2021-06-10 | 2021-10-22 | 中国电力科学研究院有限公司 | Method and system suitable for converter transformer partial discharge early warning |
| CN117542048B (en) * | 2024-01-05 | 2024-03-22 | 中信戴卡股份有限公司 | Automatic identification method for microstructure characteristics and defect characteristics of hypoeutectic aluminum-silicon alloy |
| CN117786521B (en) * | 2024-02-23 | 2024-04-19 | 兰州交通大学 | Signal classification method based on intelligent frequency spectrum sensing |
Citations (11)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103645425A (en) * | 2013-11-15 | 2014-03-19 | 广东电网公司电力科学研究院 | High-voltage cable insulation defect partial discharge on-line monitoring diagnosis method |
| CN108846418A (en) * | 2018-05-24 | 2018-11-20 | 广东电网有限责任公司 | A kind of positioning of cable machinery temperature anomaly and recognition methods |
| CN109212392A (en) * | 2018-09-25 | 2019-01-15 | 上海交通大学 | Recognition methods, system and the Caffe convolutional neural networks of direct current cables shelf depreciation defect failure |
| CN109614888A (en) * | 2018-11-23 | 2019-04-12 | 华南理工大学 | Deep learning defects detection model training method based on overhead transmission line defect auxiliary data collection |
| 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 |
| CN109785289A (en) * | 2018-12-18 | 2019-05-21 | 中国科学院深圳先进技术研究院 | A kind of transmission line of electricity defect inspection method, system and electronic equipment |
| CN109886232A (en) * | 2019-02-28 | 2019-06-14 | 燊赛(上海)智能科技有限公司 | A kind of power grid image identification system neural network based |
| WO2019216878A1 (en) * | 2018-05-07 | 2019-11-14 | Inhand Networks Inc. | System for locating fault in power distribution network based on mixed mode wave recording |
| CN110943453A (en) * | 2019-12-23 | 2020-03-31 | 北京交通大学 | Power system fault sample generation and model construction method facing transfer learning |
| CN111046581A (en) * | 2019-12-27 | 2020-04-21 | 国网江苏省电力有限公司电力科学研究院 | Power transmission line fault type identification method and system |
| CN111401384A (en) * | 2020-03-12 | 2020-07-10 | 安徽南瑞继远电网技术有限公司 | Transformer equipment defect image matching method |
-
2020
- 2020-08-12 CN CN202010807457.9A patent/CN112036450B/en active Active
Patent Citations (11)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103645425A (en) * | 2013-11-15 | 2014-03-19 | 广东电网公司电力科学研究院 | High-voltage cable insulation defect partial discharge on-line monitoring diagnosis method |
| WO2019216878A1 (en) * | 2018-05-07 | 2019-11-14 | Inhand Networks Inc. | System for locating fault in power distribution network based on mixed mode wave recording |
| CN108846418A (en) * | 2018-05-24 | 2018-11-20 | 广东电网有限责任公司 | A kind of positioning of cable machinery temperature anomaly and recognition methods |
| CN109212392A (en) * | 2018-09-25 | 2019-01-15 | 上海交通大学 | Recognition methods, system and the Caffe convolutional neural networks of direct current cables shelf depreciation defect failure |
| 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 |
| CN109614888A (en) * | 2018-11-23 | 2019-04-12 | 华南理工大学 | Deep learning defects detection model training method based on overhead transmission line defect auxiliary data collection |
| CN109785289A (en) * | 2018-12-18 | 2019-05-21 | 中国科学院深圳先进技术研究院 | A kind of transmission line of electricity defect inspection method, system and electronic equipment |
| CN109886232A (en) * | 2019-02-28 | 2019-06-14 | 燊赛(上海)智能科技有限公司 | A kind of power grid image identification system neural network based |
| CN110943453A (en) * | 2019-12-23 | 2020-03-31 | 北京交通大学 | Power system fault sample generation and model construction method facing transfer learning |
| CN111046581A (en) * | 2019-12-27 | 2020-04-21 | 国网江苏省电力有限公司电力科学研究院 | Power transmission line fault type identification method and system |
| CN111401384A (en) * | 2020-03-12 | 2020-07-10 | 安徽南瑞继远电网技术有限公司 | Transformer equipment defect image matching method |
Non-Patent Citations (1)
| Title |
|---|
| 基于迁移学习卷积神经网络的电缆隧道锈蚀识别算法;周自强等;《中国电力》;第52卷(第4期);104-110 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN112036450A (en) | 2020-12-04 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN112036450B (en) | High-voltage cable partial discharge mode identification method and system based on transfer learning | |
| CN108573225B (en) | Partial discharge signal pattern recognition method and system | |
| CN109118479B (en) | Capsule network-based insulator defect identification and positioning device and method | |
| CN108957251B (en) | Cable joint partial discharge mode identification method | |
| US20230300159A1 (en) | Network traffic anomaly detection method and apparatus, and electronic apparatus and storage medium | |
| CN113435546B (en) | Migratable image recognition method and system based on differentiation confidence level | |
| CN110161388B (en) | Fault type identification method and system of high-voltage equipment | |
| CN109188211B (en) | High-voltage equipment insulation fault diagnosis method and system | |
| CN111882554B (en) | SK-YOLOv 3-based intelligent power line fault detection method | |
| CN110726898B (en) | Power distribution network fault type identification method | |
| Adou et al. | Insulator faults detection based on deep learning | |
| CN116167010B (en) | Rapid identification method for abnormal events of power system with intelligent transfer learning capability | |
| Ma et al. | Detection method of insulator based on faster R-CNN | |
| CN112528774B (en) | Intelligent unknown radar signal sorting system and method in complex electromagnetic environment | |
| CN111476093A (en) | Cable terminal partial discharge mode identification method and system | |
| CN113447771A (en) | Partial discharge pattern recognition method based on SIFT-LDA characteristics | |
| CN114240817A (en) | Data analysis method and device, electronic equipment and storage medium | |
| CN116756594A (en) | Method, system, equipment and medium for detecting abnormal points of power grid data | |
| CN113283371A (en) | Local discharge feature extraction and classification method based on BRISK features | |
| Steiger et al. | Automatic detection of transmission towers | |
| CN112083275A (en) | Distribution network fault type identification method and system | |
| CN110543675A (en) | Power transmission line fault identification method | |
| CN112014821B (en) | Unknown vehicle target identification method based on radar broadband characteristics | |
| CN114841266A (en) | Voltage sag identification method based on triple prototype network under small sample | |
| CN111626329B (en) | Insulation pipe bus fault diagnosis method based on LDA optimizing multi-scale texture characteristics |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |