CN114675118B - Transformer winding abnormality detection method, device, equipment and storage medium - Google Patents
Transformer winding abnormality detection method, device, equipment and storage medium Download PDFInfo
- Publication number
- CN114675118B CN114675118B CN202210595995.5A CN202210595995A CN114675118B CN 114675118 B CN114675118 B CN 114675118B CN 202210595995 A CN202210595995 A CN 202210595995A CN 114675118 B CN114675118 B CN 114675118B
- Authority
- CN
- China
- Prior art keywords
- frequency response
- response curve
- preset
- transformer winding
- initial
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active
Links
Images
Classifications
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01R—MEASURING ELECTRIC VARIABLES; MEASURING MAGNETIC VARIABLES
- G01R31/00—Arrangements for testing electric properties; Arrangements for locating electric faults; Arrangements for electrical testing characterised by what is being tested not provided for elsewhere
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01B—MEASURING LENGTH, THICKNESS OR SIMILAR LINEAR DIMENSIONS; MEASURING ANGLES; MEASURING AREAS; MEASURING IRREGULARITIES OF SURFACES OR CONTOURS
- G01B7/00—Measuring arrangements characterised by the use of electric or magnetic techniques
- G01B7/16—Measuring arrangements characterised by the use of electric or magnetic techniques for measuring the deformation in a solid, e.g. by resistance strain gauge
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01R—MEASURING ELECTRIC VARIABLES; MEASURING MAGNETIC VARIABLES
- G01R31/00—Arrangements for testing electric properties; Arrangements for locating electric faults; Arrangements for electrical testing characterised by what is being tested not provided for elsewhere
- G01R31/50—Testing of electric apparatus, lines, cables or components for short-circuits, continuity, leakage current or incorrect line connections
- G01R31/62—Testing of transformers
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01R—MEASURING ELECTRIC VARIABLES; MEASURING MAGNETIC VARIABLES
- G01R31/00—Arrangements for testing electric properties; Arrangements for locating electric faults; Arrangements for electrical testing characterised by what is being tested not provided for elsewhere
- G01R31/50—Testing of electric apparatus, lines, cables or components for short-circuits, continuity, leakage current or incorrect line connections
- G01R31/72—Testing of electric windings
Landscapes
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Engineering & Computer Science (AREA)
- Power Engineering (AREA)
- Complex Calculations (AREA)
- Testing Of Short-Circuits, Discontinuities, Leakage, Or Incorrect Line Connections (AREA)
- Protection Of Transformers (AREA)
Abstract
The application discloses a transformer winding abnormity detection method, a device, equipment and a storage medium, wherein the method comprises the following steps: acquiring an initial frequency response curve of a transformer winding; inputting the initial frequency response curve into a preset variational automatic encoder to reconstruct a sample to obtain a reconstructed frequency response curve, wherein the preset variational automatic encoder comprises an encoder, a memory enhancement module, a trend analyzer and a decoder; and carrying out anomaly analysis according to the initial frequency response curve and the reconstructed frequency response curve to obtain an anomaly detection result. In order to capture the time dynamic characteristic of the frequency response curve and retain more related information, a trend analyzer and a memory enhancement module are introduced, so that the reliability of the reconstructed data is ensured, and the accuracy of the abnormal detection result is improved. Therefore, the method and the device solve the technical problem that the transformer winding abnormity detection accuracy rate is low due to the lack of sample data in the prior art.
Description
Technical Field
The present disclosure relates to the field of device anomaly detection technologies, and in particular, to a method, an apparatus, a device, and a storage medium for detecting transformer winding anomalies.
Background
The power transformer is one of important electrical equipment of a power system, and the normal and stable operation of the power transformer has great significance on the safe production and the reliability of the power system. With the increasing and drastic increase of the capacity of the power grid, the loss and the caused influence caused by the fault of the transformer are more serious. Research data show that winding deformation is one of the most main failure types of transformers, and the transformers may cause winding deformation failure due to nonresistible factors such as human misoperation, external force and other natural disasters during transportation, installation and operation. Therefore, the method and the device for detecting the deformation degree of the transformer winding accurately detect the deformation degree of the transformer winding, timely master the health state of the winding inside the transformer, find potential fault hazards of the transformer as soon as possible and achieve state overhaul of the transformer are very important.
At present, Frequency Response (FRA) and Short Circuit Impedance (SCI) methods are mainly used to diagnose whether a transformer winding is deformed, the type of deformation, and the degree of deformation. Compared with the SCI method, the FRA method is sensitive to faults of windings and iron cores of the power transformer and is widely applied, and the sweep frequency range is 1 kH-1000 kHz. At present, the method for analyzing the FRA curve mainly includes an equivalent circuit model simulation method, an artificial intelligence method, a mathematical statistics method, and the like.
However, the current frequency response curve detection technology lacks test sample data, and the sample data is incomplete and incomplete, so that the accuracy of detecting the transformer winding deformation abnormality based on the frequency response curve is low.
Disclosure of Invention
The application provides a transformer winding abnormity detection method, device, equipment and storage medium, which are used for solving the technical problem of low transformer winding abnormity detection accuracy rate caused by sample data defects in the prior art.
In view of this, the first aspect of the present application provides a method for detecting an abnormality of a transformer winding, including:
acquiring an initial frequency response curve of a transformer winding;
inputting the initial frequency response curve into a preset variational automatic encoder to reconstruct a sample to obtain a reconstructed frequency response curve, wherein the preset variational automatic encoder comprises an encoder, a memory enhancement module, a trend analyzer and a decoder;
and carrying out anomaly analysis according to the initial frequency response curve and the reconstructed frequency response curve to obtain an anomaly detection result.
Preferably, the initial frequency response curve is input into a preset variational automatic encoder for sample reconstruction, so as to obtain a reconstructed frequency response curve, and the preset variational automatic encoder includes an encoder, a memory enhancement module, a trend analyzer and a decoder, and includes:
coding the initial frequency response curve through a coder in a preset variational automatic coder to obtain an initial hidden variable and a corresponding initial probability distribution;
performing memory enhancement processing on the initial hidden variable through a memory enhancement module in the preset variational automatic encoder based on an attention mechanism and a sparse operation mechanism to obtain an enhanced hidden variable and corresponding enhanced probability distribution;
performing dynamic trend analysis on the enhanced hidden variables and the enhanced probability distribution by a trend analyzer in the preset variational automatic encoder by adopting a polynomial function to obtain a frequency response time trend sequence;
and performing data integration on the basis of the frequency response time trend sequence and a decoding reconstruction sequence obtained by adopting a decoder to analyze the enhanced hidden variable to obtain a reconstruction frequency response curve.
Preferably, the obtaining a frequency response time trend sequence by performing dynamic trend analysis on the enhancement hidden variable and the enhancement probability distribution by a trend analyzer in the preset variation automatic encoder through a polynomial function includes:
configuring a preset time parameter of a polynomial function;
respectively calculating expansion coefficients corresponding to the enhanced hidden variables according to polynomial functions based on the preset times parameters to obtain expansion coefficient matrixes;
and carrying out dynamic trend analysis of time dimension according to the expansion coefficient matrix to obtain a frequency response time trend sequence.
Preferably, the performing anomaly analysis according to the initial frequency response curve and the reconstructed frequency response curve to obtain an anomaly detection result includes:
calculating a reconstruction error value according to the initial frequency response curve and the reconstruction frequency response curve;
and carrying out transformer winding abnormity analysis according to the reconstructed error value to obtain abnormity detection.
The present application provides in a second aspect a transformer winding anomaly detection device, including:
the curve acquisition module is used for acquiring an initial frequency response curve of the transformer winding;
the data reconstruction module is used for inputting the initial frequency response curve into a preset variational automatic encoder for sample reconstruction to obtain a reconstructed frequency response curve, and the preset variational automatic encoder comprises an encoder, a memory enhancement module, a trend analyzer and a decoder;
and the anomaly detection module is used for carrying out anomaly analysis according to the initial frequency response curve and the reconstructed frequency response curve to obtain an anomaly detection result.
Preferably, the data reconstruction module includes:
the curve coding submodule is used for coding the initial frequency response curve through a coder in a preset variation automatic coder to obtain an initial hidden variable and a corresponding initial probability distribution;
the information enhancement submodule is used for carrying out memory enhancement processing on the initial hidden variable through a memory enhancement module in the preset variational automatic encoder based on an attention mechanism and a sparse operation mechanism to obtain an enhanced hidden variable and corresponding enhanced probability distribution;
the trend analysis submodule is used for carrying out dynamic trend analysis on the enhanced hidden variable and the enhanced probability distribution by adopting a polynomial function through a trend analyzer in the preset variation automatic encoder to obtain a frequency response time trend sequence;
and the data integration submodule is used for carrying out data integration on the basis of the frequency response time trend sequence and a decoding reconstruction sequence obtained by adopting a decoder to analyze the enhanced hidden variable so as to obtain a reconstruction frequency response curve.
Preferably, the trend analysis submodule is specifically configured to:
configuring a preset time parameter of a polynomial function;
respectively calculating expansion coefficients corresponding to the enhanced hidden variables according to polynomial functions based on the preset times parameters to obtain expansion coefficient matrixes;
and carrying out dynamic trend analysis of time dimension according to the expansion coefficient matrix to obtain a frequency response time trend sequence.
Preferably, the abnormality detection module is specifically configured to:
calculating a reconstruction error value according to the initial frequency response curve and the reconstruction frequency response curve;
and carrying out transformer winding abnormity analysis according to the reconstructed error value to obtain abnormity detection.
A third aspect of the present application provides a transformer winding anomaly detection device, the device 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 transformer winding abnormality detection method according to the first aspect according to 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 winding abnormality detection method of the first aspect.
According to the technical scheme, the embodiment of the application has the following advantages:
the application provides a transformer winding abnormity detection method, which comprises the following steps: acquiring an initial frequency response curve of a transformer winding; inputting the initial frequency response curve into a preset variational automatic encoder to reconstruct a sample to obtain a reconstructed frequency response curve, wherein the preset variational automatic encoder comprises an encoder, a memory enhancement module, a trend analyzer and a decoder; and carrying out anomaly analysis according to the initial frequency response curve and the reconstructed frequency response curve to obtain an anomaly detection result.
According to the transformer winding abnormity detection method, new sample data, namely a frequency response curve, is generated through a variational automatic encoder; for normal windings, the reconstructed curve is similar to the initial curve, but for abnormal windings, the difference between the reconstructed curve and the initial curve is larger. Moreover, in order to capture the time dynamic characteristic of the frequency response curve and retain more related information, a trend analyzer and a memory enhancement module are introduced, so that the reliability of the reconstructed data is ensured, and the accuracy of the abnormal detection result is improved. Therefore, the method and the device can solve the technical problem that the transformer winding abnormity detection accuracy rate is low due to sample data defects in the prior art.
Drawings
Fig. 1 is a schematic flowchart of a method for detecting an abnormality of a transformer winding according to an embodiment of the present disclosure;
fig. 2 is another schematic flow chart of a transformer winding abnormality detection method according to an embodiment of the present disclosure;
fig. 3 is a schematic structural diagram of a transformer winding abnormality detection apparatus according to an embodiment of the present disclosure;
fig. 4 is a schematic diagram of an original network structure of a variational self-encoder according to an embodiment of the present application;
fig. 5 is a schematic diagram of an overall network structure of a preset variational automatic encoder according to an embodiment of the present application;
FIG. 6 is a schematic diagram of a network structure of a memory enhancement module according to an embodiment of the present disclosure;
fig. 7 is a schematic network structure diagram of a trend analyzer according to an embodiment of the present disclosure.
Detailed Description
In order to make those skilled in the art better understand the technical solutions of the present application, the technical solutions in the embodiments of the present application will be clearly and completely described below with reference to the drawings in the embodiments of the present application, and it is obvious that the described embodiments are only a part of the embodiments of the present application, and not all the embodiments. All other embodiments obtained by a person of ordinary skill in the art based on the embodiments in the present application without making any creative effort belong to the protection scope of the present application.
Interpretation of terms:
variational automatic encoder: variational Auto-Encoders, VAE;
time dynamic variational automatic encoder: time Dynamics VAE, TD-VAE.
For easy understanding, referring to fig. 1, a first embodiment of a method for detecting an abnormality of a transformer winding provided by the present application includes:
The deformation frequency curve of the power transformer winding is the pass frequencyThe principle of the frequency response method is that a sine frequency-sweeping voltage signal is applied to the head end of a winding of a tested transformer, a corresponding frequency response signal is measured at the tail end of the winding, and the logarithm of the ratio of the output voltage to the input voltage effective value is taken to obtain a frequency response curve reflecting the winding characteristics. The sweep frequency range of the frequency response method is 1k-1MHz, the low frequency band of the amplitude-frequency characteristic curve is 1k-100kHz, the middle frequency band is 100kHz-600kHz, and the high frequency band is 600kHz-1MHz, so the winding deformation frequency response curve can be regarded as time sequence data. Suppose a certain frequency band hasInitial frequency response curve time series of power transformer winding of sample setWhereinIndicating that each sample data containsThe number of the variables is one,。
and 102, inputting the initial frequency response curve into a preset variational automatic encoder to reconstruct a sample to obtain a reconstructed frequency response curve, wherein the preset variational automatic encoder comprises an encoder, a memory enhancement module, a trend analyzer and a decoder.
The original variational automatic encoder VAE comprises an encoder and a decoder, and the improved preset variational automatic encoder of the embodiment further comprises a memory enhancement module and a trend analyzer. The original variational automatic encoder is shown in fig. 4 and comprises an input layer, an implicit variational layer and an output layer; the variational automatic encoder learns the distribution rule of the samples by establishing a probability density distribution model of the samples. In the variation automatic encoder, the encoder is used for establishing variation inference of original input data and generating hidden dataA variation probability distribution of the variables; the decoder generates a probability density distribution close to the original data according to the hidden variable variation probability distribution generated by the inference network. As shown in FIG. 4, the encoder converts the raw dataEncoding to generate hidden variablesThe generating network may be based on hidden variablesThe reduction generates an approximate probability distribution of the original data.
The transmission-variable automatic encoder establishes a model by training normal data samples, judges whether the sample data is abnormal or not according to the size of reconstruction errors, and determines the sample with larger reconstruction errors as an abnormal sample. However, since both the encoder and the decoder are composed of deep neural networks, overfitting and no discrimination between normal and abnormal data may occur during the training of the model. In order to solve the problem, the present embodiment introduces a memory enhancement mechanism into the variational automatic encoder, and the core part of the present embodiment is to determine important information of sample data through the memory enhancement mechanism to perform the next transmission. The memory enhancement module is beneficial to keeping the associated information in the sample data, thereby ensuring the accuracy and reliability of the subsequent detection result.
In addition, the low, medium and high frequency bands of the deformation frequency response curve of the power transformer winding can be regarded as Time series data, the Time series data have the characteristics of instantaneity and Dynamics, and in order to be capable of well capturing the characteristics of the Time series data of the deformation frequency response curve of the power transformer winding, a trend analyzer is further introduced into a variation automatic encoder, or the Time dynamic variation automatic encoder (TD-VAE) can be introduced into the variation automatic encoder; the time series data can be analyzed, the time dynamic characteristic is reserved, the frequency response curve characteristic of the transformer winding can be reflected, and then the detection accuracy is improved.
And 103, carrying out anomaly analysis according to the initial frequency response curve and the reconstructed frequency response curve to obtain an anomaly detection result.
The essence of the anomaly analysis is that the difference between predicted data and actual measured data, which is the degree of deviation of the reconstructed frequency response curve from the initial ordinary response curve, is judged as normal data when the difference is small, and abnormal data, which is the deformation of the transformer winding, is judged when the difference is large.
According to the transformer winding abnormity detection method provided by the embodiment of the application, new sample data, namely a frequency response curve, is generated through a variational automatic encoder; for normal windings, the reconstructed resulting curve is similar to the initial curve, but for abnormal windings, the difference is larger. Moreover, in order to capture the time dynamic characteristic of the frequency response curve and retain more related information, a trend analyzer and a memory enhancement module are introduced, so that the reliability of the reconstructed data is ensured, and the accuracy of the abnormal detection result is improved. Therefore, the technical problem that the accuracy rate of the abnormal detection of the transformer winding is low due to the defect of sample data in the prior art can be solved.
For easy understanding, please refer to fig. 2, the present application provides a second embodiment of a transformer winding abnormality detection method, including:
Referring to fig. 4, the encoder and decoder each include a plurality of convolutional layers, and the specific sizes and numbers of convolutional cores and the number of convolutional layers can be configured as required. Due to hidden variablesThe distribution of the variable automatic encoder model is unknown, and the variable automatic encoder model cannot directly utilize the maximum expectation value algorithm to carry out variable inference solution (namely a decoding process), so that an identification model is introduced into an encoder by the variable automatic encoder modelTo replace the undetermined true probability distributionThus model ofCan be used as the encoder part of the variational automatic encoder, condition distributionAs part of a decoder. In order to make the recognition model approximately equal to the true probability density distribution, the variational autoencoder training process introduces noise and makes the distribution of the hidden variables close to the normal distribution as possibleIn whichIs taken as the average value of the values,is the variance. The loss function of the VAE model is a negative log-likelihood function with a regular term, and the expression of the loss function is as follows:
wherein the content of the first and second substances,data representing a frequency response curve of a deformation of a winding of the power transformer, i.e. an initial frequency response curve,for identifying the model, the representation is represented by hidden variablesTo generate new sample data,To identify parameters of the model;to generate the model, the representation is represented by sample dataTo generate hidden variables,In order to generate the parameters of the model,the KL divergence calculation function is represented,indicating that the desired value is to be solved for,for prior distribution, it is usually set to a standard normal distribution,And solving the maximum value of the lower variation bound through a maximum likelihood method. The first large term on the right side of the formula is a reconstruction loss function, and the second large term is KL divergence; the reconstruction loss function is used for controlling the generated data and the original data to be as close as possible, and KL is scatteredThe degree can measure the approximation degree of the generated probability distribution and the real probability distribution, and the specific calculation process of the KL divergence is as follows:
wherein, the first and the second end of the pipe are connected with each other,for recognition of the model, representation by hidden variablesTo generate new sample data,To identify parameters of the model;to generate the model, the representation is represented by sample dataTo generate hidden variables,To generate the parameters of the model.
And 203, performing memory enhancement processing on the initial hidden variable through a memory enhancement module in the preset variational automatic encoder based on an attention mechanism and a sparse operation mechanism to obtain an enhanced hidden variable and corresponding enhanced probability distribution.
Referring to fig. 5 and 6, the principle of the memory enhancement module is to add a memory module and a corresponding read/write mechanism on the basis of the neural network model, similar to a memory bank, capable of recording the slave encoderThe obtained characteristic mode of the typical hidden variable, namely the initial hidden variable probability distribution, determines important relevant information through an attention mechanism and a coefficient operation mechanism, and stores the important relevant information in a memory bank. If the memory matrix in the memory enhancement module is expressed asIs composed ofKStripe memoryEach memory bank is expressed in the form of a column vector, and the length of the column vectorHThe length of the data is consistent with the data length of the initial hidden variable probability distribution, and the enhanced hidden variable sequence obtained by the memory enhancement mechanism can be expressed as follows:
wherein the content of the first and second substances,for memorizing the weight row vector, the sum of its elements is 1, eachRepresenting each line of memoryThe weight of (a) is calculated,the calculation by attention mechanism is as follows:
wherein the content of the first and second substances,in order to be an initial hidden variable,to enhance latent variables;the method is used for measuring the similarity of two vectors, specifically, the cosine similarity principle can be adopted for calculation, and the method can be expressed as follows:
in order to strictly control the number of used memory banks and ensure that effective information is adopted to obtain new hidden variables, sparse operation pairs are adoptedAnd (5) correcting:
wherein the content of the first and second substances,in order to calculate the maximum value of the values,is as followsiThe weight of each of the plurality of weights,、in order to correct the coefficients of the coefficients,in the intervalThen to the one obtained after correctionAnd (3) carrying out standardization treatment:
wherein the content of the first and second substances,representing the 1-norm of the column vector.
And 204, carrying out dynamic trend analysis on the enhanced hidden variable and the enhanced probability distribution by adopting a polynomial function through a trend analyzer in the preset variation automatic encoder to obtain a frequency response time trend sequence.
Further, step 204 includes:
configuring a preset time parameter of a polynomial function;
respectively calculating expansion coefficients corresponding to the enhanced hidden variables according to a polynomial function based on preset time parameters to obtain an expansion coefficient matrix;
and carrying out dynamic trend analysis of the time dimension according to the expansion coefficient matrix to obtain a frequency response time trend sequence.
Referring to fig. 5 and 7, the trend analyzer mainly includes two fully connected layers and a matrix-shaped reconstruction layer, and the polynomial function adjusts the extracted trend characteristics in the form of a matrix to obtain a time trend sequence reflecting the dynamic change of the frequency response curveIt is the actual uncertainty and trend of variation in capturing the time series data. The variational autoencoder introducing the concept of trend analysis is called a time-dynamic variational autoencoder, TD-VAE. The enhanced hidden variables are input to the fully-connected layer. By usingThe function is activated.Expressed as:
The preset order parameter of the configuration polynomial function is expressed asFor each enhanced latent variableCan all calculateExpansion coefficient matrix of the secondary termOf dimension ofWhen is coming into contact withThe time is an average value of the original time series, which indicates that the time series data does not have a trend of rising or falling and tends to a steady state. Based on the expansion coefficient matrixPerforming a time dynamic trend analysis may be expressed as:
wherein, the first and the second end of the pipe are connected with each other,as a time matrixrIs/are as followspA power matrix of dimensionsTime matrix,TIs a time step.
The trend analyzer performs network model layout according to a polynomial function and calculates a time trend sequence of the enhanced hidden variable。
And step 205, performing data integration on the basis of the frequency response time trend sequence and a decoding reconstruction sequence obtained by analyzing the enhanced hidden variable by a decoder to obtain a reconstruction frequency response curve.
Referring to fig. 5 and 7, a time trend sequence is obtainedThen, the decoded reconstructed sequence output by the decoder is neededThe integration is performed to obtain the reconstructed frequency response sequenceAnd then drawing a reconstructed frequency response curve. The encoder in this embodiment is composed of a fully-connected layer, a matrix-shaped reconstruction layer, two deconvolution layers, and a time-sequence distributed fully-connected layer, and the output of the fully-connected layer can be expressed as:
wherein the content of the first and second substances,in order to be a linear activation function,wherein, in the process,Yis the output of the network of the upper layer,、Ware all weight matrices in the network model,in order to be a kernel function, the kernel function,is a bias matrix.
It should be noted that the preset variational automatic encoder includes an encoder, a memory enhancement module, a trend analyzer and a decoder. In the training phase, because a memory enhancing module is introduced, a loss function needs to be adjusted, and the adjusted loss function is as follows:
wherein the content of the first and second substances,is a predetermined parameter, typically an empirical value, such as 0.00025;Sis composed ofThe dimension (c) of (a) is,for calculating a memory weight row vectorEntropy value of (a).
And step 206, calculating a reconstruction error value according to the initial frequency response curve and the reconstruction frequency response curve.
And step 207, carrying out transformer winding abnormity analysis according to the reconstructed error value to obtain abnormity detection.
The essence is that whether the sample is abnormal is judged by calculating the reconstruction error of the reconstruction data, and the criterion is as follows:
wherein the content of the first and second substances,in order to reconstruct the error value(s),the sample data is tested for the frequency response curve,the anomaly threshold value may be set according to actual conditions, such as 0.5. It is understood that anomaly detection is detecting whether the transformer winding is deformed.
For easy understanding, please refer to fig. 3, the present application further provides an embodiment of a transformer winding abnormality detection apparatus, including:
the curve acquisition module 301 is used for acquiring an initial frequency response curve of the transformer winding;
the data reconstruction module 302 is configured to input the initial frequency response curve into a preset variational automatic encoder for sample reconstruction, so as to obtain a reconstructed frequency response curve, where the preset variational automatic encoder includes an encoder, a memory enhancement module, a trend analyzer, and a decoder;
and the anomaly detection module 303 is configured to perform anomaly analysis according to the initial frequency response curve and the reconstructed frequency response curve to obtain an anomaly detection result.
Further, the data reconstruction module 302 includes:
a curve coding submodule 3021, configured to perform coding processing on an initial frequency response curve through an encoder in a preset variational automatic encoder to obtain an initial hidden variable and a corresponding initial probability distribution;
the information enhancer module 3022 is configured to perform memory enhancement processing on the initial hidden variable through a memory enhancement module in the preset variational automatic encoder based on an attention mechanism and a sparse operation mechanism to obtain an enhanced hidden variable and a corresponding enhanced probability distribution;
the trend analysis submodule 3023 is configured to perform dynamic trend analysis on the enhanced hidden variable and the enhanced probability distribution by using a polynomial function through a trend analyzer in the preset variation autoencoder to obtain a frequency response time trend sequence;
and the data integration sub-module 3024 is configured to perform data integration on the basis of the frequency response time trend sequence and a decoding reconstruction sequence obtained by analyzing the enhanced hidden variable with a decoder, so as to obtain a reconstruction frequency response curve.
Further, the trend analysis submodule 3023 is specifically configured to:
configuring a preset time parameter of a polynomial function;
respectively calculating expansion coefficients corresponding to the enhanced hidden variables according to a polynomial function based on preset time parameters to obtain an expansion coefficient matrix;
and carrying out dynamic trend analysis of the time dimension according to the expansion coefficient matrix to obtain a frequency response time trend sequence.
Further, the anomaly detection module 303 is specifically configured to:
calculating a reconstruction error value according to the initial frequency response curve and the reconstruction frequency response curve;
and carrying out transformer winding abnormity analysis according to the reconstructed error value to obtain abnormity detection.
The application also provides transformer winding abnormity detection equipment which comprises a processor and a memory;
the memory is used for storing the program codes and transmitting the program codes to the processor;
the processor is used for executing the transformer winding abnormality detection method in the above method embodiment according to instructions in the program code.
The present application further provides a computer-readable storage medium for storing program code for executing the transformer winding abnormality detection method in the above method embodiment.
In the several embodiments provided in the present application, it should be understood that the disclosed apparatus and method may be implemented in other ways. For example, the above-described apparatus embodiments are merely illustrative, and for example, the division of the units is only one type of logical functional division, and other divisions may be realized in practice, for example, multiple units or components may be combined or integrated into another system, or some features may be omitted, or not executed. In addition, the shown or discussed mutual coupling or direct coupling or communication connection may be an indirect coupling or communication connection through some interfaces, devices or units, and may be in an electrical, mechanical or other form.
The units described as separate parts may or may not be physically separate, and parts displayed 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 can be selected according to actual needs to achieve the purpose of the solution of the embodiment.
In addition, functional units in the embodiments of the present application may be integrated into one processing unit, or each unit may exist alone physically, or two or more units are integrated into one unit. The integrated unit can be realized in a form of hardware, and can also be realized in a form of a software functional unit.
The integrated unit, if implemented in the form of a software functional unit and sold or used as a stand-alone product, may be stored in a computer readable storage medium. Based on such understanding, the technical solution of the present application may be substantially implemented or contributed to by the prior art, or all or part of the technical solution may be embodied in a software product, which is stored in a storage medium and includes instructions for executing all or part of the steps of the method described in the embodiments of the present application through a computer device (which may be a personal computer, a server, or a network device). And the aforementioned storage medium includes: a U disk, a removable hard disk, a Read-Only Memory (ROM), a Random Access Memory (RAM), a magnetic disk or an optical disk, and other various media capable of storing program codes.
The above embodiments are only used for illustrating the technical solutions 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 solutions described in the foregoing embodiments may still be modified, or some technical features may be equivalently replaced; and such modifications or substitutions do not depart from the spirit and scope of the corresponding technical solutions of the embodiments of the present application.
Claims (8)
1. A transformer winding abnormality detection method is characterized by comprising the following steps:
acquiring an initial frequency response curve of a transformer winding;
inputting the initial frequency response curve into a preset variational automatic encoder to reconstruct a sample to obtain a reconstructed frequency response curve, wherein the preset variational automatic encoder comprises an encoder, a memory enhancement module, a trend analyzer and a decoder, and the specific process of generating the reconstructed frequency response curve is as follows:
coding the initial frequency response curve through a coder in a preset variational automatic coder to obtain an initial hidden variable and a corresponding initial probability distribution;
performing memory enhancement processing on the initial hidden variable through a memory enhancement module in the preset variational automatic encoder based on an attention mechanism and a sparse operation mechanism to obtain an enhanced hidden variable and a corresponding enhanced probability distribution;
performing dynamic trend analysis on the enhanced hidden variables and the enhanced probability distribution by a trend analyzer in the preset variational automatic encoder by adopting a polynomial function to obtain a frequency response time trend sequence;
performing data integration on the basis of the frequency response time trend sequence and a decoding reconstruction sequence obtained by adopting a decoder to analyze the enhanced hidden variable to obtain a reconstruction frequency response curve;
and carrying out anomaly analysis according to the initial frequency response curve and the reconstructed frequency response curve to obtain an anomaly detection result.
2. The method for detecting transformer winding abnormality according to claim 1, wherein the obtaining a frequency response time trend sequence by performing dynamic trend analysis on the enhanced hidden variable and the enhanced probability distribution by a trend analyzer in the preset variational automatic encoder using a polynomial function includes:
configuring a preset time parameter of a polynomial function;
respectively calculating expansion coefficients corresponding to the enhanced hidden variables according to polynomial functions based on the preset times parameters to obtain expansion coefficient matrixes;
and carrying out dynamic trend analysis of time dimension according to the expansion coefficient matrix to obtain a frequency response time trend sequence.
3. The method for detecting the abnormality of the winding of the transformer according to claim 1, wherein the performing abnormality analysis according to the initial frequency response curve and the reconstructed frequency response curve to obtain an abnormality detection result includes:
calculating a reconstruction error value according to the initial frequency response curve and the reconstruction frequency response curve;
and carrying out transformer winding abnormity analysis according to the reconstructed error value to obtain abnormity detection.
4. An abnormality detection device for a transformer winding, comprising:
the curve acquisition module is used for acquiring an initial frequency response curve of the transformer winding;
the data reconstruction module is used for inputting the initial frequency response curve into a preset variational automatic encoder for sample reconstruction to obtain a reconstructed frequency response curve, the preset variational automatic encoder comprises an encoder, a memory enhancement module, a trend analyzer and a decoder, and the data reconstruction module comprises:
the curve coding submodule is used for coding the initial frequency response curve through a coder in a preset variation automatic coder to obtain an initial hidden variable and a corresponding initial probability distribution;
the information enhancement submodule is used for carrying out memory enhancement processing on the initial hidden variable through a memory enhancement module in the preset variational automatic encoder based on an attention mechanism and a sparse operation mechanism to obtain an enhanced hidden variable and corresponding enhanced probability distribution;
the trend analysis submodule is used for carrying out dynamic trend analysis on the enhanced hidden variable and the enhanced probability distribution by adopting a polynomial function through a trend analyzer in the preset variation automatic encoder to obtain a frequency response time trend sequence;
the data integration submodule is used for carrying out data integration on the basis of the frequency response time trend sequence and a decoding reconstruction sequence obtained by adopting a decoder to analyze the enhanced implicit variable to obtain a reconstruction frequency response curve;
and the anomaly detection module is used for carrying out anomaly analysis according to the initial frequency response curve and the reconstructed frequency response curve to obtain an anomaly detection result.
5. The transformer winding abnormality detection apparatus according to claim 4, wherein the trend analysis submodule is specifically configured to:
configuring a preset time parameter of a polynomial function;
respectively calculating expansion coefficients corresponding to the enhanced hidden variables according to polynomial functions based on the preset times parameters to obtain expansion coefficient matrixes;
and carrying out dynamic trend analysis of time dimension according to the expansion coefficient matrix to obtain a frequency response time trend sequence.
6. The transformer winding abnormality detection apparatus according to claim 4, wherein the abnormality detection module is specifically configured to:
calculating a reconstruction error value according to the initial frequency response curve and the reconstruction frequency response curve;
and carrying out transformer winding abnormity analysis according to the reconstructed error value to obtain abnormity detection.
7. An apparatus for detecting abnormality of a transformer winding, 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 transformer winding anomaly detection method according to any one of claims 1-3 according to instructions in the program code.
8. A computer-readable storage medium, characterized in that the computer-readable storage medium is configured to store program code for performing the transformer winding anomaly detection method according to any one of claims 1-3.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202210595995.5A CN114675118B (en) | 2022-05-30 | 2022-05-30 | Transformer winding abnormality detection method, device, equipment and storage medium |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202210595995.5A CN114675118B (en) | 2022-05-30 | 2022-05-30 | Transformer winding abnormality detection method, device, equipment and storage medium |
Publications (2)
Publication Number | Publication Date |
---|---|
CN114675118A CN114675118A (en) | 2022-06-28 |
CN114675118B true CN114675118B (en) | 2022-08-23 |
Family
ID=82079791
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN202210595995.5A Active CN114675118B (en) | 2022-05-30 | 2022-05-30 | Transformer winding abnormality detection method, device, equipment and storage medium |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN114675118B (en) |
Family Cites Families (8)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN108510002B (en) * | 2018-04-02 | 2020-02-21 | 西南交通大学 | Method for detecting short circuit impact resistance of wound core traction transformer winding |
CN110794209A (en) * | 2019-11-14 | 2020-02-14 | 云南电网有限责任公司电力科学研究院 | Method and device for identifying and calibrating winding deformation frequency response data errors and storage medium |
CN111612078A (en) * | 2020-05-25 | 2020-09-01 | 中国人民解放军军事科学院国防工程研究院 | Transformer fault sample enhancement method based on condition variation automatic encoder |
CN112304419A (en) * | 2020-10-25 | 2021-02-02 | 广东石油化工学院 | Vibration and sound detection signal reconstruction method and system by using generalized sparse coding |
CN112488235A (en) * | 2020-12-11 | 2021-03-12 | 江苏省特种设备安全监督检验研究院 | Elevator time sequence data abnormity diagnosis method based on deep learning |
CN113569756B (en) * | 2021-07-29 | 2023-06-09 | 西安交通大学 | Abnormal behavior detection and positioning method, system, terminal equipment and readable storage medium |
CN113743016B (en) * | 2021-09-09 | 2023-06-30 | 湖南工业大学 | Engine residual life prediction method based on self-encoder and echo state network |
CN114240797B (en) * | 2021-12-22 | 2023-04-18 | 海南大学 | OCT image denoising method, device, equipment and medium |
-
2022
- 2022-05-30 CN CN202210595995.5A patent/CN114675118B/en active Active
Also Published As
Publication number | Publication date |
---|---|
CN114675118A (en) | 2022-06-28 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN111967571A (en) | MHMA-based anomaly detection method and equipment | |
KR101323714B1 (en) | Detection of defects in an electrochemical device | |
CN112731260B (en) | Online evaluation method for error state of voltage transformer based on concept drift recognition | |
CN114239725B (en) | Electric larceny detection method for data poisoning attack | |
CN113657622B (en) | Multi-dimensional state data fusion method, device, terminal and storage medium for power equipment | |
CN112508243A (en) | Training method and device for multi-fault prediction network model of power information system | |
CN116148656B (en) | Portable analog breaker fault detection method | |
CN112085111B (en) | Load identification method and device | |
CN111080168B (en) | Power communication network equipment reliability evaluation method based on capsule network | |
CN115169430A (en) | Cloud network end resource multidimensional time sequence anomaly detection method based on multi-scale decoding | |
CN109557434B (en) | Partial discharge signal identification method under strong background noise based on composite dictionary sparse representation classification | |
CN116306806A (en) | Fault diagnosis model determining method and device and nonvolatile storage medium | |
CN114675118B (en) | Transformer winding abnormality detection method, device, equipment and storage medium | |
CN114881157A (en) | Method, device and equipment for detecting working state of converter valve and storage medium | |
CN113611568A (en) | Vacuum circuit breaker based on genetic convolution depth network | |
CN115423370A (en) | Relay protection equipment health state assessment method and device | |
CN113705637B (en) | Method, system, equipment and readable storage medium for detecting mechanical faults of circuit breaker | |
CN115802401A (en) | Wireless network channel state prediction method, device, equipment and storage medium | |
CN115659851A (en) | Temperature rise evaluation method and device for cable terminal | |
CN113253069B (en) | Partial discharge risk assessment method based on multispectral intensity | |
CN115130546A (en) | Abnormal parameter detection method and device, computer readable medium and electronic equipment | |
CN112819067A (en) | Method, device, equipment and storage medium for processing bad data of power distribution network | |
CN114676593A (en) | Abnormity detection method of textile equipment and related device | |
CN116125922B (en) | Complex industrial process monitoring method and system based on parallel dictionary learning | |
CN113743355B (en) | Switch device state checking method, device, system and computer equipment |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
PB01 | Publication | ||
PB01 | Publication | ||
SE01 | Entry into force of request for substantive examination | ||
SE01 | Entry into force of request for substantive examination | ||
GR01 | Patent grant | ||
GR01 | Patent grant |