GIS partial discharge mode identification method and device, product, medium and equipment

Abstract

The invention discloses a GIS partial discharge mode recognition method and device, a product, a medium and equipment, and relates to the technical field of partial discharge classification recognition, and the method comprises the steps: obtaining the partial discharge data of a GIS; normalizing the partial discharge data and converting the partial discharge data into a grayscale image; inputting the grayscale image into a partial discharge mode recognition model, and outputting a partial discharge mode recognition result of the GIS; the partial discharge mode recognition model is obtained by introducing SKNet and a coordinate attention mechanism which are sequentially connected in series before residual connection of a first residual module and a second residual module of a ResNet model framework. According to the GIS partial discharge mode identification method and device, the product, the medium and the equipment, the accuracy of GIS partial discharge mode identification can be effectively improved.

Description

Technical Field

[0001] The present application relates to the technical field of partial discharge classification and identification, and in particular to a GIS partial discharge pattern recognition method, device, product, medium and equipment. Background Art

[0002] Gas Insulated Switchgear (GIS) is a key equipment widely used in modern power systems. It has a complex structure and a high working field strength. Once insulation defects occur inside GIS, it is very easy to cause equipment failure, causing large-scale power outages and huge economic losses. Insulation failure has always had an important impact on the reliability of GIS operation. Partial discharge, as an early manifestation of insulation failure inside GIS, is a potential threat to the normal operation of GIS. If partial discharge failures can be discovered and eliminated as early as possible during equipment operation, equipment losses can be reduced and power supply reliability can be improved. However, since GIS is a purely sealed structure, it is difficult to carry out various preventive tests like traditional open equipment to timely discover potential defects. Masoud Karimi et al. in Iran introduced deep belief networks into the field of partial discharge identification and found that the classifier based on deep belief networks can learn important distinguishable features from data and effectively distinguish three types of corona discharge, surface discharge and internal discharge. However, this method did not study the sub-classification of discharge faults, nor did it classify the discharge intensity (high, medium, and low), which limited its judgment of the severity of discharge in practical applications. Summary of the invention

[0003] The purpose of this application is to provide a GIS partial discharge pattern recognition method, device, product, medium and equipment, which can effectively improve the accuracy of GIS partial discharge pattern recognition.

[0004] To achieve the above objectives, this application provides the following solutions:

[0005] In a first aspect, the present application provides a GIS partial discharge pattern recognition method, comprising:

[0006] Obtain partial discharge data of GIS;

[0007] Normalize the partial discharge data and convert them into grayscale images;

[0008] The grayscale image is input into a partial discharge pattern recognition model, and the partial discharge pattern recognition result of the GIS is output; the partial discharge pattern recognition model is obtained by introducing a sequentially serialized SKNet and a coordinate attention mechanism before the residual connection of the first residual module and the second residual module of the ResNet model framework.

[0009] Optionally, the obtaining of partial discharge data of GIS specifically includes:

[0010] The phase-resolved partial discharge data of GIS is collected by using a UHF sensor to obtain the partial discharge data of GIS.

[0011] Optionally, the normalizing the partial discharge data and converting it into a grayscale image specifically includes:

[0012] Organizing the partial discharge data into a two-dimensional data matrix and normalizing it to obtain a normalized data matrix;

[0013] The data in the normalized data matrix are multiplied by 255 and converted to integer format, and then converted to grayscale images through the image processing library.

[0014] Optionally, the ResNet model includes a convolutional layer, a maximum pooling layer, a first residual module, a second residual module, an average pooling layer and a fully connected layer connected in sequence.

[0015] Optionally, the number of channels of the first residual module is 64; the number of channels of the second residual module is 128.

[0016] Optionally, the local discharge pattern recognition results of the GIS include: signals that normally do not generate local discharge, interference signals, high-intensity signals of solid suspended local discharge, medium-intensity signals of solid suspended local discharge, low-intensity signals of solid suspended local discharge, high-intensity signals of tip corona local discharge, medium-intensity signals of tip corona local discharge, low-intensity signals of tip corona local discharge, high-intensity signals of metal particle local discharge, medium-intensity signals of metal particle local discharge, low-intensity signals of metal particle local discharge, high-intensity signals of suspended potential local discharge, medium-intensity signals of suspended potential local discharge and low-intensity signals of suspended potential local discharge.

[0017] In a second aspect, the present application provides a GIS partial discharge pattern recognition device, comprising:

[0018] Data acquisition module, used to acquire partial discharge data of GIS;

[0019] A grayscale image acquisition module, used for normalizing partial discharge data and converting it into a grayscale image;

[0020] The local discharge pattern recognition result output module is used to input the grayscale image into the local discharge pattern recognition model and output the local discharge pattern recognition result of the GIS; the local discharge pattern recognition model is obtained by introducing the sequentially serialized SKNet and the coordinate attention mechanism before the residual connection of the first residual module and the second residual module of the ResNet model framework.

[0021] In a third aspect, the present application provides a computer program product, including a computer program, which implements the GIS local discharge pattern recognition method when executed by a processor.

[0022] In a fourth aspect, the present application provides a computer-readable storage medium having a computer program stored thereon, which implements the GIS local discharge pattern recognition method when executed by a processor.

[0023] In a fifth 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 GIS local discharge pattern recognition method.

[0024] According to the specific embodiments provided in this application, this application has the following technical effects:

[0025] The present application provides a GIS local discharge pattern recognition method, device, product, medium and equipment, which mainly relates to a GIS local discharge pattern recognition method based on a deep learning algorithm. The local discharge pattern recognition model of the present application is constructed by introducing the sequentially connected SKNet and the coordinate attention mechanism CoordAttention before the residual connection of the first residual module and the second residual module of the ResNet model framework. The local discharge pattern recognition model of the present application is used to output the local discharge pattern recognition result of the GIS, which can effectively improve the accuracy of the GIS local discharge pattern recognition. BRIEF DESCRIPTION OF THE DRAWINGS

[0026] In order to more clearly illustrate the technical solutions in the embodiments of the present application or related technologies, 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.

[0027] Figure 1 A schematic diagram of the process flow of the GIS partial discharge pattern recognition method provided in this application;

[0028] Figure 2 Schematic diagram of the ResNet model framework structure provided for this application;

[0029] Figure 3 Schematic diagram of the SKNet model framework structure provided for this application;

[0030] Figure 4 Schematic diagram of the implementation process of the coordinate attention mechanism provided for this application;

[0031] Figure 5A schematic diagram of the classification effect of the ResNet18 model provided in this application; wherein, Figure 5 (a) is the fault type classification diagram of the ResNet18 model; Figure 5 (b) is the fault type and intensity classification diagram of the ResNet18 model;

[0032] Figure 6 A schematic diagram of the classification effect of the ResNet model provided in this application; wherein, Figure 6 (a) is the fault type classification diagram of the ResNet model; Figure 6 (b) is the fault type and intensity classification diagram of the ResNet model;

[0033] Figure 7 A schematic diagram of the classification effect of the partial discharge pattern recognition model provided in this application; wherein, Figure 7 (a) is the fault type classification diagram of the partial discharge pattern recognition model; Figure 7 (b) is the fault type and intensity classification diagram of the partial discharge pattern recognition model. DETAILED DESCRIPTION

[0034] 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.

[0035] 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 in conjunction with the accompanying drawings and specific implementation methods.

[0036] In view of the technical defects that the discharge fault cannot be subdivided and the discharge intensity (high, medium and low) cannot be classified in the background technology, the present application provides a GIS local discharge pattern recognition method based on deep learning, in which a local discharge pattern recognition model is constructed, and the local discharge pattern recognition model is used based on the local discharge data to perform depth and feature extraction, and the local discharge fault types in the GIS are classified, including 6 data types of signals that do not normally generate local discharge, interference signals, solid suspension local discharge signals, tip corona local discharge signals, metal particle local discharge signals and suspended potential local discharge signals. On this basis, in order to divide the discharge intensity, each fault type is further subdivided into high-intensity discharge, medium-intensity discharge and low-intensity discharge, which are divided into 14 types of modes. Using the local discharge pattern recognition model of the present application, effective classification of two modes of 6 classifications and 14 classifications can be achieved, and the local discharge pattern recognition model can directly output one of the 6 classifications or one of the 14 classifications.

[0037] like Figure 1 As shown, the GIS partial discharge pattern recognition method disclosed in the present application includes:

[0038] Step 1: Obtain partial discharge data of GIS.

[0039] This application uses a UHF sensor to collect phase-resolved partial discharge (PRPD) data from GIS to obtain GIS partial discharge data, also known as PRPD data. PRPD data is a conventional data form collected for partial discharge equipment, which can effectively reflect the number and intensity of partial discharges under different phase conditions.

[0040] Step 2: Normalize the PD data and convert them into grayscale images.

[0041] After collecting PRPD data, the GIS partial discharge pattern recognition method of the present application first organizes the PRPD data into a two-dimensional data matrix, and in order to effectively convert the data into a grayscale image, the two-dimensional data matrix is ​​normalized to obtain a normalized data matrix. This operation scales the amplitude of the signal to the range of [0,1], which facilitates the subsequent grayscale image generation, wherein the normalization formula is:

[0042]

[0043] Among them, normalized_data is the normalized data matrix; PD_signal is the PRPD data in the two-dimensional data matrix; PD_min and PD_max are the minimum and maximum values ​​of the PRPD data in the two-dimensional data matrix respectively.

[0044] By normalizing, the contrast of the grayscale image can be enhanced, making the signal features more obvious in the grayscale image. The normalized data matrix can be converted into a grayscale image through an image processing library. In this step, the data in the normalized data matrix is ​​multiplied by 255 and converted to an integer format (such as uint8) to adapt to the pixel value range of the grayscale image (0 to 255), so that the final generated grayscale image can clearly show the spatial distribution and intensity of partial discharge activity.

[0045] Step 3: Input the grayscale image into the partial discharge pattern recognition model, and output the partial discharge pattern recognition result of the GIS; the partial discharge pattern recognition model is obtained by introducing the sequentially serialized SKNet and the coordinate attention mechanism before the residual connection of the first residual module and the second residual module of the ResNet model framework.

[0046] After obtaining the grayscale image, the grayscale image is transmitted to the ResNet model framework. The ResNet model framework is as follows Figure 2 As shown, it specifically includes sequentially connected convolution layers, maximum pooling layers, a first residual module with 64 channels, a second residual module with 128 channels, an average pooling layer, and a fully connected layer. After average pooling in the average pooling layer, the fully connected layer converts the local discharge pattern recognition result of GIS into the classification quantity dimension for output prediction. Specifically, the grayscale image enters the convolution module (the first residual module and the second residual module) after passing through the convolution layer and the maximum pooling layer. In order to improve the deep learning ability of PRPD data, this application introduces a neural network based on a selective convolution kernel (Selective Kernel Network, SKNet) and a coordinate attention mechanism (CoordAttention) in the two residual modules. SKNet and CoordAttention are sequentially connected in series and placed before the residual connection of the two residual modules.

[0047] For SKNet, its specific structure and feature extraction process are as follows Figure 3 As shown in the figure, the feature variable R input in SKNet is the deep feature extracted in the first residual module of the ResNet model. First, two forms of group convolution are performed based on the feature variable R. and The convolution kernel sizes are 3×3 and 5×5 respectively. After convolution transformation, the feature variable is obtained go through After convolution transformation, the feature variable is obtained In the task of local discharge pattern recognition, PRPD data may involve complex spatial and temporal features, such as discharge intensity, waveform characteristics, and noise disturbance. These features have different scales, so multi-scale convolution can help extract features under different receptive fields and effectively capture local and global information. This multi-branch convolution operation enables each branch to use a different convolution kernel scale for feature extraction, ensuring that the local discharge pattern recognition model for pattern recognition can process complex and hierarchical signals.

[0048] Afterwards and Perform element-wise summation to obtain the eigenvector U:

[0049]

[0050] After the global average pooling operation F gp The spatial scale H of the eigenvector U s ×W s Compress and obtain the channel feature vector S:

[0051]

[0052] Among them, F gp represents the global average pooling operation; H s represents the height of the eigenvector U; W s represents the width of the feature vector U; (i, j) represents the pixel position in the feature vector U.

[0053] Partial discharge signals may show different information at different scales, and this information is complementary. The features extracted by different convolution kernels are superimposed, and certain forms of feature transformation and fusion are performed, so that the partial discharge pattern recognition model does not only rely on the information in a single receptive field. Partial discharge data often contains both subtle high-frequency features and global pattern information. The above operations can process this diverse data feature and improve recognition accuracy.

[0054] Then, the fully connected layer is used to reduce the dimension to obtain feature Z, and then the weights a and b of each feature scale of feature Z are obtained through dimension increase and softmax operation, and then combined with them in the Select stage. and The weighted sum is performed to obtain the final learned feature V. Through the Select operation, SKNet can dynamically select the optimal feature extraction path during the training process. This adaptive feature selection mechanism enables SKNet to automatically identify the most significant feature patterns in the discharge signal and amplify these features, thereby improving the ability to distinguish different discharge types.

[0055] For CoordAttention, the implementation process is as follows Figure 4 As shown in the figure, the input of CoordAttention is the final learned feature V obtained from SKNet. First, the input feature V is averaged pooled from the X-axis and Y-axis to obtain the feature V in width W. c and height H c The feature maps in two directions are used to obtain the accurate position information encoding of PRPD data. The specific formula is as follows:

[0056]

[0057] in, and Respectively represent the compression characteristics of V along the height and width; H c Indicates the height of V; W c represents the width of V; (h, p) and (q, w) represent the pixel positions in V, where The operation compresses the width-wise features by The operation compresses the height direction characteristics.

[0058] The feature maps of the two directions of the global receptive field obtained in the height and width scales are spliced ​​together, and the number of channels is compressed through the convolution module to reduce the amount of calculation. After batch normalization and nonlinear activation function, the feature map f is obtained. h and f w , and then the feature map is converted into attention weights along the width and height directions through convolution operation and sigmoid function:

[0059]

[0060] Among them, g h and g w are the attention weights of height (X axis) and width (Y axis), respectively; σ represents the sigmoid function; F h and F w Represents a convolution operation along the height and width.

[0061] Finally, V is multiplied and weighted to obtain the final feature vector with attention weights:

[0062] Y(α,β)=V(α,β)×g h (α)×g w (β)(6)

[0063] Among them, Y represents the learned features of channel attention, and (α, β) represents the position of feature vector and attention weight.

[0064] The coordinate attention mechanism of this application independently encodes the features in the horizontal and vertical directions and then fuses them, so that the local discharge pattern recognition model has higher flexibility and accuracy in the spatial dimension. The local discharge pattern often has specific spatial distribution characteristics. The use of this mechanism can better capture these spatial features, thereby improving the accuracy of classification.

[0065] In summary, this application effectively combines SKNet and the coordinate attention mechanism, and through the global fusion of features in different directions, enables the local discharge pattern recognition model to better fuse the features of each discharge pattern, improve the local discharge pattern recognition model's feature extraction capability for local discharge data, and enhance the classification effect. In addition, this application uses a simplified ResNet model, which reduces the number of parameters and computational overhead while maintaining a high feature extraction capability, and is suitable for deployment in resource-limited environments.

[0066] As a specific embodiment, the present application also compares the local discharge pattern recognition model with the ResNet18 model and the simplified ResNet in various aspects, and the obtained data are shown in Tables 1 and 2.

[0067] Table 1 Comparison of computational complexity and parameter quantity of different models

[0068] Model Computational complexity Model parameter quantity ResNet18 model 1.745G 11.177M Simplified ResNet model 456.503M 309.198K Partial Discharge Pattern Recognition Model 474.516M 325.510K

[0069] Table 2 Fault classification accuracy of different models

[0070] Model Fault classification Fault intensity classification ResNet18 model 96.89％ 79.32％ Simplified ResNet model 77.28％ 37.37％ Partial Discharge Pattern Recognition Model 98.83％ 88.02％

[0071] It can be found from Table 1 that the simplified ResNet model has significantly reduced model parameters and computational complexity compared to the ResNet18 model. The lightweight model has a great promoting effect on the practical application in power equipment. However, it can be seen from Table 2 that although the simplified ResNet significantly reduces the model volume, the model's classification ability for different fault types is also significantly reduced, which restricts the model's feature learning ability for complex partial discharge data. The partial discharge pattern recognition model of this application is based on the simplified ResNet and adds a key attention module, which only slightly increases the computational complexity and model parameters, but significantly improves the feature learning ability of the partial discharge pattern recognition model.

[0072] The partial discharge pattern recognition model of the present application directly outputs one of the fault codes 0-5 (or 0-13) of the 6 categories (or 14 categories). The statistical effect diagram of all the predicted categories of each model under the test condition is as follows: Figures 5 to 7 As shown. Among them, for the confusion matrix, the corresponding relationship between the category label and the mode type of partial discharge is:

[0073] (1) Six types of partial discharge data: 0-normal signal without partial discharge, 1-interference signal, 2-solid suspension partial discharge signal, 3-tip corona partial discharge signal, 4-metal particle partial discharge signal, 5-suspended potential partial discharge signal.

[0074] (2) 14 types of partial discharge data: 0-normal signal without partial discharge, 1-interference signal, 2-solid suspension partial discharge high intensity signal, 3-solid suspension partial discharge medium intensity signal, 4-solid suspension partial discharge low intensity signal, 5-tip corona partial discharge high intensity signal, 6-tip corona partial discharge medium intensity signal, 7-tip corona partial discharge low intensity signal, 8-metal particle partial discharge high intensity signal, 9-metal particle partial discharge medium intensity signal, 10-metal particle partial discharge low intensity signal, 11-suspended potential partial discharge high intensity signal, 12-suspended potential partial discharge medium intensity signal, 13-suspended potential partial discharge low intensity signal.

[0075] By observing the pattern recognition effect of the ResNet18 model for 6 types of partial discharge data, it can be found that only there are pattern misjudgments between categories 2 and 3, which can basically meet the feature learning and pattern recognition of various partial discharge faults. For the pattern recognition of 14 types of partial discharge data, the accuracy of more categories is less than 70%, and the classification effect is low. For the simplified ResNet model, the feature extraction ability of the model is significantly reduced. For the pattern recognition of 6 types of partial discharge data and 14 types of partial discharge data, the effect is difficult to meet the fault classification requirements of practical applications. The partial discharge pattern recognition model provided by this application not only ensures the lightweight of the model, but also improves the model's key feature learning ability for partial discharge data. Among the 14 classification effects, there are only more misjudgments between categories 2 and 3, and the other categories are all 80% or above. The model can achieve effective division of various partial discharge data, proving that the model of this application can correctly identify different fault types and meet the actual application requirements of partial discharge scenarios.

[0076] Compared with the existing partial discharge pattern recognition technology, the GIS partial discharge pattern recognition method of this application has the following advantages:

[0077] 1) This application integrates SKNet into the residual module of the simplified ResNet model, enhances the flexibility of feature extraction through selective convolution kernels with different receptive fields, and makes the local discharge pattern recognition model more advantageous in local discharge feature extraction through multi-scale information fusion.

[0078] (2) By introducing CoordAttention, the local discharge pattern recognition model can simultaneously focus on the global information of space and channels, and combine the position-sensitive coordinate attention mechanism to strengthen feature expression, thereby improving the local discharge pattern recognition model's ability to extract complex spatiotemporal features in local discharge data.

[0079] (3) The local discharge pattern recognition model of this application uses group convolution, selective convolution and dimensionality reduction design in SKNet and CoordAttention, and adopts a simplified ResNet model. It effectively reduces the computational complexity and parameter quantity while ensuring the performance of the local discharge pattern recognition model, which is conducive to the lightweight deployment of the model.

[0080] Based on the GIS partial discharge pattern recognition method, the present application also provides a GIS partial discharge pattern recognition device, including:

[0081] The data acquisition module is used to acquire the partial discharge data of GIS.

[0082] The grayscale image acquisition module is used to normalize the partial discharge data and convert it into a grayscale image.

[0083] The local discharge pattern recognition result output module is used to input the grayscale image into the local discharge pattern recognition model and output the local discharge pattern recognition result of the GIS; the local discharge pattern recognition model is obtained by introducing the sequentially serialized SKNet and the coordinate attention mechanism before the residual connection of the first residual module and the second residual module of the ResNet model framework.

[0084] In some embodiments, the present application also provides a computer program product, including a computer program, which implements the GIS local discharge pattern recognition method when executed by a processor.

[0085] In some embodiments, the present application also provides a computer-readable storage medium having a computer program stored thereon, and when the computer program is executed by a processor, the GIS partial discharge pattern recognition method is implemented.

[0086] In some embodiments, the present application further provides a computer device, including a processor, a memory, an input / output interface (I / O for short), a communication interface, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the GIS partial discharge pattern recognition method. Specifically, the operating device of the GIS partial discharge pattern recognition method provided in the present application is a computer, the processor version is i5-12600KF, and the graphics card version is RTX-3060Ti.

[0087] The processor, memory and input / output interface are connected through a system bus, and the communication interface is connected to the system bus through 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 database of the computer device is used to store pending transactions. 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 through a network connection. When the computer program is executed by the processor, the GIS partial discharge pattern recognition method is implemented.

[0088] It should be noted that the object information (including but not limited to object device information, object 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 object or fully authorized by all parties, and the collection, use and processing of relevant data must comply with relevant laws, regulations and standards.

[0089] 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 (ReadOnlyMemory, ROM), magnetic tape, floppy disk, flash memory, optical memory, high-density embedded non-volatile memory, resistive random access memory (ReRAM), magnetic 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. As an illustration and not limitation, RAM can be in various forms, such as static random access memory (SRAM) or dynamic random access memory (DRAM). The database involved in each embodiment provided in this application may include at least one of a relational database and a non-relational database. Non-relational databases may include distributed databases based on blockchains, etc., but are not limited to this. 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 are not limited to this.

[0090] 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.

[0091] 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 GIS partial discharge pattern recognition method, characterized in that: include: Obtain partial discharge data of GIS; Normalize the partial discharge data and convert them into grayscale images; The grayscale image is input into a partial discharge pattern recognition model, and the partial discharge pattern recognition result of the GIS is output; the partial discharge pattern recognition model is obtained by introducing a sequentially serialized SKNet and a coordinate attention mechanism before the residual connection of the first residual module and the second residual module of the ResNet model framework.

2. The GIS partial discharge pattern recognition method according to claim 1, characterized in that: The obtaining of partial discharge data of GIS specifically includes: The phase-resolved partial discharge data of GIS is collected by using a UHF sensor to obtain the partial discharge data of GIS.

3. The GIS partial discharge pattern recognition method according to claim 1 is characterized in that: The normalizing of the partial discharge data and converting the data into a grayscale image specifically includes: Organizing the partial discharge data into a two-dimensional data matrix and normalizing it to obtain a normalized data matrix; The data in the normalized data matrix are multiplied by 255 and converted to integer format, and then converted to grayscale images through the image processing library.

4. The GIS partial discharge pattern recognition method according to claim 1 is characterized in that: The ResNet model includes a convolutional layer, a maximum pooling layer, a first residual module, a second residual module, an average pooling layer and a fully connected layer which are connected in sequence.

5. The GIS partial discharge pattern recognition method according to claim 4 is characterized in that: The number of channels of the first residual module is 64; the number of channels of the second residual module is 128.

6. The GIS partial discharge pattern recognition method according to claim 1 is characterized in that: The local discharge pattern recognition results of the GIS include: signals without normal local discharge, interference signals, high-intensity signals of solid suspended local discharge, medium-intensity signals of solid suspended local discharge, low-intensity signals of solid suspended local discharge, high-intensity signals of tip corona local discharge, medium-intensity signals of tip corona local discharge, low-intensity signals of tip corona local discharge, high-intensity signals of metal particle local discharge, medium-intensity signals of metal particle local discharge, low-intensity signals of metal particle local discharge, high-intensity signals of suspended potential local discharge, medium-intensity signals of suspended potential local discharge and low-intensity signals of suspended potential local discharge.

7. A GIS partial discharge pattern recognition device, characterized in that: include: Data acquisition module, used to acquire partial discharge data of GIS; A grayscale image acquisition module, used for normalizing partial discharge data and converting it into a grayscale image; The local discharge pattern recognition result output module is used to input the grayscale image into the local discharge pattern recognition model and output the local discharge pattern recognition result of the GIS; the local discharge pattern recognition model is obtained by introducing the sequentially serialized SKNet and the coordinate attention mechanism before the residual connection of the first residual module and the second residual module of the ResNet model framework.

8. A computer program product, comprising a computer program, characterized in that When the computer program is executed by a processor, the GIS partial discharge pattern recognition method described in any one of claims 1 to 7 is implemented.

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 GIS partial discharge pattern recognition method described in any one of claims 1 to 7 is implemented.

10. 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 GIS partial discharge pattern recognition method according to any one of claims 1 to 7.

Images