CN116049638A - Transformer vibration detection method, system, equipment and storage medium - Google Patents
Transformer vibration detection method, system, equipment and storage medium Download PDFInfo
- Publication number
- CN116049638A CN116049638A CN202310188265.8A CN202310188265A CN116049638A CN 116049638 A CN116049638 A CN 116049638A CN 202310188265 A CN202310188265 A CN 202310188265A CN 116049638 A CN116049638 A CN 116049638A
- Authority
- CN
- China
- Prior art keywords
- fault
- matrix
- attention network
- gray scale
- transformer
- 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
Images
Classifications
-
- 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
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- Mathematical Physics (AREA)
- Theoretical Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Data Mining & Analysis (AREA)
- Software Systems (AREA)
- General Engineering & Computer Science (AREA)
- Pure & Applied Mathematics (AREA)
- Mathematical Optimization (AREA)
- Mathematical Analysis (AREA)
- Computational Mathematics (AREA)
- Computing Systems (AREA)
- Health & Medical Sciences (AREA)
- Databases & Information Systems (AREA)
- Algebra (AREA)
- Life Sciences & Earth Sciences (AREA)
- Artificial Intelligence (AREA)
- Biomedical Technology (AREA)
- Biophysics (AREA)
- Computational Linguistics (AREA)
- Evolutionary Computation (AREA)
- General Health & Medical Sciences (AREA)
- Molecular Biology (AREA)
- Complex Calculations (AREA)
Abstract
The application discloses a transformer vibration detection method, a system, equipment and a storage medium, wherein the first aspect provides a global-local cross comparison attention network to enhance interaction between a global image and a local highlight region; the second aspect proposes a fault-normal cross comparison attention network to establish a comparison between a fault image and a normal image, further determine local features of the fault image, find more complementary parts for identification and more effective assessment of layering and classification of faults with smaller granularity; and thirdly, carrying out off-line training on the dual-cross comparison attention network, further optimizing the attention network, simultaneously accelerating the operation speed of on-line fault detection, finally carrying out short-time fault prediction on the image without fault detection, and continuously detecting the data image after fault prediction. The problem of prior art vibrations fault detection and location rate of accuracy low, and can't carry out the early warning to the trouble is solved.
Description
Technical Field
The present disclosure relates to the field of power technologies, and in particular, to a method, a system, an apparatus, and a storage medium for detecting vibration of a transformer.
Background
Along with the proposal of the three concepts of 'intellectualization', 'stabilization', 'safety' of the power system, the automation machine room has been promoted and operated in a large area. With the development of technology, devices such as sensors and the like are arranged and used in an automation machine room in a large quantity, the level and the precision of data acquisition are greatly improved, but the pluripotency and the isomerism of data are multiplied, and the vibration fault detection method and the fault positioning method of the traditional transformer are difficult to show satisfactory precision and timeliness under the background of the large data.
At present, three main types of vibration fault detection methods for power transformers are provided: the first is to construct an accurate mathematical physical model based on an analytical model to diagnose the transformer. The second category is to collect imperfect fault experience, build mathematical qualitative model, and deduce fault category through model, typically with artificial decision tree method. The third category is a data-driven-based intelligent classification method for power transformer faults, such as a support vector machine, a convolutional neural network and the like.
However, the above detection method has some problems. The first type of method is limited by the accuracy of mathematical models, and is not sensitive enough to the reaction of external factors such as temperature and illumination, and has low positioning accuracy. The second type of method is based on the past fault data, and the lack of timeliness can not update the fault library in real time, so that various novel vibration faults in an intelligent environment are difficult to identify. The third type of method has inherent defects, such as high requirement on the quality of a learning sample by a neural network, and is easy to fall into local optimum; the support vector machine has outstanding performance in processing small sample data, is a classifier in nature, and is inefficient in processing multi-classification problems.
Disclosure of Invention
The application provides a transformer vibration detection method, a system, equipment and a storage medium, which are used for solving the technical problems that the vibration fault detection and positioning accuracy rate is low and the fault cannot be early-warned in the prior art.
In view of this, a first aspect of the present application provides a method for detecting vibration of a transformer, the method comprising:
s1, collecting and preprocessing operation signals of a transformer, wherein the operation signals comprise: coil vibration frequency signals of the transformer, temperature signals of the coil and the box body, and brightness signals around the switch and the coil are acquired by infrared shooting;
s2, converting the operation signal into a two-dimensional data matrix, and carrying out normalization processing on the two-dimensional data matrix to obtain a two-dimensional gray scale map;
s3, inputting a two-dimensional gray level diagram of a time period to be detected into a trained double-cross contrast attention network, and outputting a first fault detection result of a transformer, wherein the double-cross contrast attention network is obtained by training a historical two-dimensional gray level diagram, and the double-cross contrast attention network comprises: global-local cross-contrast attention network and fault-normal cross-contrast attention network;
s4, determining a two-dimensional gray scale map corresponding to the transformer which does not generate faults according to the first fault detection result as a fault two-dimensional gray scale map, inputting the fault two-dimensional gray scale map into a trained short-term fault data prediction model, outputting a two-dimensional gray scale map of a prediction time period, inputting the two-dimensional gray scale map into the double cross contrast attention network, and outputting a second fault detection result of the transformer.
Optionally, the training process of the global-local cross contrast attention network specifically includes:
regarding the two-dimensional gray scale map obtained in the step S3 as a query matrix R= [ R ] formed by N query vectors 1 ；r 2 ；...；r N ]Calculating a cumulative attention score S for an ith image block in accordance with an attention display i ；
According to the cumulative attention score S i From R i The first T query vectors corresponding to the first T highest responses in the accumulated weights of the CLS are selected to construct a new query matrix R l Representing the most interesting local embedding;
calculating cross-concerns between the selected local query and global key-value pairs based on the output function;
wherein the output function is:
in the method, in the process of the invention,
as a scale factor, the query matrix Q, key matrix K and value matrix V are embedded from the same input with different linear transformations, respectively N×D To calculate q=xw Q ,K＝XW K ，V＝XW V L and g are coefficients, and a new matrix F is output G Namely, a fault feature matrix;
according to the fault characteristic matrix F G And determining the fault characteristic weight by using GRA to obtain a fault weight matrix, and taking the fault weight matrix as an output matrix of the global-local cross contrast attention network.
Optionally, the training process of the fault-normal cross contrast attention network specifically includes:
constructing a fault-normal cross contrast attention network, and inputting the fault two-dimensional gray scale map and the normal two-dimensional gray scale map into the fault-normal cross contrast attention network in sequence for training;
calculating a normal data matrix and a fault characteristic weight matrix after the comparison of the data matrix containing faults according to the fault two-dimensional gray level diagram and the normal two-dimensional gray level diagram based on the output function, and taking the fault characteristic matrix and the fault characteristic weight matrix as an output matrix of the fault-normal cross comparison attention network;
combining the output matrixes of the fault-normal cross contrast attention network and the global-local cross contrast attention network through a contribution rate function to obtain a fault characteristic matrix and a fault characteristic weight matrix;
and layering the classified faults according to the fault characteristic weight matrix to obtain a fault classification layered characteristic set, and improving and iterating the contrast attention network parameters by the characteristic set trained by the offline network to obtain the dual cross contrast attention network.
Optionally, the normalizing the two-dimensional matrix to obtain a two-dimensional gray scale map specifically includes:
based on a normalization formula, carrying out normalization processing on the two-dimensional data matrix to obtain a two-dimensional gray scale map;
wherein, the normalization formula is:
wherein L (i) is the two-dimensional data matrix, where i=1, 2 … …, n×m; max (L) represents the maximum value in the two-dimensional gray scale map; min (L) represents the minimum value in the two-bit gray scale map; r (N, M) (n=1, 2 … …, N, m=1, 2 … …, M) represents the pixel intensity of the image normalized; the circle function is a normalization function.
A second aspect of the present application provides a transformer vibration detection system, the system comprising:
the collection unit is used for collecting operation signals of the transformer and preprocessing the operation signals, and the operation signals comprise: coil vibration frequency signals of the transformer, temperature signals of the coil and the box body, and brightness signals around the switch and the coil are acquired by infrared shooting;
the normalization unit is used for converting the operation signals into a two-dimensional data matrix, and then normalizing the two-dimensional data matrix to obtain a two-dimensional gray level map;
the first detection unit is used for inputting the two-dimensional gray level diagram of the time period to be detected into a trained dual cross contrast attention network, and outputting a first fault detection result of the transformer, wherein the dual cross contrast attention network is obtained by training a historical two-dimensional gray level diagram, and the dual cross contrast attention network comprises: global-local cross-contrast attention network and fault-normal cross-contrast attention network;
the second detection unit is used for determining a two-dimensional gray scale image corresponding to the transformer which does not generate faults according to the first fault detection result as a fault two-dimensional gray scale image, inputting the fault two-dimensional gray scale image into a trained short-term fault data prediction model, outputting a two-dimensional gray scale image of a prediction time period, inputting the two-dimensional gray scale image into the double cross contrast attention network, and outputting a second fault detection result of the transformer.
Optionally, the training process of the global-local cross contrast attention network specifically includes:
regarding the two-dimensional gray scale map obtained in the step S3 as a query matrix R= [ R ] formed by N query vectors 1 ；r 2 ；…；r N ]Calculating a cumulative attention score S for an ith image block in accordance with an attention display i ；
According to the cumulative attention score S i From R i The first T query vectors corresponding to the first T highest responses in the accumulated weights of the CLS are selected to construct a new query matrix R l Representing the most interesting local embedding;
calculating cross-concerns between the selected local query and global key-value pairs based on the output function;
wherein the output function is:
in the method, in the process of the invention,
as a scale factor, the query matrix Q, key matrix K and value matrix V are embedded from the same input with different linear transformations, respectively N×D To calculate q=xw Q ,K＝XW K ，V＝XW V L and g are coefficients, and a new matrix F is output G Namely, a fault feature matrix;
according to the fault characteristic matrix F G And determining the fault characteristic weight by using GRA to obtain a fault weight matrix, and taking the fault weight matrix as an output matrix of the global-local cross contrast attention network.
Optionally, the training process of the fault-normal cross contrast attention network specifically includes:
constructing a fault-normal cross contrast attention network, and inputting the fault two-dimensional gray scale map and the normal two-dimensional gray scale map into the fault-normal cross contrast attention network in sequence for training;
calculating a normal data matrix and a fault characteristic weight matrix after the comparison of the data matrix containing faults according to the fault two-dimensional gray level diagram and the normal two-dimensional gray level diagram based on the output function, and taking the fault characteristic matrix and the fault characteristic weight matrix as an output matrix of the fault-normal cross comparison attention network;
combining the output matrixes of the fault-normal cross contrast attention network and the global-local cross contrast attention network through a contribution rate function to obtain a fault characteristic matrix and a fault characteristic weight matrix;
and layering the classified faults according to the fault characteristic weight matrix to obtain a fault classification layered characteristic set, and improving and iterating the contrast attention network parameters by the characteristic set trained by the offline network to obtain the dual cross contrast attention network.
Optionally, the normalizing the two-dimensional matrix to obtain a two-dimensional gray scale map specifically includes:
based on a normalization formula, carrying out normalization processing on the two-dimensional data matrix to obtain a two-dimensional gray scale map;
wherein, the normalization formula is:
wherein L (i) is the two-dimensional data matrix, where i=1, 2 … …, n×m; max (L) represents the maximum value in the two-dimensional gray scale map; min (L) represents the minimum value in the two-bit gray scale map; r (N, M) (n=1, 2 … …, N, m=1, 2 … …, M) represents the pixel intensity of the image normalized; the circle function is a normalization function.
A third aspect of the present application provides a transformer vibration detection apparatus, the apparatus comprising a processor and a memory:
the memory is used for storing program codes and transmitting the program codes to the processor;
the processor is configured to execute the steps of the method for detecting a vibration of a transformer according to the first aspect according to the instructions in the program code.
A fourth aspect of the present application provides a computer-readable storage medium for storing program code for executing the transformer vibration detection method of the first aspect described above.
From the above technical scheme, the application has the following advantages:
the application provides a transformer vibration detection method, which comprises the following steps: s1, collecting and preprocessing operation signals of a transformer, wherein the operation signals comprise: coil vibration frequency signals of the transformer, temperature signals of the coil and the box body, and brightness signals around the switch and the coil are acquired by infrared shooting; s2, converting the operation signal into a two-dimensional data matrix, and carrying out normalization processing on the two-dimensional data matrix to obtain a two-dimensional gray scale map; s3, inputting a two-dimensional gray level diagram of a time period to be detected into a trained double-cross contrast attention network, outputting a first fault detection result of a transformer, wherein the double-cross contrast attention network is obtained by training a historical two-dimensional gray level diagram, and the double-cross contrast attention network comprises: global-local cross-contrast attention network and fault-normal cross-contrast attention network; s4, determining a two-dimensional gray scale image corresponding to the transformer which does not generate faults according to the first fault detection result as a fault two-dimensional gray scale image, inputting the fault two-dimensional gray scale image into a trained short-term fault data prediction model, outputting a two-dimensional gray scale image of a prediction time period, inputting the two-dimensional gray scale image into a double cross contrast attention network, and outputting a second fault detection result of the transformer.
1. Compared with the traditional transformer diagnosis method for constructing the mathematical model, the method can comprehensively consider the influence of external environmental factors by establishing the fault feature set and the fault feature weight matrix, is sensitive to the related environmental factors, and avoids fault detection errors in extreme environments.
2. Compared with a fault positioning method based on a convolutional neural network, the method for offline training of the positioning model is adopted, so that time spent on vibration fault detection is greatly reduced, and accurate recognition and positioning of vibration fault positions are realized.
Drawings
Fig. 1 is a schematic flow chart of an embodiment of a method for detecting vibration of a transformer according to an embodiment of the present application;
fig. 2 is a schematic structural diagram of an embodiment of a transformer vibration detection system provided in the embodiments of the present application;
fig. 3 is a schematic diagram of a process of performing normalization processing on a converted two-dimensional matrix to convert the two-dimensional matrix into a two-dimensional gray scale map according to an embodiment of the present application;
FIG. 4 is a schematic diagram of a portion of an offline training model provided in an embodiment of the present application;
FIG. 5 is a schematic diagram of a global-local cross-contrast attention network training process provided in an embodiment of the present application;
FIG. 6 is a schematic diagram of a fault-normal cross-contrast attention network training process provided in an embodiment of the present application;
fig. 7 is a schematic diagram of a network structure of a short-term failure data prediction model based on a multi-layer neural network according to an embodiment of the present application.
Detailed Description
In order to make the present application solution better understood by those skilled in the art, the following description will clearly and completely describe the technical solution in the embodiments of the present application with reference to the accompanying drawings in the embodiments of the present application, and it is apparent that the described embodiments are only some embodiments of the present application, not all embodiments. All other embodiments, which can be made by one of ordinary skill in the art without undue burden from the present disclosure, are within the scope of the present disclosure.
Referring to fig. 1, an embodiment of the present application provides a method for detecting vibration of a transformer, including:
in the application, the multi-mode sensor device, the signal amplifier, the signal processing device and the fault detection and positioning device are arranged; the fault prediction device and the early warning device;
the multi-modal sensor arrangement is placed in different positions of the transformer tank under test. Wherein the vibration sensor is arranged around the transformer coil and used for detecting vibration frequency information of the transformer coil; the temperature sensor is arranged on a coil and a box body of the transformer and is used for detecting temperature data of key devices of the transformer; an infrared camera is mainly placed around the switch and coil of the transformer for measuring temperature data of the core area.
The collected signals are stored according to the time sequence, and the signals amplified in the time period are output through the signal amplifier according to the requirement, so that the next processing is facilitated.
sampling the amplified signals in the time period, and respectively obtaining N equally-spaced signal sampling points with the same length on M sensor signals, wherein the vertical axis of the sampling points is frequency, temperature or brightness;
the data N of the sampling points and the sequence M of the sensors are respectively used as the abscissa and the ordinate of the matrix, and the sampled data are filled into the matrix to form three N multiplied by M two-dimensional data matrixes;
the converted two-dimensional matrix is normalized to be converted into a two-dimensional gray scale map, as shown in fig. 3.
Where L (i) represents a transformed two-dimensional matrix, where i=1, 2 … …, n×m; max (L) represents the maximum value in the two-bit gray scale map; min (L) represents the minimum value in the two-bit gray scale map; r (N, M) (n=1, 2 … …, N, m=1, 2 … …, M) represents the pixel intensity of the image normalized; the circle function is a normalization function that ensures that the normalized pixel intensities are in the interval 0-255.
it should be noted that, the dual cross contrast attention network is divided into an offline training model and an online detection, which are as follows;
offline training model section, as in fig. 4:
the fault data obtained off-line are input into the global-local cross contrast attention network after being processed by the steps 101-102, and meanwhile, the fault data and the normal data which are off-line are input into the fault-normal cross contrast attention network after being processed by the steps 101-102. And obtaining the characteristic matrix of the fault classification hierarchy more comprehensively after the two kinds of network cross training.
Specifically, the global-local cross contrast attention network training process is as follows, as in fig. 5:
1) The matrix R= [ R ] formed by N query vectors in the processed two-dimensional gray scale map 1 ；r 2 ；…；r N ]Calculating a cumulative attention score for the ith image block according to the attention display:
wherein the method comprises the steps of
Representing re-normalized attention weights taking into account the remaining connections using identity matrix E, +.>
Representing the cumulative weight of the class embedded in the CLS. />
Representing a matrix multiplication operation. In this way we keep track of information that propagates from the input layer to higher layers.
2) From R i Corresponding to the first T highest responses in the cumulative weight of the CLSThe first T query vectors to construct a new query matrix R l Representing the most interesting local embedding.
3) The cross-interests between the selected local query and global key-value pairs are calculated as follows:
wherein the method comprises the steps of
Is a scale factor. Embedding X E R from the same input by different linear transformations for query matrix, key matrix and value matrix, respectively N×D To calculate q=xw Q ,K＝XW K ，V＝XW V 。S∈R N×N Representing the attention weight matrix, the new matrix F is output G And the fault characteristic matrix is obtained.
4) Based on the principal component of the fault signature and the pre-fault signature matrix F G And determining the fault characteristic weight by using GRA to obtain a fault weight matrix.
Further, the fault-normal cross contrast attention network training procedure is as follows, as shown in fig. 6:
5) And inputting the two-dimensional gray level map of the fault and the normal two-dimensional gray level map into a fault-normal cross contrast attention network in sequence for training.
6) Both graphs are processed in steps S41-S44, and the output function is modified as follows:
and obtaining a fault characteristic matrix and a fault characteristic weight matrix after the normal data matrix and the data matrix containing faults are compared.
7) The fault-normal cross-contrast attention network is focused on the cross-contrast of the highlight and local areas in the fault map. Focusing on the difference between the fault diagram and the normal diagram, combining the output matrixes of the two networks together through a contribution rate function to obtain an accurate fault characteristic matrix and a fault characteristic weight matrix. Fault-normal cross contrast attention network as follows
8) And classifying the detected faults according to the characteristics (frequency, temperature and brightness) of the faults, and layering the classified faults according to a fault characteristic weight matrix to obtain a fault classification layering characteristic set. And simultaneously improving the attention network parameters by the feature set trained by the offline network, and stopping iteration when the error is smaller than a specified value so as to obtain the optimal dual cross contrast attention network model.
On-line detection part:
9) And if a fault exists, flashing a red alarm lamp through an early warning device, and positioning a sensor at the position, namely the position of the fault. If no fault is detected, the next fault prediction is performed.
It should be noted that, further, the step S5 mainly includes 5 steps:
1) Constructing a network structure of a short-term fault data prediction model based on a multi-layer neural network by using a fault feature matrix, as shown in fig. 7;
2) Sequentially dividing a data set of the fault characteristic matrix and the true value vector into a training set, a verification set and a test set according to the proportion of 6:3:1;
3) Performing parameter adjustment on the fault data prediction model based on the multi-layer neural network by using a training set and a verification set to obtain an optimized model SPLF-MNN;
4) And calculating the prediction precision of the SPLF-MNN model by using the test set to evaluate the performance of the model, and predicting the related data of one day or even one week in the future by using the SPLF-MNN model after the two-dimensional gray level diagram after fault detection meets the requirements.
5) And detecting the predicted data in the constructed double cross contrast attention network, if a fault exists, flashing a yellow warning lamp through a warning device, and positioning a sensor at the position, namely the position of the fault. If no fault is detected, the green safety lamp is turned on for a long time.
The foregoing is a method for detecting vibration of a transformer provided in an embodiment of the present application, and the following is a system for detecting vibration of a transformer provided in an embodiment of the present application.
Referring to fig. 2, a transformer vibration detection system is provided in an embodiment of the present application. Comprising the following steps:
the acquisition unit 201 is configured to acquire and pre-process an operation signal of the transformer, where the operation signal includes: coil vibration frequency signals of the transformer, temperature signals of the coil and the box body, and brightness signals around the switch and the coil are acquired by infrared shooting;
the normalization unit 202 is configured to convert the operation signal into a two-dimensional data matrix, and perform normalization processing on the two-dimensional data matrix to obtain a two-dimensional gray scale map;
the first detecting unit 203 is configured to input a two-dimensional gray scale map of a time period to be detected into a trained dual cross contrast attention network, and output a first fault detection result of the transformer, where the dual cross contrast attention network is obtained by training a historical two-dimensional gray scale map, and the dual cross contrast attention network includes: global-local cross-contrast attention network and fault-normal cross-contrast attention network;
the second detection unit 204 is configured to determine, according to the first fault detection result, a two-dimensional gray scale map corresponding to the transformer that has not failed as a fault two-dimensional gray scale map, input the fault two-dimensional gray scale map into a trained short-term fault data prediction model, output a two-dimensional gray scale map of a prediction time period, input the two-dimensional gray scale map into a dual cross contrast attention network, and output a second fault detection result of the transformer.
Further, in an embodiment of the present application, there is provided a transformer vibration detection apparatus, which is characterized in that the apparatus includes a processor and a memory:
the memory is used for storing program codes and transmitting the program codes to the processor;
the processor is configured to execute the transformer vibration detection method according to the above method embodiment according to the instructions in the program code.
Further, there is provided a computer readable storage medium in an embodiment of the present application, where the computer readable storage medium is configured to store a program code, where the program code is configured to execute the method for detecting a transformer vibration according to the embodiment of the method described above.
It will be clear to those skilled in the art that, for convenience and brevity of description, the specific working procedures of the above-described system and unit may refer to the corresponding procedures in the foregoing method embodiments, which are not repeated here.
The terms "first," "second," "third," "fourth," and the like in the description of the present application and in the above-described figures, if any, are used for distinguishing between similar objects and not necessarily for describing a particular sequential or chronological order. It is to be understood that the data so used may be interchanged where appropriate such that embodiments of the present application described herein may be capable of operation in sequences other than those illustrated or described herein, for example. Furthermore, the terms "comprises," "comprising," and "having," and any variations thereof, are intended to cover a non-exclusive inclusion, such that a process, method, system, article, or apparatus that comprises a list of steps or elements is not necessarily limited to those steps or elements expressly listed but may include other steps or elements not expressly listed or inherent to such process, method, article, or apparatus.
It should be understood that in this application, "at least one" means one or more, and "a plurality" means two or more. "and/or" for describing the association relationship of the association object, the representation may have three relationships, for example, "a and/or B" may represent: only a, only B and both a and B are present, wherein a, B may be singular or plural. The character "/" generally indicates that the context-dependent object is an "or" relationship. "at least one of" or the like means any combination of these items, including any combination of single item(s) or plural items(s). For example, at least one (one) of a, b or c may represent: a, b, c, "a and b", "a and c", "b and c", or "a and b and c", wherein a, b, c may be single or plural.
In the several embodiments provided in this application, it should be understood that the disclosed systems, apparatuses, and methods may be implemented in other ways. For example, the apparatus embodiments described above are merely illustrative, e.g., the division of the units is merely a logical function division, and there may be additional divisions when actually implemented, e.g., multiple units or components may be combined or integrated into another system, or some features may be omitted or not performed. Alternatively, the coupling or direct coupling or communication connection shown or discussed with each other may be an indirect coupling or communication connection via some interfaces, devices or units, which may be in electrical, mechanical or other form.
The units described as separate units may or may not be physically separate, and units shown as units may or may not be physical units, may be located in one place, or may be distributed on a plurality of network units. Some or all of the units may be selected according to actual needs to achieve the purpose of the solution of this embodiment.
In addition, each functional unit in each embodiment of the present application may be integrated in one processing unit, or each unit may exist alone physically, or two or more units may be integrated in one unit. The integrated units may be implemented in hardware or in software functional units.
The integrated units, if implemented in the form of software functional units and sold or used as stand-alone products, may be stored in a computer readable storage medium. Based on such understanding, the technical solution of the present application may be embodied in essence or a part contributing to the prior art or all or part of the technical solution in the form of a software product stored in a storage medium, including several instructions to cause a computer device (which may be a personal computer, a server, or a network device, etc.) to perform all or part of the steps of the methods described in the embodiments of the present application. And the aforementioned storage medium includes: u disk, mobile hard disk, read-Only Memory (ROM), random access Memory (Random Access Memory, RAM), magnetic disk or optical disk, etc.
The above embodiments are merely for illustrating the technical solution of the present application, and not for limiting the same; although the present application has been described in detail with reference to the foregoing embodiments, it should be understood by those of ordinary skill in the art that: the technical scheme described in the foregoing embodiments can be modified or some technical features thereof can be replaced by equivalents; such modifications and substitutions do not depart from the spirit and scope of the corresponding technical solutions.
Claims (10)
1. A method for detecting vibration of a transformer, comprising:
s1, collecting and preprocessing operation signals of a transformer, wherein the operation signals comprise: coil vibration frequency signals of the transformer, temperature signals of the coil and the box body, and brightness signals around the switch and the coil are acquired by infrared shooting;
s2, converting the operation signal into a two-dimensional data matrix, and carrying out normalization processing on the two-dimensional data matrix to obtain a two-dimensional gray scale map;
s3, inputting a two-dimensional gray level diagram of a time period to be detected into a trained double-cross contrast attention network, and outputting a first fault detection result of a transformer, wherein the double-cross contrast attention network is obtained by training a historical two-dimensional gray level diagram, and the double-cross contrast attention network comprises: global-local cross-contrast attention network and fault-normal cross-contrast attention network;
s4, determining a two-dimensional gray scale map corresponding to the transformer which does not generate faults according to the first fault detection result as a fault two-dimensional gray scale map, inputting the fault two-dimensional gray scale map into a trained short-term fault data prediction model, outputting a two-dimensional gray scale map of a prediction time period, inputting the two-dimensional gray scale map into the double cross contrast attention network, and outputting a second fault detection result of the transformer.
2. The method for detecting the vibration of the transformer according to claim 1, wherein the training process of the global-local cross contrast attention network specifically comprises:
regarding the two-dimensional gray scale map obtained in the step S3 as a query matrix R= [ R ] formed by N query vectors 1 ；r 2 ；…；r N ]Calculating a cumulative attention score S for an ith image block in accordance with an attention display i ；
According to the cumulative attention score S i From R i The first T query vectors corresponding to the first T highest responses in the accumulated weights of the CLS are selected to construct a new query matrix R l Representing the most interesting local embedding;
calculating cross-concerns between the selected local query and global key-value pairs based on the output function;
wherein the output function is:
in the method, in the process of the invention,
as a scale factor, the query matrix Q, key matrix K and value matrix V are embedded from the same input with different linear transformations, respectively N×D To calculate q=xw Q ,K＝XW K ，V＝XW V L and g are coefficients, and a new matrix F is output G Namely, a fault feature matrix;
according to the fault characteristic matrix F G And determining the fault characteristic weight by using GRA to obtain a fault weight matrix, and taking the fault weight matrix as an output matrix of the global-local cross contrast attention network.
3. The method for detecting the vibration of the transformer according to claim 2, wherein the training process of the fault-normal cross contrast attention network specifically comprises:
constructing a fault-normal cross contrast attention network, and inputting the fault two-dimensional gray scale map and the normal two-dimensional gray scale map into the fault-normal cross contrast attention network in sequence for training;
calculating a normal data matrix and a fault characteristic weight matrix after the comparison of the data matrix containing faults according to the fault two-dimensional gray level diagram and the normal two-dimensional gray level diagram based on the output function, and taking the fault characteristic matrix and the fault characteristic weight matrix as an output matrix of the fault-normal cross comparison attention network;
combining the output matrixes of the fault-normal cross contrast attention network and the global-local cross contrast attention network through a contribution rate function to obtain a fault characteristic matrix and a fault characteristic weight matrix;
and layering the classified faults according to the fault characteristic weight matrix to obtain a fault classification layered characteristic set, and improving and iterating the contrast attention network parameters by the characteristic set trained by the offline network to obtain the dual cross contrast attention network.
4. The method for detecting vibration of a transformer according to claim 1, wherein the normalizing the two-dimensional matrix to obtain a two-dimensional gray scale map specifically comprises:
based on a normalization formula, carrying out normalization processing on the two-dimensional data matrix to obtain a two-dimensional gray scale map;
wherein, the normalization formula is:
wherein L (i) is the two-dimensional data matrix, where i=1, 2 … …, n×m; max (L) represents the maximum value in the two-dimensional gray scale map; min (L) represents the minimum value in the two-bit gray scale map; r (N, M) (n=1, 2 … …, N, m=1, 2 … …, M) represents the pixel intensity of the image normalized; the circle function is a normalization function.
5. A transformer vibration detection system, comprising:
the collection unit is used for collecting operation signals of the transformer and preprocessing the operation signals, and the operation signals comprise: coil vibration frequency signals of the transformer, temperature signals of the coil and the box body, and brightness signals around the switch and the coil are acquired by infrared shooting;
the normalization unit is used for converting the operation signals into a two-dimensional data matrix, and then normalizing the two-dimensional data matrix to obtain a two-dimensional gray level map;
the first detection unit is used for inputting the two-dimensional gray level diagram of the time period to be detected into a trained dual cross contrast attention network, and outputting a first fault detection result of the transformer, wherein the dual cross contrast attention network is obtained by training a historical two-dimensional gray level diagram, and the dual cross contrast attention network comprises: global-local cross-contrast attention network and fault-normal cross-contrast attention network;
the second detection unit is used for determining a two-dimensional gray scale image corresponding to the transformer which does not generate faults according to the first fault detection result as a fault two-dimensional gray scale image, inputting the fault two-dimensional gray scale image into a trained short-term fault data prediction model, outputting a two-dimensional gray scale image of a prediction time period, inputting the two-dimensional gray scale image into the double cross contrast attention network, and outputting a second fault detection result of the transformer.
6. The transformer vibration detection system of claim 5, wherein the training process of the global-local cross contrast attention network specifically comprises:
regarding the two-dimensional gray scale map obtained in the step S3 as a query matrix R= [ R ] formed by N query vectors 1 ；r 2 ；…；r N ]Calculating a cumulative attention score S for an ith image block in accordance with an attention display i ；
According to the cumulative attention score S i From R i The first T query vectors corresponding to the first T highest responses in the accumulated weights of the CLS are selected to construct a new query matrix R l Representing the most interesting local embedding;
calculating cross-concerns between the selected local query and global key-value pairs based on the output function;
wherein the output function is:
in the method, in the process of the invention,
as a scale factor, the query matrix Q, key matrix K and value matrix V are embedded from the same input with different linear transformations, respectively N×D To calculate q=xw Q ,K＝XW K ，V＝XW V L and g are coefficients, and a new matrix F is output G Namely, a fault feature matrix;
according to the fault characteristic matrix F G And determining the fault characteristic weight by using GRA to obtain a fault weight matrix, and taking the fault weight matrix as an output matrix of the global-local cross contrast attention network.
7. The transformer vibration detection system of claim 5, wherein the training process of the fault-normal cross-contrast attention network specifically comprises:
constructing a fault-normal cross contrast attention network, and inputting the fault two-dimensional gray scale map and the normal two-dimensional gray scale map into the fault-normal cross contrast attention network in sequence for training;
calculating a normal data matrix and a fault characteristic weight matrix after the comparison of the data matrix containing faults according to the fault two-dimensional gray level diagram and the normal two-dimensional gray level diagram based on the output function, and taking the fault characteristic matrix and the fault characteristic weight matrix as an output matrix of the fault-normal cross comparison attention network;
combining the output matrixes of the fault-normal cross contrast attention network and the global-local cross contrast attention network through a contribution rate function to obtain a fault characteristic matrix and a fault characteristic weight matrix;
and layering the classified faults according to the fault characteristic weight matrix to obtain a fault classification layered characteristic set, and improving and iterating the contrast attention network parameters by the characteristic set trained by the offline network to obtain the dual cross contrast attention network.
8. The transformer vibration detection system according to claim 5, wherein the normalizing the two-dimensional matrix to obtain a two-dimensional gray scale map comprises:
based on a normalization formula, carrying out normalization processing on the two-dimensional data matrix to obtain a two-dimensional gray scale map;
wherein, the normalization formula is:
wherein L (i) is the two-dimensional data matrix, where i=1, 2, … …, n×m; max (L) represents the maximum value in the two-dimensional gray scale map; min (L) represents the minimum value in the two-bit gray scale map; r (N, M) (n=1, 2 … …, N, m=1, 2 … …, M) represents the pixel intensity of the image normalized; the circle function is a normalization function.
9. A transformer vibration detection apparatus, the apparatus comprising a processor and a memory:
the memory is used for storing program codes and transmitting the program codes to the processor;
the processor is configured to perform the transformer vibration detection method of any one of claims 1-4 according to instructions in the program code.
10. A computer readable storage medium for storing program code for performing the transformer vibration detection method of any one of claims 1-4.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202310188265.8A CN116049638A (en) | 2023-02-28 | 2023-02-28 | Transformer vibration detection method, system, equipment and storage medium |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202310188265.8A CN116049638A (en) | 2023-02-28 | 2023-02-28 | Transformer vibration detection method, system, equipment and storage medium |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN116049638A true CN116049638A (en) | 2023-05-02 |
Family
ID=86133264
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202310188265.8A Pending CN116049638A (en) | 2023-02-28 | 2023-02-28 | Transformer vibration detection method, system, equipment and storage medium |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN116049638A (en) |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN117910120A (en) * | 2024-03-20 | 2024-04-19 | 西华大学 | Buffeting response prediction method for wind-bridge system based on lightweight transducer |
-
2023
- 2023-02-28 CN CN202310188265.8A patent/CN116049638A/en active Pending
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN117910120A (en) * | 2024-03-20 | 2024-04-19 | 西华大学 | Buffeting response prediction method for wind-bridge system based on lightweight transducer |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| US11921566B2 (en) | Abnormality detection system, abnormality detection method, abnormality detection program, and method for generating learned model | |
| US20230152187A1 (en) | Fault signal locating and identifying method of industrial equipment based on microphone array | |
| EP3701708B1 (en) | Machine diagnosis using mobile devices and cloud computers | |
| CN108776808A (en) | A kind of method and apparatus for detecting ladle corrosion defect | |
| US11715190B2 (en) | Inspection system, image discrimination system, discrimination system, discriminator generation system, and learning data generation device | |
| CN108830837A (en) | A kind of method and apparatus for detecting ladle corrosion defect | |
| CN114509266B (en) | Bearing health monitoring method based on fault feature fusion | |
| KR20200014129A (en) | Diagnosis method of electric transformer using Deep Learning | |
| CN108121961A (en) | Inspection Activity recognition method, apparatus, computer equipment and storage medium | |
| US12007745B2 (en) | Apparatus for predicting equipment damage | |
| CN109978870A (en) | Method and apparatus for output information | |
| CN116049638A (en) | Transformer vibration detection method, system, equipment and storage medium | |
| Silva et al. | Assets predictive maintenance using convolutional neural networks | |
| CN114139589A (en) | Fault diagnosis method, device, equipment and computer readable storage medium | |
| CN116861331A (en) | Expert model decision-fused data identification method and system | |
| CN117473411A (en) | Bearing life prediction method based on improved transducer model | |
| JP2019158684A (en) | Inspection system, identification system, and discriminator evaluation device | |
| CN115904916A (en) | Hard disk failure prediction method and device, electronic equipment and storage medium | |
| CN117934963B (en) | Gas sensor drift compensation method | |
| CN118052825B (en) | Glass insulator surface flaw detection method | |
| Liu et al. | Artificial enabled communications energy trade management in sustainable generator based fault detection for work substation recordings | |
| CN117494588B (en) | Method, equipment and medium for optimizing residual effective life of fan bearing | |
| CN117723782B (en) | Sensor fault identification positioning method and system for bridge structure health monitoring | |
| CN118089743B (en) | Unmanned aerial vehicle intelligent navigation and camera system special for converter station | |
| JP2008083866A (en) | Emergency monitoring device |
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 |