Model training method and transformer winding state monitoring method based on compressed sensing

Abstract

The invention provides a model training method and a transformer winding state monitoring method based on compressed sensing, and the model training method employs a compressed sensing technology to reconstruct a transformer vibration signal, employs a small amount of data to reconstruct an original signal, and employs the reconstructed signal to train a CNN network, thereby obtaining a winding state monitoring model. And further monitoring the state of the transformer winding by using the model, reconstructing a small amount of transformer vibration signals with original characteristics to obtain reconstructed signals similar to the original signals, and inputting the reconstructed signals into the trained model for state monitoring, thereby obtaining the state of the winding or performing fault judgment on the winding. According to the method provided by the invention, the workload of data acquisition and the data transmission pressure can be greatly reduced, and further, the frequency domain characteristics of the reconstructed signal are taken as the input of model training, so that the dynamic characteristics of the vibration signal are better mined.

Description

Technical Field

[0001] The present invention relates to transformer state monitoring technology, and in particular to a model training method and a transformer winding state monitoring method based on compressed sensing. Background Art

[0002] Power transformers are one of the most critical and important equipment in the power system. Their stable and safe operation is of great significance to the safety and reliability of the power grid. If such transformers fail during operation, not only will the maintenance cycle be long and expensive, but they may also cause sudden large-scale power outages, causing huge economic losses. Therefore, it is particularly important to conduct real-time online status monitoring and fault diagnosis of large power transformers in operation. The windings and cores of transformers are the parts with more failures. By analyzing the vibration signals on the surface of the oil tank, abnormalities or damages of the transformer windings can be detected in time, potential accidents can be prevented, and traditional regular maintenance can be transformed into status-based maintenance, which will greatly extend the service life of the transformer and reduce operating costs.

[0003] With the development of sensor technology and artificial intelligence, deep learning methods are increasingly used in transformer equipment vibration fault identification. It has the following advantages: 1) High judgment accuracy. 2) Stable judgment effect. 3) Clear judgment results, no ambiguous diagnosis results. However, deep learning models require a large number of samples for training, which greatly increases the amount of data that the data acquisition system needs to obtain, and puts higher requirements on subsequent signal processing and transmission. Summary of the invention

[0004] Based on this, the present invention aims to propose a model training method and a transformer winding state monitoring method based on compressed sensing, which uses the compressed sensing method to reconstruct the transformer vibration signal, uses a small amount of data to reconstruct the original signal, and then uses the reconstructed signal to train the CNN network, thereby completing the state monitoring of the transformer winding and realizing fault identification.

[0005] In a first aspect, the present invention provides a model training method, wherein a model trained by the method is used for transformer winding state monitoring based on compressed sensing, comprising:

[0006] Obtain historical transformer vibration signals;

[0007] The historical transformer vibration signal is reconstructed using the compressed sensing method to obtain a reconstructed signal set;

[0008] The reconstructed signal set is used as the input of the convolutional neural network for model training to obtain the winding state monitoring model.

[0009] Furthermore, the reconstructed signal set is used as the input of the convolutional neural network for model training, including:

[0010] Performing a time domain transformation on the reconstructed signal to obtain a frequency domain feature representation of the reconstructed signal;

[0011] The frequency domain feature representation of the reconstructed signal is used as the input of the convolutional neural network for model training.

[0012] Furthermore, the FFT transform method is selected for time domain transform.

[0013] Furthermore, the compressed sensing method is used to reconstruct the historical transformer vibration signal, and the reconstructed signal set includes:

[0014] According to the characteristics of transformer vibration signal, the compression measurement matrix and signal sparse transformation basis are designed;

[0015] Based on the compressed measurement matrix and the signal sparse transformation basis, the transformer vibration signal is reconstructed using the sparse signal reconstruction algorithm to obtain the reconstructed signal set.

[0016] Furthermore, the sparse signal reconstruction algorithm is a generalized orthogonal matching pursuit algorithm.

[0017] Furthermore, the signal sparse transformation basis selects a discrete cosine transform basis.

[0018] In a second aspect, the present invention provides a transformer winding state monitoring method based on compressed sensing, comprising:

[0019] Collect transformer vibration signals;

[0020] The transformer vibration signal is reconstructed using the compressed sensing method to obtain a reconstructed signal;

[0021] The reconstructed signal is input into the winding state monitoring model trained by the model training method of the first aspect to obtain a state monitoring result.

[0022] Furthermore, the transformer vibration signal is reconstructed using the compressed sensing method, and the reconstructed signal includes:

[0023] According to the characteristics of transformer vibration signal, the compression measurement matrix and signal sparse transformation basis are designed;

[0024] Based on the compressed measurement matrix and the signal sparse transformation basis, the transformer vibration signal is reconstructed using the sparse signal reconstruction algorithm to obtain the reconstructed signal.

[0025] Furthermore, before the reconstructed signal is input into the winding condition monitoring model, it includes:

[0026] The reconstructed signal is transformed in the time domain to obtain the frequency domain feature representation of the reconstructed signal, and the frequency domain feature representation of the reconstructed signal is used as the input of the winding state monitoring model.

[0027] Furthermore, the FFT transform method is selected for time domain transform.

[0028] In a third aspect, the present invention provides a model training device, comprising:

[0029] A signal acquisition module, used to acquire historical transformer vibration signals;

[0030] A first signal reconstruction module is used to reconstruct the historical transformer vibration signal by using a compressed sensing method to obtain a reconstructed signal set;

[0031] The model training module is used to use the reconstructed signal set as the input of the convolutional neural network for model training to obtain a winding state monitoring model.

[0032] In a fourth aspect, the present invention provides a transformer winding state monitoring device, comprising:

[0033] A signal acquisition module, used to collect transformer vibration signals;

[0034] The second signal reconstruction module is used to reconstruct the transformer vibration signal by using a compressed sensing method to obtain a reconstructed signal;

[0035] The state monitoring module is used to input the reconstructed signal into the winding state monitoring model trained by the model training device of the third aspect to obtain the state monitoring result.

[0036] In a fifth aspect, the present invention provides an electronic device, comprising: a memory and a processor;

[0037] The memory is used to store programs;

[0038] The processor is used to call the program stored in the memory to execute the model training method provided by the above-mentioned first aspect embodiment and / or any possible implementation method combined with the first aspect embodiment, or to execute the transformer winding state monitoring method based on compressed sensing provided by the above-mentioned second aspect embodiment.

[0039] In a sixth aspect, the present invention further provides a readable storage medium having a computer program stored thereon. When the computer program is executed by a processor, it implements the model training method provided in the above-mentioned first aspect embodiment and / or any possible implementation method in combination with the first aspect embodiment, or executes the transformer winding state monitoring method based on compressed sensing provided in the above-mentioned second aspect embodiment.

[0040] Compared with the prior art, the present invention has the following advantages:

[0041] The present invention proposes a model training method and a transformer winding state monitoring method based on compressed sensing, wherein the model training method utilizes compressed sensing technology to reconstruct transformer vibration signals, uses a small amount of data to reconstruct the original signal, and then uses the reconstructed signal to train a CNN network, thereby obtaining a winding state monitoring model, and further uses the model to monitor the transformer winding state, reconstructs a small amount of transformer vibration signals with original features, obtains a reconstructed signal that is approximate to the original signal, and inputs the reconstructed signal into the trained model for state monitoring, thereby knowing the state of the winding or performing fault judgment on it; the method provided by the present invention can greatly reduce the workload of data collection and the pressure of data transmission, and a further embodiment uses the frequency domain characteristics of the reconstructed signal as input for model training, so as to better explore the dynamic characteristics of the vibration signal. BRIEF DESCRIPTION OF THE DRAWINGS

[0042] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on the provided drawings without paying creative work.

[0043] Figure 1 is a flow chart for implementing the model training method provided by an embodiment of the present invention;

[0044] Figure 2 is a schematic diagram of transformer vibration signal reconstruction based on compressed sensing provided by an embodiment of the present invention;

[0045] Figure 3 It is a waveform comparison diagram of a reconstructed signal after compressing and reconstructing a simulation signal and an original signal provided by an embodiment of the present invention;

[0046] Figure 4 This is a waveform comparison diagram of a reconstructed signal after 50% compression and reconstruction of a transformer mechanical vibration signal provided by an embodiment of the present invention and an original signal;

[0047] Figure 5 is a CNN network architecture diagram of a winding state monitoring model provided by an embodiment of the present invention;

[0048] Figure 6 It is a framework diagram for model training of a winding state monitoring model provided by an embodiment of the present invention;

[0049] Figure 7 is a flow chart of a transformer winding state monitoring method based on compressed sensing provided by an embodiment of the present invention;

[0050] Figure 8is a schematic diagram of the structure of a model training device provided by an embodiment of the present invention;

[0051] Fig. 9 It is a schematic structural diagram of a transformer winding state monitoring device provided by an embodiment of the present invention;

[0052] Fig.10 It is a diagram of the electronic device architecture provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0053] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0054] See also Figure 1 An embodiment of the present invention provides a model training method, wherein the model trained by the method is used for transformer winding state monitoring based on compressed sensing, and comprises the following steps:

[0055] Step S110: Acquire historical transformer vibration signals.

[0056] Specifically, this step obtains a small amount of transformer tank vibration signals, which contain a large number of original features, are usually collected by vibration sensors and stored as time series data. Furthermore, the obtained signals can be preprocessed before analysis, including denoising, normalization, etc.

[0057] Step S120: Reconstruct the historical transformer vibration signal using the compressed sensing method to obtain a reconstructed signal set.

[0058] Specifically, this step uses compressed sensing to reconstruct a small amount of vibration signals to obtain a reconstructed signal that is close to the original signal. The main step of the compressed sensing method is to convert the vibration signal into a sparse representation. Compressed sensing relies on the sparsity of the signal under a certain basis, that is, the signal can be represented as a linear combination of a small number of non-zero coefficients in a specific transform domain. In this framework, a small amount of random linear measurements are used to obtain information about the vibration signal. The key to this process lies in the design of the compressed measurement matrix and the selection of the signal sparse representation basis.

[0059] Step S130: Use the reconstructed signal set as the input of the convolutional neural network to perform model training to obtain a winding state monitoring model.

[0060] Specifically, this step uses a convolutional neural network as the architecture of the model. The convolutional neural network is a multi-layer deep neural network that can directly extract sensitive features from the input signal through convolution operations. At the same time, it can effectively reduce the complexity and computational complexity of the model through local weight sharing. It mainly includes an input layer, a convolution layer, a downsampling layer, a fully connected layer, and an output layer. Among them, the convolution layer is used to perform convolution calculations on the input signal to extract key features, the downsampling layer is used to represent the features in reduced dimensions, and the fully connected layer sequentially connects the downsampling results to obtain a one-dimensional feature set.

[0061] In a further embodiment, the process of compressed sensing signal reconstruction in step S120 is as follows:

[0062] The measurement matrix is ​​designed according to the characteristics of the transformer vibration signal. The measurement matrix directly affects the value of the observation number and the reconstruction accuracy. The observation compression of the signal is expressed as follows:

[0063]

[0064] in, Represents the observation matrix The observation vector obtained after observation, is the original signal.

[0065] According to the theory of compressed sensing, the observed signal is required to be sparsely represented in a certain transform domain, that is:

[0066]

[0067] in, is the perception matrix, and its compression sampling process is as follows Figure 2 As shown, from the measured value Recovery The necessary condition is that the observation matrix must satisfy the constrained equidistance principle. If the perception matrix A satisfies the following formula, it is said to satisfy the constrained equidistance principle.

[0068]

[0069] in, is the isometry constant. Since it is sometimes not easy to verify whether the perception matrix satisfies the constraint isometry, an equivalent irrelevance criterion is proposed as follows:

[0070]

[0071] In the formula, Represents the observation matrix With sparse transform basis The relevance of is the column of the sparse basis, is the row of the observation matrix, The larger the value of , the higher the correlation between the measurement matrix and the sparse basis, and vice versa. A high correlation means that more observations are needed. It can be seen that when selecting the measurement matrix, the measurement matrix that is uncorrelated with the sparse transformation basis should be selected as much as possible.

[0072] In some embodiments, the measurement matrix may be selected from a Bernoulli random matrix, a partial Hadamard measurement matrix, or a Gaussian random matrix.

[0073] In some embodiments, since the compressed sensing framework requires that the signal is sparse or approximately sparse, and such signals rarely exist in vibration signals, the compressed sensing theory points out that as long as the signal is sparse in a certain domain after transformation, it can also meet the requirements of the theory. Discrete cosine transform (DCT) is a commonly used sparse transform basis, and the vibration signal has better sparsity in the DCT sparse domain. Therefore, DCT can be selected as the sparse transform basis of the transformer vibration signal in this embodiment.

[0074] Then, a sparse signal reconstruction algorithm is used to reconstruct the measured values Recovery , sparse reconstruction can be expressed as the following optimization problem in the compressed sensing process:

[0075]

[0076] To solve the above optimization problem, considering that it is an NP problem in this embodiment, it can be converted into the following expression: Norm minimization problem to solve the original signal:

[0077]

[0078] In a further embodiment, the generalized orthogonal matching pursuit algorithm (GOMP) is used for signal reconstruction. The algorithm uses the conjugate gradient method instead of the least squares method in OMP to solve the estimated value of the original signal. The conjugate gradient method has fast convergence and high precision. It aims to use the first-order derivative information to solve the problem of oscillating convergence during the convergence of the steepest descent method, which causes the operation time to be too long. It also reduces the amount of data occupied by the Newton method to save the matrix and the amount of matrix inversion operations. The specific steps include the following process:

[0079] Step S121. Input original signal , dictionary matrix , sparsity ;

[0080] Step S122. Output the original signal of -Sparse approximation ;

[0081] Step S123. Initialize the following parameters: , index collection , number of iterations ;

[0082] Step S124. Calculate dictionary All atoms in The inner product of the residual, record the column number of the atom corresponding to the maximum absolute value of the inner product , ;

[0083] Step S125. Update index set , update the sub-dictionary ;

[0084] Step S126: Calculate the estimated value using the conjugate gradient method , the initial value is 0, The step length is:

[0085]

[0086] Calculate and update the step size And the signal , until the number of conjugate gradient method iterations reaches n (the length of x).

[0087] Step S127. Calculate estimated value The N that can cause the largest residual change, and record the corresponding index into the support set Λ, , ,, Update atomic set Record the observation matrix column vector corresponding to the maximum inner product value.

[0088] Step S128. Update residual , update the candidate set , let the number of iterations .

[0089] Step S129. Determination If the condition is met, the iteration is stopped; otherwise, step S121 is executed again.

[0090] Figure 3 and Figure 4 The signal compression effect diagrams of the compressed sensing framework using the GOMP algorithm are shown respectively. Figure 3 The diagram shows the compression and reconstruction of the simulation signal. Figure 4 The diagram shows the compression reconstruction of the actual transformer mechanical vibration signal with a compression rate of 50%. It can be seen that the original signal and the reconstructed signal basically coincide with each other, and the compression reconstruction effect is good.

[0091] In a further embodiment, for the vibration of the transformer, its amplitude and phase information can better reflect the state of the winding. The frequency, amplitude and other characteristics generated by different vibration faults are different from those in the normal state. It is more meaningful to analyze the fault mechanism and locate the fault source based on the characteristic frequency and amplitude of the fault signal for state monitoring. Therefore, the reconstructed vibration signal can be converted into the time domain, and the frequency domain feature representation of the signal can be extracted. The frequency domain feature representation is used as the input for model training, including the frequency component of the signal and its distribution, amplitude, spectrum diagram, etc., so that the convolutional neural network learns the winding state characteristics from the frequency domain signal for state monitoring.

[0092] Exemplarily, the vibration signal is converted into a time-frequency signal using FFT transformation to obtain a characteristic vector of the signal.

[0093] In a further embodiment, the architecture of the CNN network used as the model framework is as follows Figure 5 shown.

[0094] The input layer receives the vibration signal of the input network, which can be a frequency domain feature representation of the signal after time-frequency conversion.

[0095] The convolution layer is the process of performing convolution calculation on the input signal to extract key features, which is specifically expressed as follows:

[0096]

[0097] in, Indicates Layer The convolution kernel weights, Indicates Layer The local area is convolved; Indicates the width of the convolution kernel.

[0098] The downsampling layer is used to reduce the dimension of the feature vector when the feature dimension obtained after the convolution operation is large. The main downsampling methods are the maximum method and the mean method, which are expressed as follows:

[0099]

[0100] in, Indicates Tier Frame No. The activation value of a neuron; Indicates the width of the pooling area; Indicates The weights corresponding to the neurons in the layer pooling.

[0101] The fully connected layer and the output layer connect the downsampling results in order to form a one-dimensional feature set. A classifier is connected after the fully connected layer for pattern recognition. In this embodiment, the Softmax function is selected as the classifier of the output layer.

[0102] The above model training process can be Figure 6 As shown in the figure, the transformer vibration signal is reconstructed using compressed sensing technology, and the original signal is reconstructed using a small amount of data. The reconstructed signal is converted into a frequency feature representation by time-frequency conversion, and then the frequency feature representation of the vibration signal is used to train the CNN network to obtain a winding state monitoring model.

[0103] In a further embodiment, when the signal set used for model training includes sufficient fault status data, the prediction output of the model can not only make a binary judgment on whether there is a fault, but also predict the specific state of the winding according to the specific characteristics of the input signal.

[0104] The above model training method can obtain a winding condition monitoring model with higher accuracy, small data transmission volume and accurate prediction. Figure 7 A method for monitoring transformer winding status based on compressed sensing using the above model according to an embodiment of the present invention is described.

[0105] See also Figure 7 An embodiment of the present invention provides a transformer winding state monitoring method based on compressed sensing, comprising the following steps:

[0106] Step S710: Collect transformer vibration signals.

[0107] Step S720: Reconstruct the transformer vibration signal using a compressed sensing method to obtain a reconstructed signal.

[0108] Step S730: Input the reconstructed signal into the trained winding state monitoring model to obtain a state monitoring result.

[0109] In a specific embodiment, as mentioned in the aforementioned embodiment, a compressed measurement matrix and a signal sparse transformation basis are designed according to the characteristics of the transformer vibration signal. Based on the compressed measurement matrix and the signal sparse transformation basis, a sparse signal reconstruction algorithm is used to reconstruct the transformer vibration signal to obtain a reconstructed signal. The specific signal reconstruction process can refer to the relevant parts of the aforementioned embodiment and will not be repeated here.

[0110] In a further embodiment, considering that the amplitude and phase information can better reflect the state of the winding, the frequency, amplitude and other characteristics generated by different vibration faults are different from the normal state. It is more meaningful for state monitoring to analyze the fault mechanism and locate the fault source based on the characteristic frequency and amplitude of the fault signal. Before inputting the reconstructed signal into the model for state prediction, the reconstructed signal is transformed in the time domain to obtain the frequency domain feature representation of the reconstructed signal. The frequency domain feature representation of the reconstructed signal is used as the input of the winding state monitoring model, so that the model predicts and classifies the winding state according to the frequency characteristics of the signal.

[0111] In a further embodiment, when the signal set used for model training includes sufficient fault status data, the prediction output of the model can not only make a binary judgment on whether there is a fault, but also predict the specific state of the winding according to the specific characteristics of the input signal.

[0112] The above-disclosed embodiments describe in detail a model training method and a transformer winding state monitoring method based on compressed sensing. The above-disclosed method can be implemented using various forms of equipment. Therefore, the present invention also discloses a device corresponding to the above-mentioned method. Specific embodiments are given below for detailed description.

[0113] See also Figure 8 One embodiment of the present invention provides a model training device, comprising:

[0114] A signal acquisition module 802 is used to acquire historical transformer vibration signals;

[0115] A first signal reconstruction module 804 is used to reconstruct the historical transformer vibration signal using a compressed sensing method to obtain a reconstructed signal set;

[0116] The model training module 806 is used to use the reconstructed signal set as the input of the convolutional neural network to perform model training to obtain a winding state monitoring model.

[0117] See also Fig. 9 One embodiment of the present invention provides a transformer winding state monitoring device, comprising:

[0118] A signal acquisition module 902 is used to acquire transformer vibration signals;

[0119] The second signal reconstruction module 904 is used to reconstruct the transformer vibration signal by using a compressed sensing method to obtain a reconstructed signal;

[0120] The state monitoring module 905 is used to input the reconstructed signal into the winding state monitoring model trained by the above-mentioned model training device to obtain the state monitoring result.

[0121] The methods and related devices mentioned in the above embodiments are described with reference to the method flow charts and / or structural diagrams provided in the embodiments of the present application. Specifically, each process and / or block in the method flow charts and / or structural diagrams, as well as the combination of processes and / or blocks in the flow charts and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to generate a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the process in the process. Figure 1 Schematic diagram of one or more processes and / or structures Figure 1 These computer program instructions can also be stored in a computer-readable memory that can guide a computer or other programmable data processing device to work in a specific way, so that the instructions stored in the computer-readable memory produce a product including an instruction device, which implements the functions specified in the process. Figure 1 Schematic diagram of one or more processes and / or structures Figure 1 These computer program instructions can also be loaded onto a computer or other programmable data processing device, so that a series of operation steps are executed on the computer or other programmable device to produce a computer-implemented process, so that the instructions executed on the computer or other programmable device provide for implementing the process in the process. Figure 1 A flow or multiple flows and / or structures illustrate the steps of the functions specified in one block or multiple blocks.

[0122] The following embodiments are described by taking the method applied to a computer device as an example. It can be understood that the computer device can be any device with computing and processing functions, and can be but not limited to a server or a personal laptop computer, etc. In one embodiment, the computer device can be an application server, and the application server can be a server for running an application to be tested.

[0123] See also Fig.10 , which shows a hardware block diagram of an electronic device, the electronic device is intended to represent various forms of digital computers, such as laptop computers, desktop computers, workstations, personal digital assistants, servers, blade servers, mainframe computers, and other suitable computers. The electronic device can also represent various forms of mobile devices, such as personal digital processing, cellular phones, smart phones, wearable devices, and other similar computing devices. The components shown herein, their connections and relationships, and their functions are merely examples and are not intended to limit the implementation of the present application described and / or required herein.

[0124] like Fig.10As shown, the electronic device includes: at least one processor 1, at least one communication interface 2, at least one memory 3 and at least one communication bus 4;

[0125] In the embodiment of the present application, the number of the processor 1, the communication interface 2, the memory 3, and the communication bus 4 is at least one, and the processor 1, the communication interface 2, and the memory 3 communicate with each other through the communication bus 4;

[0126] The processor 1 may be a central processing unit CPU, or an application-specific integrated circuit ASIC (Application Specific Integrated Circuit), or one or more integrated circuits configured to implement the embodiments of the present invention, etc.;

[0127] The memory 3 may include a high-speed RAM memory, and may also include a non-volatile memory (non-volatile memory), etc., such as at least one disk memory;

[0128] Among them, the memory stores a program, and the processor can call the program stored in the memory, and the program is used to: implement the model training method provided by the above-mentioned embodiment and / or any possible implementation method combined with the embodiment, or, a transformer winding state monitoring method based on compressed sensing.

[0129] An embodiment of the present invention also provides a readable storage medium having a computer program stored thereon. When the computer program is executed by a processor, it implements the model training method provided in the above embodiment and / or any possible implementation method in combination with the embodiment, or a transformer winding state monitoring method based on compressed sensing.

[0130] The above-mentioned embodiments have described the present invention in particular detail with respect to possible scenarios, and those skilled in the art will recognize that the present invention can be practiced through other embodiments. The specific naming of components, the capitalization of terms, attributes, data structures, or any other programming or structural aspects are not mandatory or important, and the mechanisms or features of the present invention may have different names, forms, or procedures. The system may be implemented by a combination of hardware and software (as described), entirely by hardware elements, or entirely by software elements. The specific division of functions between the various system components described herein is exemplary only and not mandatory; on the contrary, the functions performed by a single system component may be performed by multiple components, or the functions performed by multiple components may be performed by a single component.

[0131] Those skilled in the art should understand that the various steps of the above disclosed method can be implemented by a general computing device, they can be concentrated on a single computing device, or distributed on a network composed of multiple computing devices, optionally, they can be implemented with program codes executable by the computing device, so that they can be stored in a storage device and executed by the computing device, or they can be made into individual integrated circuit modules, or multiple modules or steps therein can be made into a single integrated circuit module for implementation. Thus, the embodiments of the present invention are not limited to any specific combination of hardware and software.

[0132] These computing device executable programs (also referred to as programs, software, software applications, or code) include machine instructions for programmable processors, and these computing programs can be implemented using high-level procedural and / or object-oriented programming languages, and / or assembly / machine languages. As used herein, the terms "machine-readable medium" and "computer-readable medium" refer to any computer program product, device, and / or device (e.g., disk, optical disk, memory, programmable logic device (PLD)) for providing machine instructions and / or data to a programmable processor, including a machine-readable medium that receives machine instructions as a machine-readable signal. The term "machine-readable signal" refers to any signal for providing machine instructions and / or data to a programmable processor.

[0133] Certain aspects of the present invention include process steps and instructions described herein in the form of algorithms. It should be noted that the process steps and instructions of the present invention can be implemented in software, firmware and / or hardware, and when implemented by software, it can be downloaded, stored on different platforms used by various operating systems and operated from the platforms.

[0134] Those skilled in the art will understand that the structures shown in the accompanying drawings are merely block diagrams of partial structures related to the scheme of the present application, and do not constitute a limitation on the terminal device to which the scheme of the present application is applied. The specific terminal device may include more or fewer components than shown in the figures, or combine certain components, or have a different arrangement of components.

[0135] In the description of this specification, the description with reference to the terms "one embodiment", "some embodiments", "example", "specific example", or "possible design" etc. means that the specific features, structures, materials or characteristics described in conjunction with the embodiment or example are included in at least one embodiment or example of the present application. In this specification, the schematic representations of the above terms do not necessarily refer to the same embodiment or example. Moreover, the specific features, structures, materials or characteristics described may be combined in any one or more embodiments or examples in a suitable manner. In addition, those skilled in the art may combine and combine the different embodiments or examples described in this specification and the features of the different embodiments or examples, without contradiction.

[0136] Finally, it should be noted that, in this article, relational terms such as first and second, etc. are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Moreover, the terms "include", "comprise" or any other variants thereof are intended to cover non-exclusive inclusion, so that a process, method, article or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, article or device. In the absence of further restrictions, the elements defined by the sentence "comprise a ..." do not exclude the presence of other identical elements in the process, method, article or device including the elements.

[0137] The above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit the same. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that the technical solutions described in the aforementioned embodiments may still be modified, or some of the technical features thereof may be replaced by equivalents. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A model training method, characterized in that: The model trained by the method is used for transformer winding condition monitoring based on compressed sensing, including: Obtain historical transformer vibration signals; Reconstructing the historical transformer vibration signal using a compressed sensing method to obtain a reconstructed signal set; The reconstructed signal set is used as the input of the convolutional neural network for model training to obtain a winding state monitoring model.

2. The method according to claim 1, characterized in that The step of using the reconstructed signal set as an input of a convolutional neural network for model training comprises: Performing a time domain transformation on the reconstructed signal to obtain a frequency domain feature representation of the reconstructed signal; The frequency domain feature representation of the reconstructed signal is used as the input of the convolutional neural network for model training.

3. The method according to claim 1, characterized in that The compressed sensing method is used to reconstruct the historical transformer vibration signal to obtain a reconstructed signal set including: According to the characteristics of transformer vibration signal, the compression measurement matrix and signal sparse transformation basis are designed; Based on the compressed measurement matrix and the signal sparse transformation basis, the historical transformer vibration signal is reconstructed using a sparse signal reconstruction algorithm to obtain a reconstructed signal set.

4. The method according to claim 3, characterized in that The sparse signal reconstruction algorithm is a generalized orthogonal matching pursuit algorithm.

5. A transformer winding state monitoring method based on compressed sensing, characterized in that: include: Collect transformer vibration signals; Reconstructing the transformer vibration signal using a compressed sensing method to obtain a reconstructed signal; The reconstructed signal is input into a winding state monitoring model trained by the model training method according to any one of claims 1 to 4 to obtain a state monitoring result.

6. The method according to claim 5, characterized in that Before inputting the reconstructed signal into the winding condition monitoring model, the method further includes: The reconstructed signal is transformed in the time domain to obtain a frequency domain feature representation of the reconstructed signal, and the frequency domain feature representation of the reconstructed signal is used as an input of the winding state monitoring model.

7. A model training device, characterized in that: include: A signal acquisition module, used to acquire historical transformer vibration signals; A first signal reconstruction module is used to reconstruct the historical transformer vibration signal by using a compressed sensing method to obtain a reconstructed signal set; The model training module is used to use the reconstructed signal set as the input of the convolutional neural network for model training to obtain a winding state monitoring model.

8. A transformer winding state monitoring device, characterized in that: include: A signal acquisition module, used to collect transformer vibration signals; The second signal reconstruction module is used to reconstruct the transformer vibration signal by using a compressed sensing method to obtain a reconstructed signal; The state monitoring module is used to input the reconstructed signal into the winding state monitoring model trained by the model training device described in claim 7 to obtain the state monitoring result.

9. An electronic device, characterized in that: include: Memory and processor; The memory is used to store programs; The processor is used to call the program stored in the memory to execute the model training method as described in any one of claims 1 to 4, or to execute the transformer winding state monitoring method based on compressed sensing as described in any one of claims 5 to 6.

10. A readable storage medium, characterized in that: A computer program is stored thereon, and when the computer program is executed by a processor, the model training method as described in any one of claims 1 to 4 is implemented, or the transformer winding state monitoring method based on compressed sensing as described in any one of claims 5 to 6 is executed.

Images