CN117312930A - Transformer zero sample fault diagnosis method, device, storage medium and equipment - Google Patents
Transformer zero sample fault diagnosis method, device, storage medium and equipment Download PDFInfo
- Publication number
- CN117312930A CN117312930A CN202311606768.9A CN202311606768A CN117312930A CN 117312930 A CN117312930 A CN 117312930A CN 202311606768 A CN202311606768 A CN 202311606768A CN 117312930 A CN117312930 A CN 117312930A
- Authority
- CN
- China
- Prior art keywords
- data
- attribute
- transformer
- fault
- sample
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
- 238000000034 method Methods 0.000 title claims abstract description 53
- 238000003745 diagnosis Methods 0.000 title claims abstract description 35
- 238000012549 training Methods 0.000 claims abstract description 36
- 239000011159 matrix material Substances 0.000 claims description 27
- 239000013598 vector Substances 0.000 claims description 17
- 238000003066 decision tree Methods 0.000 claims description 6
- 238000003672 processing method Methods 0.000 claims description 6
- 230000011218 segmentation Effects 0.000 claims description 6
- 238000012545 processing Methods 0.000 claims description 5
- 238000004590 computer program Methods 0.000 claims description 4
- 230000006870 function Effects 0.000 claims description 3
- 230000007246 mechanism Effects 0.000 abstract description 2
- 230000010354 integration Effects 0.000 abstract 1
- 239000007789 gas Substances 0.000 description 5
- 230000008569 process Effects 0.000 description 5
- 238000001514 detection method Methods 0.000 description 4
- 230000000694 effects Effects 0.000 description 4
- 238000001834 photoacoustic spectrum Methods 0.000 description 3
- 238000010586 diagram Methods 0.000 description 2
- 230000004927 fusion Effects 0.000 description 2
- 238000010801 machine learning Methods 0.000 description 2
- 238000012986 modification Methods 0.000 description 2
- 230000004048 modification Effects 0.000 description 2
- 238000004088 simulation Methods 0.000 description 2
- 238000013526 transfer learning Methods 0.000 description 2
- 238000004458 analytical method Methods 0.000 description 1
- 238000013528 artificial neural network Methods 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 230000005540 biological transmission Effects 0.000 description 1
- 230000015556 catabolic process Effects 0.000 description 1
- 238000013135 deep learning Methods 0.000 description 1
- 230000007547 defect Effects 0.000 description 1
- 238000006731 degradation reaction Methods 0.000 description 1
- 239000002360 explosive Substances 0.000 description 1
- 230000005484 gravity Effects 0.000 description 1
- 230000036541 health Effects 0.000 description 1
- 239000007788 liquid Substances 0.000 description 1
- 238000012544 monitoring process Methods 0.000 description 1
- 238000005457 optimization Methods 0.000 description 1
- 238000011160 research Methods 0.000 description 1
- 230000035945 sensitivity Effects 0.000 description 1
- 238000001228 spectrum Methods 0.000 description 1
- 238000012706 support-vector machine Methods 0.000 description 1
- 238000012360 testing method Methods 0.000 description 1
- 238000001931 thermography 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
- G06F18/24—Classification techniques
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F17/00—Digital computing or data processing equipment or methods, specially adapted for specific functions
- G06F17/10—Complex mathematical operations
- G06F17/16—Matrix or vector computation, e.g. matrix-matrix or matrix-vector multiplication, matrix factorization
-
- 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
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/25—Fusion techniques
- G06F18/253—Fusion techniques of extracted features
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N20/00—Machine learning
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Data Mining & Analysis (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- General Engineering & Computer Science (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Evolutionary Computation (AREA)
- Artificial Intelligence (AREA)
- Mathematical Physics (AREA)
- Evolutionary Biology (AREA)
- Bioinformatics & Computational Biology (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Life Sciences & Earth Sciences (AREA)
- Software Systems (AREA)
- Mathematical Analysis (AREA)
- Mathematical Optimization (AREA)
- Computing Systems (AREA)
- Pure & Applied Mathematics (AREA)
- Computational Mathematics (AREA)
- Algebra (AREA)
- Databases & Information Systems (AREA)
- Medical Informatics (AREA)
- Testing Relating To Insulation (AREA)
Abstract
The invention discloses a method, a device, a storage medium and equipment for diagnosing zero sample faults of a transformer, and belongs to the technical field of transformer fault diagnosis. The method comprises the following steps: acquiring sample data of various sensors of a target transformer; acquiring fault types based on a pre-trained attribute learner according to the acquired sample data of various sensors of the target transformer; wherein the training of the attribute learner comprises: respectively acquiring sample data of a one-dimensional sensor and a multi-dimensional sensor in a target transformer; first data features corresponding to sample data of the one-dimensional sensor and second data features corresponding to sample data of the multi-dimensional sensor are obtained. According to the invention, the first data features and the second data features corresponding to the sample data of various sensors of the transformer are obtained, and the semantics are embedded into the first data features and the second data features, so that the integration of multiple modes is realized, and a flexible online supervision mechanism for the fault diagnosis of the transformer is realized.
Description
Technical Field
The invention relates to a method, a device, a storage medium and equipment for diagnosing zero sample faults of a transformer, and belongs to the technical field of transformer fault diagnosis.
Background
The power transformer is used as large industrial core equipment, and plays a key supporting role in the safe operation of the whole power system. In recent years, with the optimization of a power system and the explosive access of a large number of new energy systems, the installed amount of a transformer is multiplied, so how to diagnose the faults of the transformer efficiently and accurately becomes a difficult problem to be solved in the field.
At present, the traditional transformer fault diagnosis method mostly depends on single fault information, and only evaluates the health condition of the transformer through the content, ratio or vibration, temperature and the like of dissolved gas in oil, but neglects to acquire other important fault parameters from the multi-mode angle. However, the algorithms such as neural networks, deep learning and the like which are used as research hotspots at present require a large amount of data to be driven and have long training time, and once the internal structure of the transformer is changed, the algorithms need to be retrained, which is time-consuming, labor-consuming and difficult to popularize. More importantly, since the early monitoring level of the power grid is not perfect enough and the transformer is a large-scale power transmission and distribution equipment, a large-scale fault can rarely occur, so that certain specific fault data are very difficult to acquire, and the data size is insufficient to support the training requirement of a conventional intelligent algorithm.
Disclosure of Invention
The invention aims to overcome the defects in the prior art, provides a transformer zero sample fault diagnosis method, device, storage medium and equipment, and solves the problems of single target detection angle, low diagnosis precision and excessive dependence on fault sample data in the prior art.
In order to achieve the above purpose/solve the above technical problems, the present invention is realized by adopting the following technical scheme:
in a first aspect, a method for diagnosing a zero sample fault of a transformer includes:
acquiring sample data of various sensors of a target transformer;
acquiring fault types based on a pre-trained attribute learner according to the acquired sample data of various sensors of the target transformer;
wherein the training of the attribute learner comprises:
respectively acquiring sample data of a one-dimensional sensor and a multi-dimensional sensor in a target transformer;
acquiring first data features corresponding to sample data of a one-dimensional sensor and second data features corresponding to sample data of a multi-dimensional sensor;
an attribute learner is acquired from the first data characteristic and the second data characteristic.
Further, the method for acquiring the second data features corresponding to the sample data of the multidimensional sensor comprises the following steps:
acquiring four video characteristic values of contrast, energy, entropy and relativity of images in sample data of a multidimensional sensor;
acquiring brightness, saturation and color characteristic values of an image in sample data of a multidimensional sensor;
and taking the obtained four video characteristic values, brightness, saturation and color characteristic values as second data characteristics.
Further, the method for obtaining the contrast, energy, entropy and correlation of the image in the sample data of the multidimensional sensor comprises a gray level co-occurrence matrix processing method, wherein the gray level co-occurrence matrix processing method processes the image according to the formula:
；
wherein,Conis the contrast of the image;
iis a pixel point on a gray level co-occurrence matrix row vector,jthe pixel points are pixel points on the gray level symbiotic matrix array vector;
P(i,j) Representing grey levelsiFor starting point, grey level occursjProbability of (2);
Asmis the energy of the image;
Ententropy of the image;
corris the correlation of the image;
μ x is the average value of the transverse vector,μ y is the mean value of the longitudinal vectors;
δ x is the standard deviation of the transverse vector,δ y is the standard deviation of the longitudinal vector.
Further, the method for obtaining the brightness, saturation and color characteristic values of the image in the sample data of the multidimensional sensor comprises an HSV color segmentation method, wherein the HSV color segmentation method processes the image according to the formula:
；
；
wherein,R、G、Brespectively representing three color channels of red, green and blue,R’、G’、B’respectively representing three color channels of red, green and blue after treatment;
Vrepresenting the brightness of the image, whenV≠ At 0:
；
Srepresenting the saturation of the image and,
when (when)V = R’ When (1):
；
when (when)V = G’ When (1):
；
when (when)V = B’ When (1):
；
Hrepresenting the color characteristic value of the image.
Further, the method for acquiring the attribute learner according to the first data feature and the second data feature comprises:
embedding semantics into the acquired first data features and second data features to generate fault description;
and training and classifying the attribute learner according to the generated fault description to obtain a trained attribute learner.
Further, the loss function for training the attribute learner is:
；
wherein,P l is the firstlClass sample data in total samplesLThe proportion of the components;
Uis the number of sample data;
nfor the number of branches of the decision tree,V m is the firstmThe number of branches of the data of a sample,I m is the firstmInformation entropy of the individual sample data branches;
Gfor information gain, whenGWhen the value falls to 0, training of the attribute learner is completed.
Further, after the training of the attribute learner is completed, the method further comprises:
acquiring attribute characteristics corresponding to the first data characteristics and the second data characteristics of the embedded semantics;
constructing a fault attribute matrix based on attribute features corresponding to the first data features and the second data features;
and acquiring the corresponding fault type according to the fault attribute matrix.
In a second aspect, a transformer zero sample fault diagnosis apparatus includes:
the first acquisition module is used for acquiring sample data of various sensors of the target transformer;
the second acquisition module is used for acquiring fault types based on a pre-trained attribute learner according to the acquired sample data of various sensors of the target transformer;
and the training module is used for training the attribute learner.
In a third aspect, a computer readable storage medium has stored thereon a computer program which, when executed by a processor, implements the above-described transformer zero sample fault diagnosis method.
In a fourth aspect, a computer device comprises:
a memory for storing instructions;
and the processor is used for executing the instructions to enable the computer equipment to execute the operation of realizing the transformer zero sample fault diagnosis method.
Compared with the prior art, the invention has the beneficial effects that:
1. according to the invention, through acquiring the first data features and the second data features corresponding to the sample data of various sensors of the target transformer and embedding semantics into the first data features and the second data features, multi-mode fusion is realized, a flexible online supervision mechanism of transformer fault diagnosis is achieved, and the problems of single target detection angle and low diagnosis precision in the existing method are solved;
2. according to the invention, the fault attribute matrix is constructed based on the attribute features corresponding to the first data feature and the second data feature, and the corresponding fault type is obtained according to the fault attribute matrix, so that fault diagnosis of a zero sample is realized, excessive dependence on fault sample data is avoided, and the same effect of fault diagnosis of the zero sample is good while the existing sample diagnosis efficiency is high.
Drawings
Fig. 1 is a flowchart of a transformer zero sample fault diagnosis method provided in embodiment 1 of the present invention;
fig. 2 is a training flowchart of an attribute learner of a transformer zero sample fault diagnosis method provided in embodiment 1 of the present invention;
fig. 3 is a schematic semantic embedding diagram of a transformer zero sample fault diagnosis method according to embodiment 1 of the present invention;
fig. 4 is a Fault3 confusion matrix based on KNN in the method for diagnosing zero sample faults of a transformer according to embodiment 1 of the present invention;
fig. 5 is a Fault3 confusion matrix based on MSVM in the method for diagnosing zero sample faults of a transformer according to embodiment 1 of the present invention;
fig. 6 is a Fault3 confusion matrix based on NB for a transformer zero sample Fault diagnosis method according to embodiment 1 of the present invention;
fig. 7 is a schematic diagram of the working principle of a transformer zero sample fault diagnosis method provided in embodiment 1 of the present invention.
Description of the embodiments
The invention is further described below with reference to the accompanying drawings. The following examples are only for more clearly illustrating the technical aspects of the present invention, and are not intended to limit the scope of the present invention.
Embodiment 1, as shown in fig. 1, a method for diagnosing zero sample faults of a transformer includes:
acquiring sample data of various sensors of a target transformer;
acquiring fault types based on a pre-trained attribute learner according to the acquired sample data of various sensors of the target transformer;
wherein training of the attribute learner includes:
as shown in fig. 2, sample data of a one-dimensional sensor and a multi-dimensional sensor in a target transformer are respectively acquired;
acquiring first data features corresponding to sample data of a one-dimensional sensor and second data features corresponding to sample data of a multi-dimensional sensor; specific:
the one-dimensional sensor comprises a temperature sensor, a humidity sensor, a vibration sensor, a current sensor, a partial discharge sensor, an audio sensor, a liquid leakage sensor and the like; the multidimensional sensor comprises a video sensor, an infrared thermal imaging sensor and the like;
processing sample data acquired by a one-dimensional sensor by adopting a traditional machine learning method to acquire first data characteristics;
extracting image features acquired by the multidimensional sensor, thereby obtaining the damage and degradation conditions of the appearance of the transformer, the sleeve, the oil and the like;
acquiring four video characteristic values of contrast, energy, entropy and relativity of images in sample data of a multidimensional sensor;
acquiring brightness, saturation and color characteristic values of an image in sample data of a multidimensional sensor;
the method for acquiring the contrast, energy, entropy and correlation four video characteristic values of the image in the sample data of the multidimensional sensor comprises a gray level co-occurrence matrix processing method, wherein the formula for processing the image by the gray level co-occurrence matrix processing method is as follows:
；
wherein,Conthe contrast of the image reflects the definition degree and the texture depth degree of the image, and the smaller the value is, the shallower the groove is;
iis a pixel point on a gray level co-occurrence matrix row vector,jthe pixel points are pixel points on the gray level symbiotic matrix array vector;
P(i，j) Representing grey levelsiFor starting point, grey level occursjProbability of (2); 1394 in the image training, 1955;
Asmthe gray level distribution uniformity degree and the texture thickness degree of the image are reflected as the energy of the image, and the value of the gray level distribution uniformity degree and the texture thickness degree are in inverse proportion to the uniformity degree;
Entthe entropy of the image reflects the complexity of gray level distribution of the image, and the entropy value and the complexity are in a direct proportion relation;
corrfor the correlation of the image, it measures the similarity of the gray level co-occurrence matrix elements in the row or column direction, the value of which reflectsLocal gray scale correlation degree in the image;
μ x 、μ y is the mean value;
δ x is the standard deviation of the transverse vector,δ y is the standard deviation of the longitudinal vector;
the method for acquiring the brightness, saturation and color characteristic values of the image in the sample data of the multidimensional sensor comprises an HSV color segmentation method, wherein the HSV color segmentation method processes the image according to the formula:
；
；
wherein,R、G、B、R’、G’、B’respectively representing color channels;
Vrepresenting the brightness of the image, whenV≠ 0, the following formula is given:
；
Srepresenting the saturation of the image and,
when (when)V = R’ The following formula is given:
；
when v=g'When (1):
；
when (when)V = B’ When (1):
；
Hcolor feature values representing images;
and taking the obtained four video characteristic values, brightness, saturation and color characteristic values as second data characteristics.
The multi-dimensional sensor also comprises a photoacoustic spectrum sensor, wherein the photoacoustic spectrum sensor uses an infrared wide-spectrum light source, infrared light modulated by wavelength passes through a rotary filter disc, and target gas is repeatedly excited at a certain specific frequency to generate vibration signals which can be captured by a photogenerated microphone, the concentrations of different gases are measured according to the characteristics of the vibration signals, the photoacoustic spectrum sensor is used for detecting the concentration of the gases in transformer oil, so that good sensitivity can be obtained, and the concentration data and the vibration signals of the gases are also used as second data characteristics.
Acquiring an attribute learner according to the first data characteristic and the second data characteristic, and specifically:
embedding semantics into the acquired first data features and second data features to generate fault description;
as shown in fig. 3 and table 1, the extracted first data features and second data features are semantically embedded, and mapped to an attribute feature space to realize multi-mode fusion, and 16 types of fault descriptions are generated by semantically embedding the first data features and the second data features;
TABLE 1 description of faults
。
Training and classifying the attribute learner according to the generated fault description to obtain a trained attribute learner, and specifically:
as shown in fig. 7, classification of attribute learning is performed on sample data based on semantics, and whether each fault type has an attribute feature of the sample data (correlation attribute is 1, irrelevant attribute is 0) is determined;
the training mode of the attribute learner adopts a classical ID3 algorithm of a decision tree, which constructs a top-down decision tree, calculates the maximum information entropy and the conditional entropy contained in the sample data in each layer of nodes, and calculates the information gain in the sample data according to the difference between the two entropy; when the information gain is reduced to 0, the training of the high-entropy data is completed, the decision tree is successfully constructed, namely, the attribute learning is completed, and the loss function of the training attribute learner is as follows:
；
wherein,P l is the firstlClass sample data in total samplesLThe proportion of the components;
Uis the number of sample data;
nfor the number of branches of the decision tree,V m is the firstmThe number of branches of the data of a sample,I m is the firstmInformation entropy of the individual sample data branches;
Gfor information gain, whenGWhen the value falls to 0, training of the attribute learner is completed.
As shown in table 2, fault training classification is performed on the attribute features learned by the attribute learner, and fault diagnosis accuracy of the multi-mode transformer based on data driving is often limited by quality and quantity of samples, so in this embodiment, three methods are used to perform simulation analysis on 288 groups of fault data of the transformer, the 288 groups of fault data are classified according to 80% as training set and 20% as testing level, 5 groups of data are randomly sampled for simulation experiment, and MSVM (multi-classification support vector machine), KNN (K nearest neighbor search) and NB algorithm (naive bayes) in machine learning are adopted to compare algorithm results; the three algorithms are integrated, the visible fault diagnosis rate of the embodiment is higher than 96.55%, and the invention has the advantages even if zero sample data are not processed, and the MSVM has higher stability and best comprehensive effect when processing the data; the diagnosis probability reaches 100% when the NB processes the first three groups of data, but the diagnosis effect of D, E groups deviates from the expected value; KNN is the worst of the above methods and the diagnostic rate for group D data only reaches 86.21%;
TABLE 2 failure training class diagnostic results
。
Acquiring attribute characteristics corresponding to the first data characteristics and the second data characteristics of the embedded semantics;
constructing a fault attribute matrix based on attribute features corresponding to the first data features and the second data features;
obtaining a corresponding fault type according to the fault attribute matrix, and specifically:
as shown in table 3, constructing a transformer knowledge graph based on expert experience, and deducing a fault attribute matrix by combining the embedded semantic attributes; the invention divides 12 fault types altogether, each fault corresponds to the number of data sets;
TABLE 3 failure type
。
As shown in table 4, combining table 1 and table 3, constructing a transformer fault attribute matrix in a fine granularity mode, respectively taking 12 fault types as invisible faults in sequence, performing attribute transfer learning on the invisible faults by using other 11 fault types, substituting the fault attribute matrix into a training set as expansion of training samples in the transfer learning process, and diagnosing the fault types as shown in table 5;
table 4 transformer fault attributes
；
TABLE 5 zero sample fault diagnosis results
。
The comprehensive diagnosis effect of the MSVM is still the best, the average fault diagnosis rate of the zero sample reaches 72.92%, and the KNN and NB respectively only have 52.08% and 60.01%. From the MSVM alone, the Fault detection rate except for Fault3 reaches 50% and above. Excluding the representative reasons of lack of Fault2 and Fault7 due to too few Fault sets, fault4, fault6, fault9, fault10, fault11 and Fault12 all perform well, and are all above 87.5%. From the perspective of faults, MSVM and NB with higher detection easiness degree of Fault10 reach 100%, and KNN also has 87.5%. The diagnostic rate of Fault3 is low, and the highest KNN algorithm only reaches 25% even 0% of NB algorithm;
referring to fig. 4, 5 and 6, it can be seen that the specific gravity of Fault5 and Fault7 is the largest in the misprediction classification; the problem that the fault description number is increased can be solved by increasing the sensors corresponding to the poor contact of the tap changer due to the fact that the three parts are close to each other in the vector space.
Embodiment 2, a transformer zero sample fault diagnosis device, comprising:
the first acquisition module is used for acquiring sample data of various sensors of the target transformer;
the second acquisition module is used for acquiring fault types based on a pre-trained attribute learner according to the acquired sample data of various sensors of the target transformer;
and the training module is used for training the attribute learner.
Embodiment 3, a computer readable storage medium having stored thereon a computer program which, when executed by a processor, performs the steps of the method of:
acquiring sample data of various sensors of a target transformer;
acquiring fault types based on a pre-trained attribute learner according to the acquired sample data of various sensors of the target transformer;
wherein training of the attribute learner includes:
respectively acquiring sample data of a one-dimensional sensor and a multi-dimensional sensor in a target transformer;
acquiring first data features corresponding to sample data of a one-dimensional sensor and second data features corresponding to sample data of a multi-dimensional sensor;
embedding semantics into the acquired first data features and second data features to generate fault description;
and training and classifying the attribute learner according to the generated fault description to obtain a trained attribute learner.
Embodiment 4, a computer device, comprising:
a memory for storing instructions;
a processor for executing instructions to cause a computer device to perform operations implementing the method of:
acquiring sample data of various sensors of a target transformer;
acquiring fault types based on a pre-trained attribute learner according to the acquired sample data of various sensors of the target transformer;
wherein training of the attribute learner includes:
respectively acquiring sample data of a one-dimensional sensor and a multi-dimensional sensor in a target transformer;
acquiring first data features corresponding to sample data of a one-dimensional sensor and second data features corresponding to sample data of a multi-dimensional sensor;
embedding semantics into the acquired first data features and second data features to generate fault description;
and training and classifying the attribute learner according to the generated fault description to obtain a trained attribute learner.
The foregoing is merely a preferred embodiment of the present invention, and it should be noted that modifications and variations could be made by those skilled in the art without departing from the technical principles of the present invention, and such modifications and variations should also be regarded as being within the scope of the invention.
Claims (10)
1. A method for diagnosing a zero sample fault of a transformer, comprising:
acquiring sample data of various sensors of a target transformer;
acquiring fault types based on a pre-trained attribute learner according to the acquired sample data of various sensors of the target transformer;
wherein the training of the attribute learner comprises:
respectively acquiring sample data of a one-dimensional sensor and a multi-dimensional sensor in a target transformer;
acquiring first data features corresponding to sample data of a one-dimensional sensor and second data features corresponding to sample data of a multi-dimensional sensor;
an attribute learner is acquired from the first data characteristic and the second data characteristic.
2. The method of claim 1, wherein the step of obtaining second data features corresponding to the sample data of the multi-dimensional sensor comprises:
acquiring four video characteristic values of contrast, energy, entropy and relativity of images in sample data of a multidimensional sensor;
acquiring brightness, saturation and color characteristic values of an image in sample data of a multidimensional sensor;
and taking the obtained four video characteristic values, brightness, saturation and color characteristic values as second data characteristics.
3. The method for diagnosing a zero sample fault of a transformer according to claim 2, wherein the method for obtaining four video feature values of contrast, energy, entropy and correlation of an image in sample data of a multidimensional sensor comprises a gray level co-occurrence matrix processing method, and a formula for processing the image by the gray level co-occurrence matrix processing method is as follows:
；
wherein,Conis the contrast of the image;
iis a pixel point on a gray level co-occurrence matrix row vector,jthe pixel points are pixel points on the gray level symbiotic matrix array vector;
P(i,j) Representing grey levelsiFor starting point, grey level occursjProbability of (2);
Asmis the energy of the image;
Ententropy of the image;
corris the correlation of the image;
μ x is the average value of the transverse vector,μ y is the mean value of the longitudinal vectors;
δ x is the standard deviation of the transverse vector,δ y is the standard deviation of the longitudinal vector.
4. The method for diagnosing a zero sample fault of a transformer according to claim 2, wherein the method for obtaining the brightness, saturation and color characteristic values of the image in the sample data of the multidimensional sensor comprises an HSV color segmentation method, and the formula for processing the image by the HSV color segmentation method is as follows:
；
；
wherein,R、G、Brespectively representing three color channels of red, green and blue,R’、G’、B’respectively representing three color channels of red, green and blue after treatment;
Vrepresenting the brightness of the image and,
when (when)V≠ At 0:
；
Srepresenting the saturation of the image and,
when (when)V = R’ When (1):
；
when (when)V = G’ When (1):
；
when (when)V= B’ When (1):
；
Hrepresenting the color characteristic value of the image.
5. The transformer zero sample fault diagnosis method according to claim 1, wherein the method of acquiring the attribute learner from the first data characteristic and the second data characteristic comprises:
embedding semantics into the acquired first data features and second data features to generate fault description;
and training and classifying the attribute learner according to the generated fault description to obtain a trained attribute learner.
6. The transformer zero sample fault diagnosis method according to claim 5, wherein the loss function for training the attribute learner is:
；
wherein,P l is the firstlClass sample data in total samplesLThe proportion of the components;
Uis the number of sample data;
nfor the number of branches of the decision tree,V m is the firstmThe number of branches of the data of a sample,I m is the firstmInformation entropy of the individual sample data branches;
Gfor information gain, whenGWhen the value falls to 0, training of the attribute learner is completed.
7. The transformer zero sample fault diagnosis method according to claim 5, further comprising, after the training of the attribute learner is completed:
acquiring attribute characteristics corresponding to the first data characteristics and the second data characteristics of the embedded semantics;
constructing a fault attribute matrix based on attribute features corresponding to the first data features and the second data features;
and acquiring the corresponding fault type according to the fault attribute matrix.
8. A transformer zero sample fault diagnosis apparatus, comprising:
the first acquisition module is used for acquiring sample data of various sensors of the target transformer;
the second acquisition module is used for acquiring fault types based on a pre-trained attribute learner according to the acquired sample data of various sensors of the target transformer;
and the training module is used for training the attribute learner.
9. A computer-readable storage medium, on which a computer program is stored, characterized in that the computer program, when executed by a processor, implements the transformer zero sample fault diagnosis method according to any one of claims 1 to 7.
10. A computer device, comprising:
a memory for storing instructions;
a processor configured to execute the instructions, so that the computer device performs an operation of implementing the transformer zero sample fault diagnosis method according to any one of claims 1 to 7.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202311606768.9A CN117312930A (en) | 2023-11-29 | 2023-11-29 | Transformer zero sample fault diagnosis method, device, storage medium and equipment |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202311606768.9A CN117312930A (en) | 2023-11-29 | 2023-11-29 | Transformer zero sample fault diagnosis method, device, storage medium and equipment |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN117312930A true CN117312930A (en) | 2023-12-29 |
Family
ID=89285099
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202311606768.9A Pending CN117312930A (en) | 2023-11-29 | 2023-11-29 | Transformer zero sample fault diagnosis method, device, storage medium and equipment |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN117312930A (en) |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN117725529A (en) * | 2024-02-18 | 2024-03-19 | 南京邮电大学 | Transformer fault diagnosis method based on multi-mode self-attention mechanism |
| CN117951574A (en) * | 2024-03-21 | 2024-04-30 | 北京航空航天大学 | Semantic embedding-based fault diagnosis method and system for aircraft power supply system |
Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN108389137A (en) * | 2018-02-06 | 2018-08-10 | 国网山西省电力公司电力科学研究院 | Power fault detection early warning system based on infared spectrum technology |
| CN110068741A (en) * | 2019-05-29 | 2019-07-30 | 国网河北省电力有限公司石家庄供电分公司 | A method of the transformer fault diagnosis based on categorised decision tree |
| CN113378943A (en) * | 2021-06-16 | 2021-09-10 | 西北工业大学 | Engine rotor rubbing fault diagnosis method based on wavelet-gray level co-occurrence matrix |
| CN113496089A (en) * | 2021-09-07 | 2021-10-12 | 江苏宝诺铸造有限公司 | Internet of things equipment abnormal state online monitoring method and system based on artificial intelligence |
| CN116402825A (en) * | 2023-06-09 | 2023-07-07 | 华东交通大学 | Bearing fault infrared diagnosis method, system, electronic equipment and storage medium |
| CN117075582A (en) * | 2023-09-08 | 2023-11-17 | 兰州理工大学 | Industrial process generalized zero sample fault diagnosis method based on DSECMR-VAE |
-
2023
- 2023-11-29 CN CN202311606768.9A patent/CN117312930A/en active Pending
Patent Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN108389137A (en) * | 2018-02-06 | 2018-08-10 | 国网山西省电力公司电力科学研究院 | Power fault detection early warning system based on infared spectrum technology |
| CN110068741A (en) * | 2019-05-29 | 2019-07-30 | 国网河北省电力有限公司石家庄供电分公司 | A method of the transformer fault diagnosis based on categorised decision tree |
| CN113378943A (en) * | 2021-06-16 | 2021-09-10 | 西北工业大学 | Engine rotor rubbing fault diagnosis method based on wavelet-gray level co-occurrence matrix |
| CN113496089A (en) * | 2021-09-07 | 2021-10-12 | 江苏宝诺铸造有限公司 | Internet of things equipment abnormal state online monitoring method and system based on artificial intelligence |
| CN116402825A (en) * | 2023-06-09 | 2023-07-07 | 华东交通大学 | Bearing fault infrared diagnosis method, system, electronic equipment and storage medium |
| CN117075582A (en) * | 2023-09-08 | 2023-11-17 | 兰州理工大学 | Industrial process generalized zero sample fault diagnosis method based on DSECMR-VAE |
Cited By (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN117725529A (en) * | 2024-02-18 | 2024-03-19 | 南京邮电大学 | Transformer fault diagnosis method based on multi-mode self-attention mechanism |
| CN117725529B (en) * | 2024-02-18 | 2024-05-24 | 南京邮电大学 | Transformer fault diagnosis method based on multi-mode self-attention mechanism |
| CN117951574A (en) * | 2024-03-21 | 2024-04-30 | 北京航空航天大学 | Semantic embedding-based fault diagnosis method and system for aircraft power supply system |
| CN117951574B (en) * | 2024-03-21 | 2024-05-28 | 北京航空航天大学 | Semantic embedding-based fault diagnosis method and system for aircraft power supply system |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN117312930A (en) | Transformer zero sample fault diagnosis method, device, storage medium and equipment | |
| Rudolph et al. | Same same but differnet: Semi-supervised defect detection with normalizing flows | |
| CN108256482B (en) | Face age estimation method for distributed learning based on convolutional neural network | |
| WO2019015246A1 (en) | Image feature acquisition | |
| CN112580590A (en) | Finger vein identification method based on multi-semantic feature fusion network | |
| CN111222457B (en) | Detection method for identifying authenticity of video based on depth separable convolution | |
| CN111860106A (en) | Unsupervised bridge crack identification method | |
| CN111027636A (en) | Unsupervised feature selection method and system based on multi-label learning | |
| CN111598167A (en) | Small sample image identification method and system based on graph learning | |
| Patel | Bacterial colony classification using atrous convolution with transfer learning | |
| Qiao et al. | A multi-level thresholding image segmentation method using hybrid Arithmetic Optimization and Harris Hawks Optimizer algorithms | |
| CN110717602A (en) | Machine learning model robustness assessment method based on noise data | |
| Yao et al. | Learning global-local correspondence with semantic bottleneck for logical anomaly detection | |
| CN112633301A (en) | Traditional Chinese medicine tongue image greasy feature classification method based on depth metric learning | |
| CN112149804A (en) | Novel convolutional neural network model and application | |
| CN110750672B (en) | Image retrieval method based on deep measurement learning and structure distribution learning loss | |
| Liu et al. | Automatic and quantitative evaluation of attribute discovery methods | |
| Torop et al. | Unsupervised approaches for out-of-distribution dermoscopic lesion detection | |
| CN115984546B (en) | Sample base generation method for anomaly detection of fixed scene | |
| Mohammadi et al. | Odd-one-out representation learning | |
| CN108764367A (en) | A kind of characteristic image extraction element and extracting method based on relationship regularization | |
| Shen et al. | Head pose estimation in classroom scenes | |
| Farhansyah et al. | Revolutionizing Banana Grading with ResNeXt and SVM: An Automated Approach | |
| CN115393362B (en) | Method, equipment and medium for selecting automatic glaucoma recognition model | |
| Zheng et al. | Cross-modal knowledge distillation for fault detection under multiple failure modes |
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 |