Gas insulated switchgear partial discharge fault diagnosis method, system, device, medium and product

Abstract

The invention discloses a gas insulated switchgear partial discharge fault diagnosis method, system and device, a medium and a product, and relates to the field of fault diagnosis, and the method comprises the steps: obtaining a partial discharge map data set of a gas insulated switchgear; the partial discharge spectrum data set is preprocessed; performing image enhancement on the preprocessed partial discharge spectrum data set based on a generative adversarial network; training a diagnosis model according to the partial discharge spectrum data set after image enhancement; the diagnosis model comprises a convolutional neural network, a Transform model and a support vector machine which are connected in sequence; the Transform model and the support vector machine are both optimized by adopting a sparrow search algorithm; and carrying out partial discharge fault diagnosis on the gas insulated switchgear by using the trained diagnosis model. The method can effectively solve the problems of insufficient data, difficult feature extraction and low classification precision, and improves the precision and robustness of the diagnosis model.

Description

Technical Field

[0001] The present application relates to the field of fault diagnosis, and in particular to a method, system, equipment, medium and product for diagnosing partial discharge faults of gas-insulated switchgear. Background Art

[0002] Gas insulated switchgear (GIS) is a core component of high-voltage power transmission systems, and its internal partial discharge phenomenon is an important indicator of equipment aging or potential failure. Different types of partial discharge (such as tip discharge, air gap discharge, suspended discharge, and free metal particle discharge) have significant differences in characteristics. Traditional methods based on feature extraction and simple classifiers are difficult to capture complex partial discharge patterns, and problems such as insufficient data and class imbalance also limit the diagnostic capabilities of traditional models.

[0003] Therefore, in order to effectively solve the problems of insufficient data, difficult feature extraction and low classification accuracy, it is urgent to provide a new GIS partial discharge fault diagnosis method. Summary of the invention

[0004] The purpose of this application is to provide a method, system, equipment, medium and product for partial discharge fault diagnosis of gas insulated switchgear, which can effectively solve the problems of insufficient data, difficult feature extraction and low classification accuracy, and improve the accuracy and robustness of the diagnosis model.

[0005] To achieve the above objectives, this application provides the following solutions:

[0006] In a first aspect, the present application provides a method for diagnosing a partial discharge fault of a gas insulated switchgear, the method comprising:

[0007] Acquire a partial discharge spectrum data set of a gas insulated switchgear; the partial discharge spectrum data set includes partial discharge spectra of different discharge types;

[0008] Preprocessing the partial discharge spectrum data set; the preprocessing includes: grayscale processing and normalization processing;

[0009] Image enhancement is performed on the preprocessed partial discharge atlas dataset based on a generative adversarial network;

[0010] Training a diagnostic model based on the partial discharge atlas dataset after image enhancement; the diagnostic model includes a convolutional neural network, a Transformer model, and a support vector machine connected in sequence; the Transformer model and the support vector machine are both optimized using a sparrow search algorithm;

[0011] The trained diagnostic model is used to diagnose partial discharge faults in gas-insulated switchgear.

[0012] Optionally, the preprocessing of the partial discharge spectrum data set specifically includes:

[0013] Using Formula I gray (x,y)=0.2989·I R (x,y)+0.5870·I G (x,y)+0.1140·I B (x,y) is grayed out;

[0014] Using the formula Perform normalization processing;

[0015] Among them, I gray is the pixel value of the image after grayscale processing, I R ,I G and I B are the pixel values ​​of the red, green, and blue channels respectively, x, y are the pixel coordinates, I norm is the normalized image pixel value.

[0016] Optionally, the image enhancement of the preprocessed partial discharge atlas dataset based on the generative adversarial network also includes:

[0017] The generator of the generative adversarial network is used to generate generated samples with the same resolution as the real samples in the preprocessed partial discharge atlas dataset;

[0018] Use the discriminator of the generative adversarial network to evaluate the generated samples and obtain the scores of the generated samples;

[0019] According to the score of the generated sample, the Wasserstein distance method is used to calculate the gap between the generated sample and the real sample;

[0020] The generator is continuously optimized so that the gap between the generated samples and the real samples is smaller than the gap threshold.

[0021] Optionally, training a diagnostic model based on the partial discharge atlas dataset after image enhancement specifically includes:

[0022] Using the formula Determine the feature vector F output by the convolutional neural network that reflects the morphological differences of different discharge types CNN ;

[0023] Using formula F Transformer =FFN(Z)=ReLU(ZW1+b1)W2+b2 determines the output feature F of the transformer model Transformer ;

[0024] in, Represents the final feature map after multiple layers of convolution and pooling in the convolutional neural network. is the feature map after convolution, max is the maximum value in the pooling window, Z represents the output of the multi-head self-attention mechanism, W1 and W2 represent the weight matrix of the feedforward neural network FFN in the transformer model, b1 and b2 represent bias terms, ReLU is the activation function, which introduces nonlinear transformation, and Flatten is the Flatten operation, which is used to flatten the output feature map after each layer of convolution and pooling into a one-dimensional vector.

[0025] Optionally, the objective function of the support vector machine is:

[0026]

[0027] The constraints are

[0028] Among them, ω represents the weight vector of the hyperplane, x i is the output feature of the Transformer model, y i is a label indicating the discharge type; ω·x i For x i The projection to the hyperplane, b represents the bias term of the hyperplane, which determines the position of the hyperplane.

[0029] Optionally, the discharge types include: tip discharge, suspension discharge, free metal particle discharge and air gap discharge.

[0030] In a second aspect, the present application provides a gas insulated switchgear partial discharge fault diagnosis system, the gas insulated switchgear partial discharge fault diagnosis system comprising:

[0031] A data set acquisition module, used to acquire a partial discharge spectrum data set of a gas insulated switchgear; the partial discharge spectrum data set includes partial discharge spectra of different discharge types;

[0032] A data set preprocessing module, used for preprocessing the partial discharge spectrum data set; the preprocessing includes: grayscale processing and normalization processing;

[0033] An image enhancement module, used to perform image enhancement on the preprocessed partial discharge atlas dataset based on a generative adversarial network;

[0034] A diagnostic model training module is used to train a diagnostic model based on the partial discharge atlas dataset after image enhancement; the diagnostic model includes a convolutional neural network, a Transformer model and a support vector machine connected in sequence; the Transformer model and the support vector machine are both optimized using a sparrow search algorithm;

[0035] The fault diagnosis module is used to diagnose partial discharge faults of gas-insulated switchgear using the trained diagnosis model.

[0036] In a third aspect, the present application provides a computer device, comprising: a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the gas insulated switchgear partial discharge fault diagnosis method.

[0037] In a fourth aspect, the present application provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the method for diagnosing partial discharge faults of gas-insulated switchgear.

[0038] In a fifth aspect, the present application provides a computer program product, including a computer program, which implements the gas-insulated switchgear partial discharge fault diagnosis method when executed by a processor.

[0039] According to the specific embodiments provided in this application, this application has the following technical effects:

[0040] The present application provides a method, system, equipment, medium and product for diagnosing partial discharge faults of gas insulated switchgear. The diagnostic model includes a convolutional neural network, a Transformer model and a support vector machine connected in sequence; the Transformer model and the support vector machine are optimized using a sparrow search algorithm; multi-stage feature extraction and classification are performed by the diagnostic model, combined with data enhancement and intelligent optimization, which can effectively solve the problems of insufficient samples and category imbalance, thereby enhancing the adaptability of the model to actual data, improving the accuracy and robustness of the diagnostic model, and providing efficient and reliable technical support for online monitoring and intelligent maintenance of partial discharge of GIS, thereby effectively solving the problems of insufficient data, difficult feature extraction and low classification accuracy. BRIEF DESCRIPTION OF THE DRAWINGS

[0041] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the drawings required for use in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0042] Figure 1 This is a flow chart of a method for diagnosing partial discharge faults of a gas insulated switchgear in one embodiment of the present application;

[0043] Figure 2 This is a schematic diagram of the diagnostic model principle;

[0044] Figure 3 Schematic diagram of the distribution of classification results for various types of discharges. DETAILED DESCRIPTION

[0045] The following will be combined with the drawings in the embodiments of the present application to clearly and completely describe the technical solutions in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, not all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of this application.

[0046] In order to make the above-mentioned objects, features and advantages of the present application more obvious and easy to understand, the present application is further described in detail below with reference to the accompanying drawings and specific implementation methods.

[0047] In an exemplary embodiment, Figure 1 and Figure 2 As shown, a method for diagnosing partial discharge faults of gas insulated switchgear is provided, the method comprising the following S101 to S105. Wherein:

[0048] S101, obtaining a partial discharge spectrum data set of a gas insulated switchgear; the partial discharge spectrum data set includes partial discharge spectra of different discharge types; the discharge types include: tip discharge, suspension discharge, free metal particle discharge and air gap discharge;

[0049] Among them, an ultra-high frequency sensor (frequency band of 0.3-3GHz) is used to measure the partial discharge signal at the insulation gap of GIS. After the electrical signal is amplified and processed, a partial discharge spectrum is generated;

[0050] S102, preprocessing the partial discharge spectrum data set; the preprocessing includes: grayscale processing and normalization processing;

[0051] The specific process of preprocessing is:

[0052] The partial discharge spectrum data set is converted into a matrix using the imread() function of MATLAB. gray (x,y)=0.2989·I R (x,y)+0.5870·I G (x,y)+0.1140·I B(x, y) is grayscaled; after grayscale conversion, the value of each pixel is between 0 (black) and 255 (white).

[0053] Among them, I gray is the pixel value of the image after grayscale processing, I R ,I G and I B are the pixel values ​​of the red, green, and blue channels respectively, and x, y are the pixel coordinates;

[0054] Using the formula Normalize the pixel values ​​of the grayscale image to [0,1], that is, convert the pixel values ​​from 8-bit integer type (uint8, range [0,255]) to floating point type (float32);

[0055] Among them, I norm is the normalized image pixel value.

[0056] S103, performing image enhancement on the preprocessed partial discharge atlas dataset based on a generative adversarial network (WGAN);

[0057] In order to meet the specific range requirements of WGAN input, formula I is used WGAN =2I norm -1 will normalize the image to [0,1] and further convert it to the range of [-1,1];

[0058] The generator G of WGAN is used to input random noise z to generate Figure 1 The generated sample G(x) with the same resolution is generated; the discriminator D of WGAN is then used to evaluate G(x) and output the score D(G(x)) of the generated sample; finally, the Wasserstein distance is used to calculate the difference between the generated sample and the real sample.

[0059]

[0060] Among them, P real is the distribution of the real samples, P Z is the distribution of normal distribution noise, D(x) is the output of the discriminator, which means sample x (i.e., the input I WGAN )’s authenticity score, is the distribution P from real data real The expected value P of the sampled data x real is the image dataset distribution, we can sample N images x1,x2,...,xN, and then calculate: Denotes the fake data distribution P generated from the generator G Z The expected value of the data z sampled in, for the input z of the generator, assuming PZ If it is a standard normal distribution, then we can calculate by sampling N noise z1,z2,...,zN:

[0061] By optimizing the generator G, the generated sample G(x) is made as close to the real sample x as possible, that is, L is minimized, and the samples generated by the generator G cannot be distinguished by the discriminator D, thereby generating high-quality discharge images to expand the data set and ensure that the distribution of each type of data is balanced after enhancement, that is, the number of data sets of the four discharge type maps is the same.

[0062] S104, training a diagnostic model according to the partial discharge atlas dataset after image enhancement; the diagnostic model includes a convolutional neural network (CNN), a Transformer model and a support vector machine (SVM) connected in sequence; the Transformer model and the support vector machine are both optimized using a sparrow search algorithm (SSA);

[0063] In CNN, using the formula The image-enhanced partial discharge map dataset is converted to [0,1], and the spatial feature vector of the discharge map is output through the convolution layer, pooling layer and fully connected layer. The convolution operation extracts local features from the image-enhanced partial discharge map dataset, including the edge, texture, shape and other features of the discharge map. The convolution formula is:

[0064]

[0065] X l-1 is the input data (output of the previous layer), W l is the convolution kernel (weight), * represents the convolution operation, b1 represents the bias term, and σ represents the ReLU activation function.

[0066] The pooling layer reduces the dimension of the feature map after convolution, reduces the amount of data, and retains key information. It improves the translation invariance of the feature and reduces the computational complexity. The maximum pooling is used, that is, the maximum value in the pooling window is taken.

[0067]

[0068] The fully connected layer expands the convolution and pooling feature map into a feature vector F that can reflect the morphological differences of different discharge types. CNN , F CNN The feature dimension of F is determined by the CNN network structure. CNN As the input of the next transformer model.

[0069]

[0070] Represents the final feature map after multiple layers of convolution and pooling.

[0071] F CNN As input, it needs to be vectorized and positionally encoded in the transformer model first;

[0072] Using formula F CNN ={f1,f2,…,f N} CNN Convert to a sequence representation;

[0073] Among them, f i Represents the i-th feature vector output by CNN.

[0074] Using the formula Determine the positional encoding of the sequence representation;

[0075] Among them, pos is the position in the sequence representation, and D is the feature dimension. Through position encoding, Transformer can capture the position information of different features in the sequence.

[0076] Using formula F Transformer =F CNN +PE combines the position encoding with the feature F extracted by CNN CNN Add together to form the final input F Transformer ;

[0077] The input of the Transformer model is F Transformer , processed by multiple layers of self-attention mechanism and feed-forward neural network.

[0078] The Transformer model first calculates the correlation between different feature vectors through the self-attention mechanism to focus on the important features of the graph.

[0079]

[0080] Among them, Q represents the query matrix, K represents the key matrix, and V represents the value matrix, which are obtained by linear transformation of the input. k It represents the feature dimension of 9 (the same as the input feature dimension), which plays a scaling role to prevent the attention score from being too large.

[0081] The multi-head self-attention mechanism performs multiple independent attention operations and then concatenates the results. The output of the multi-head self-attention mechanism is recorded as Z, which has the same dimension as the input:

[0082] Z=MultiHead(Q,K,V)=Concat(head1,…,head h )W O ;

[0083] Where h represents the number of attention heads, W O It is the linear transformation weight matrix after the final concatenation.

[0084] head i =Attention(QW i Q ,KW i K ,VW i V );

[0085] Among them, W i Q ,W i K ,W i V Represents the parameter matrix for each head.

[0086] The feedforward neural network FFN processes the output Z from the multi-head self-attention mechanism, enhances the feature representation capability, and obtains the output of the transformer network.

[0087] F Transformer =FFN(Z)=ReLU(ZW1+b1)W2+b2;

[0088] Among them, Z represents the output of the multi-head self-attention mechanism, W1, W2 represent the weight matrix of FFN, which is obtained through back propagation and gradient descent training, b1, b2 represent bias terms, and ReLU is the activation function, which introduces nonlinear transformation.

[0089] The hyperparameter optimization process of the Transformer model is:

[0090] 1. Randomly initialize a set of Transformer model hyperparameters and randomly generate N sparrow individuals, each of which is a hyperparameter combination [d k ,n heads ,d hidden ,η,n layers ]. Among them, d k is the dimension of the key and query vector in the attention mechanism, N heads is the number of heads in multi-head attention, d hidden is the number of neurons in the hidden layer, η is the optimizer learning rate of the Transformer model, N layers为 The number of encoder stacking layers;

[0091] 2. Use SSA to iteratively search for the optimal hyperparameter combination:

[0092] First, the performance of each set of hyperparameters is evaluated through the fitness function (defined as the accuracy of the model on the validation set);

[0093] Fitness=Accuracy validation (d k ,N heads ,d hidden ,η,n layers );

[0094] Then, through the mechanism of discoverers (responsible for searching for food and having a large search range) and followers (observing the behavior of discoverers and conducting local searches), the hyperparameters are continuously optimized through iterative updates, the fitness function is repeatedly calculated, and the population position is updated until the fitness function is no longer significantly improved, thus finding the hyperparameter combination that maximizes Fitness. The final output is the optimized Transformer model.

[0095] The support vector machine (SVM) objective function is as follows:

[0096] The constraints are

[0097] Among them, ω represents the weight vector of the hyperplane, x i is the output feature of the Transformer model, y i is a label indicating the discharge type; ω·x i For x i The projection to the hyperplane, b represents the bias term of the hyperplane, which determines the position of the hyperplane.

[0098] The objective function means minimizing the norm of the hyperplane weight vector ω, which measures the complexity of the hyperplane. Minimizing the norm of the hyperplane weight vector ω is equivalent to maximizing the interval from the sample to the hyperplane, making the classification more robust and the generalization ability stronger.

[0099] In order to improve the SVM classification performance, SSA is used to optimize these hyperparameters C and γ.

[0100] SSA optimizes SVM, which is similar to SSA optimizes transformer model:

[0101] (1) Initialize the SVM hyperparameter combination. The SVM hyperparameters include the penalty coefficient C (which controls the model’s tolerance to misclassification) and the kernel function parameter γ (which affects the high-dimensional feature mapping effect).

[0102] (2) Use the fitness function (validation set accuracy) to evaluate hyperparameter performance.

[0103] (3) Output the optimal hyperparameter combination (C, γ).

[0104] The optimized SVM is used for fault diagnosis and the discharge type classification results are output.

[0105] Evaluate the classification accuracy, recall, precision, F1 score on the test set, and generate a confusion matrix.

[0106] (1) Accuracy: Accuracy refers to the overall accuracy of the model in classifying all samples.

[0107]

[0108] Where: TP i is the number of samples correctly classified as the i-th class, TN is the number of samples correctly classified as non-i-th class, FP is the number of samples incorrectly classified as i-th class, and FN is the number of samples not correctly classified as i-th class.

[0109] (2) Recall: The recall rate reflects the model’s ability to recognize samples of the target category.

[0110]

[0111] (3) Precision: Precision indicates the proportion of samples that are actually of a certain category among the samples predicted by the model.

[0112]

[0113] (4) F1 score: The F1 score is the harmonic mean of precision and recall, and is used to weigh the relationship between the two.

[0114]

[0115] S105, using the trained diagnostic model to perform partial discharge fault diagnosis on gas insulated switchgear

[0116] The following is an explanation through a specific example. The number of GIS partial discharge maps collected is 100 for tip discharge, 200 for air gap discharge, 150 for suspension discharge, and 50 for free metal particle discharge. WGAN generates 700 synthetic images, which are expanded to 1200, that is, 300 maps for each discharge type. The training set and test set are divided in a ratio of 7:3. Comparison of diagnostic results of different models

[0117] The diagnosis results of each type of discharge are shown in Table 1 and Table 2, where the classification results of each type of discharge are distributed and the sources of misclassification are analyzed. Figure 3 shown.

[0118] Table 1 Comparison of diagnostic results of different models

[0119] Model Accuracy Recall F1 score Original data + CNN-SVM 85.3％ 82.7％ 83.9％ Enhanced Data + CNN-Transformer-SVM 91.2％ 89.8％ 90.5％ Enhanced Data + CNN-Transformer-SVM + SSA 94.2％ 94.2％ 94.2％

[0120] Table 2

[0121] Discharge type Accuracy Recall F1 score Tip discharge 91.4％ 94.4％ 92.9％ Air Gap Discharge 93.5％ 95.6％ 94.5％ Suspension discharge 95.5％ 94.4％ 94.9％ Free metal particle discharge 96.5％ 92.2％ 94.3％

[0122] Based on the same inventive concept, the embodiment of the present application also provides a gas insulated switchgear partial discharge fault diagnosis system for implementing the gas insulated switchgear partial discharge fault diagnosis method involved above. The implementation scheme for solving the problem provided by the system is similar to the implementation scheme recorded in the above method, so the specific limitations in one or more gas insulated switchgear partial discharge fault diagnosis system embodiments provided below can refer to the limitations of the gas insulated switchgear partial discharge fault diagnosis method above, and will not be repeated here.

[0123] In an exemplary embodiment, a gas insulated switchgear partial discharge fault diagnosis system is provided, comprising:

[0124] A data set acquisition module, used to acquire a partial discharge spectrum data set of a gas insulated switchgear; the partial discharge spectrum data set includes partial discharge spectra of different discharge types;

[0125] A data set preprocessing module, used for preprocessing the partial discharge spectrum data set; the preprocessing includes: grayscale processing and normalization processing;

[0126] An image enhancement module, used to perform image enhancement on the preprocessed partial discharge atlas dataset based on a generative adversarial network;

[0127] A diagnostic model training module is used to train a diagnostic model based on the partial discharge atlas dataset after image enhancement; the diagnostic model includes a convolutional neural network, a Transformer model and a support vector machine connected in sequence; the Transformer model and the support vector machine are both optimized using a sparrow search algorithm;

[0128] The fault diagnosis module is used to diagnose partial discharge faults of gas-insulated switchgear using the trained diagnosis model.

[0129] In an exemplary embodiment, a computer device is provided, which may be a server or a terminal. The computer device includes a processor, a memory, an input / output interface (I / O for short) and a communication interface. The processor, the memory and the input / output interface are connected via a system bus, and the communication interface is connected to the system bus via the input / output interface. The processor of the computer device is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system, a computer program and a database. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The input / output interface of the computer device is used to exchange information between the processor and an external device. The communication interface of the computer device is used to communicate with an external terminal via a network connection. When the computer program is executed by the processor, a method for diagnosing partial discharge faults of a gas-insulated switchgear is implemented.

[0130] In an exemplary embodiment, a computer-readable storage medium is provided, storing a computer program, and when the computer program is executed by a processor, the steps in the above method embodiments are implemented.

[0131] In an exemplary embodiment, a computer program product is provided, including a computer program, and when the computer program is executed by a processor, the steps in the above method embodiments are implemented.

[0132] It should be noted that the user information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data used for analysis, stored data, displayed data, etc.) involved in this application are all information and data authorized by the user or fully authorized by all parties, and the collection, use and processing of relevant data must comply with relevant regulations.

[0133] Those skilled in the art can understand that all or part of the processes in the above-mentioned embodiment methods can be completed by instructing the relevant hardware through a computer program, and the computer program can be stored in a non-volatile computer-readable storage medium. When the computer program is executed, it can include the processes of the embodiments of the above-mentioned methods. Among them, any reference to the memory, database or other medium used in the embodiments provided in the present application can include at least one of non-volatile and volatile memory. Non-volatile memory can include read-only memory (ROM), magnetic tape, floppy disk, flash memory, optical memory, high-density embedded non-volatile memory, resistive random access memory (ReRAM), magnetoresistive random access memory (MRAM), ferroelectric random access memory (FRAM), phase change memory (PCM), graphene memory, etc. Volatile memory can include random access memory (RAM) or external cache memory, etc. By way of illustration and not limitation, RAM may be in various forms, such as static random access memory (SRAM) or dynamic random access memory (DRAM).

[0134] The database involved in each embodiment provided in this application may include at least one of a relational database and a non-relational database. The non-relational database may include a distributed database based on blockchain, etc., but is not limited thereto. The processor involved in each embodiment provided in this application may be a general-purpose processor, a central processing unit, a graphics processor, a digital signal processor, a programmable logic device, a data processing logic device based on quantum computing, etc., but is not limited thereto.

[0135] In this application, all actions to obtain signals, information or data are carried out in compliance with the relevant data protection laws and policies of the country where they are located and with the authorization given by the owner of the corresponding device.

[0136] The technical features of the above embodiments may be combined arbitrarily. To make the description concise, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0137] This article uses specific examples to illustrate the principles and implementation methods of this application. The description of the above embodiments is only used to help understand the method and core ideas of this application. At the same time, for those skilled in the art, according to the ideas of this application, there will be changes in the specific implementation methods and application scope. In summary, the content of this specification should not be understood as limiting this application.

Claims

1. A method for diagnosing partial discharge faults of gas insulated switchgear, characterized in that: The gas insulated switchgear partial discharge fault diagnosis method comprises: Acquire a partial discharge spectrum data set of a gas insulated switchgear; the partial discharge spectrum data set includes partial discharge spectra of different discharge types; Preprocessing the partial discharge spectrum data set; the preprocessing includes: grayscale processing and normalization processing; Image enhancement is performed on the preprocessed partial discharge atlas dataset based on a generative adversarial network; Training a diagnostic model based on the partial discharge atlas dataset after image enhancement; the diagnostic model includes a convolutional neural network, a Transformer model, and a support vector machine connected in sequence; the Transformer model and the support vector machine are both optimized using a sparrow search algorithm; The trained diagnostic model is used to diagnose partial discharge faults in gas-insulated switchgear.

2. The method for diagnosing partial discharge faults of gas insulated switchgear according to claim 1, characterized in that: The preprocessing of the partial discharge spectrum data set specifically includes: Using Formula I gray (x,y)=0.2989·I R (x,y)+0.5870·I G (x,y)+0.1140·I B (x,y) is grayed out; Using the formula Perform normalization processing; Among them, I gray is the pixel value of the image after grayscale processing, I R ,I G and I B are the pixel values ​​of the red, green, and blue channels respectively, x, y are the pixel coordinates, I norm is the normalized image pixel value.

3. The method for diagnosing partial discharge faults of gas insulated switchgear according to claim 1, characterized in that: The image enhancement of the preprocessed partial discharge atlas dataset based on the generative adversarial network also includes: The generator of the generative adversarial network is used to generate generated samples with the same resolution as the real samples in the preprocessed partial discharge atlas dataset; Use the discriminator of the generative adversarial network to evaluate the generated samples and obtain the scores of the generated samples; According to the score of the generated sample, the Wasserstein distance method is used to calculate the gap between the generated sample and the real sample; The generator is continuously optimized so that the gap between the generated samples and the real samples is smaller than the gap threshold.

4. The method for diagnosing partial discharge faults of gas insulated switchgear according to claim 1, characterized in that: Based on the image-enhanced partial discharge atlas dataset, the diagnostic model is trained, including: Using the formula Determine the feature vector F output by the convolutional neural network that reflects the morphological differences of different discharge types CNN ; Using formula F Transformer =FFN(Z)=ReLU(ZW1+b1)W2+b2 determines the output feature F of the transformer model Transformer ; in, Represents the final feature map after multiple layers of convolution and pooling in the convolutional neural network. is the feature map after convolution, max is the maximum value in the pooling window, Z is the output of the multi-head self-attention mechanism, W1 and W2 are the weight matrices of the feedforward neural network FFN in the transformer model, b1 and b2 are bias terms, ReLU is the activation function, which introduces nonlinear transformation, and Flatten is the Flatten operation, which is used to flatten the output feature map after each layer of convolution and pooling into a one-dimensional vector.

5. The method for diagnosing partial discharge faults of gas insulated switchgear according to claim 1, characterized in that: The objective function of the support vector machine is: The constraints are Among them, ω represents the weight vector of the hyperplane, x i is the output feature of the Transformer model, y i is a label indicating the discharge type; ω·x i For x i The projection to the hyperplane, b represents the bias term of the hyperplane, which determines the position of the hyperplane.

6. The method for diagnosing partial discharge faults of gas insulated switchgear according to claim 5, characterized in that: Discharge types include: tip discharge, suspension discharge, free metal particle discharge and air gap discharge.

7. A gas insulated switchgear partial discharge fault diagnosis system, characterized in that: The gas insulated switchgear partial discharge fault diagnosis system comprises: A data set acquisition module, used to acquire a partial discharge spectrum data set of a gas-insulated switchgear; the partial discharge spectrum data set includes partial discharge spectra of different discharge types; A data set preprocessing module, used for preprocessing the partial discharge spectrum data set; the preprocessing includes: grayscale processing and normalization processing; An image enhancement module, used to perform image enhancement on the preprocessed partial discharge atlas dataset based on a generative adversarial network; A diagnostic model training module is used to train a diagnostic model based on the partial discharge atlas dataset after image enhancement; the diagnostic model includes a convolutional neural network, a Transformer model and a support vector machine connected in sequence; the Transformer model and the support vector machine are both optimized using a sparrow search algorithm; The fault diagnosis module is used to diagnose partial discharge faults of gas-insulated switchgear using the trained diagnosis model.

8. A computer device comprising: A memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the method for diagnosing partial discharge faults of gas-insulated switchgear according to any one of claims 1 to 6.

9. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the method for diagnosing partial discharge faults of a gas-insulated switchgear according to any one of claims 1 to 6 is implemented.

10. A computer program product, comprising a computer program, characterized in that When the computer program is executed by a processor, the method for diagnosing partial discharge faults of a gas-insulated switchgear according to any one of claims 1 to 6 is implemented.

Images