Capacitive equipment fault diagnosis method of improved YOLO model

Abstract

The invention provides a capacitive equipment fault diagnosis method based on an improved YOLO model, and relates to the technical field of electric power digital data processing, and the method comprises the steps: collecting an infrared image of capacitive equipment; denoising the image by using a median filtering algorithm; a YOLOv7 model is used to carry out identification and target extraction on the denoised image; clustering the processed infrared image by using an FCM algorithm; and the clustering result is evaluated by adopting an F-measure method. According to the method, efficient recognition of the capacitive equipment is achieved through image denoising and target extraction, clustering is conducted on the infrared images of the capacitive equipment through the FCM algorithm, the fault type is diagnosed, intelligent management and control of a power grid are achieved conveniently, the fault diagnosis efficiency is improved, and the risk of manual diagnosis is reduced.

Description

Technical Field

[0001] The present invention relates to the technical field of electrical digital data processing, and in particular to a capacitive device fault diagnosis method based on an improved YOLO model. Background Art

[0002] As the power system becomes increasingly complex, the importance of capacitive equipment continues to increase, and its insulation status has a decisive impact on the stable operation of the system. Traditional monitoring methods are no longer suitable for the dynamically changing operating environment, and often fail to detect insulation defects in a timely manner, increasing system risks. Capacitive equipment accounts for the majority of substation equipment, and its insulation medium may deteriorate due to multiple factors, which may lead to functional failure in severe cases, threatening safe operation and personnel safety. Therefore, real-time monitoring of the insulation status of capacitive equipment is particularly important.

[0003] Faced with the growth of power system scale and complexity, more intelligent and accurate monitoring methods are needed to improve equipment operation efficiency and safety. Deep learning and machine learning technologies provide new approaches for intelligent monitoring. This paper proposes an online monitoring method for capacitive equipment in the whole station based on the YOLOv7 algorithm to ensure the safe and stable operation of the power system. Summary of the invention

[0004] In view of the above existing problems, the present invention is proposed.

[0005] Therefore, the present invention provides a capacitive device fault diagnosis method based on an improved YOLO model, which can solve the problems of low efficiency and low accuracy of the existing capacitive device fault diagnosis method.

[0006] In order to solve the above technical problems, the present invention provides the following technical solutions, a capacitive device fault diagnosis method of an improved YOLO model, comprising: Collect infrared images of capacitive devices; preprocess the infrared images and use the median filter algorithm to denoise them; apply the YOLOv7 model to identify and extract devices in the preprocessed infrared images; apply the FCM mean clustering method to cluster the infrared images processed by the YOLOv7 model; introduce the F-measure evaluation indicator to evaluate the clustering results.

[0007] As a preferred solution of the capacitive device fault diagnosis method of the improved YOLO model described in the present invention, collecting infrared images of the capacitive device includes the following steps: automatically scanning and collecting data of the capacitive device by using an inspection robot with an infrared thermal imager.

[0008] As a preferred solution of the method for generating an adaptive user interface based on user preferences described in the present invention, the following steps are included: preprocessing the infrared image using the median filtering algorithm: performing preliminary noise reduction processing on the image using the median filtering algorithm, and the specific operations are: selecting a square filtering window according to the shape of the picture taken by the infrared imager; selecting a 3×3 filtering window to ensure that the filtered image retains more details; sorting all pixel values ​​in the window to generate a monotonous two-dimensional data sequence; taking the middle value from the sorted pixel values, and replacing the number of pixels of the current image with the median value.

[0009] As a preferred solution of the capacitive device fault diagnosis method of the improved YOLO model described in the present invention, the expression for selecting the middle value of the pixel point is: , in, Indicates the selected middle value, Med indicates the median operation, Represents the pixel value of the box diagram, N represents a natural number, m represents the window length, and v represents the number of samples contained on both sides of the window.

[0010] As a preferred solution of the capacitive device fault diagnosis method of the improved YOLO model described in the present invention, the application of the YOLOv7 model to identify and extract the device in the preprocessed infrared image includes the following steps: using the resize_image function to scale the input image and adjust the input image to a size of 640×640.

[0011] As a preferred solution of the capacitive device fault diagnosis method of the improved YOLO model described in the present invention, the SE module is added after the convolution layer of the traditional backbone network. The module consists of two main parts: Squeeze stage: The spatial information of the feature map is compressed into channel-level global information through global average pooling. The purpose of the squeeze stage is to compress the information of the spatial dimension into the information of the channel dimension. Given an input feature map , where H and W are the spatial dimensions of the feature map, and C is the number of channels. The global average pooling operation GAP converts the The dimensions are compressed to a scalar: , in, is the input feature map X at position and the value on channel c, is the global descriptor of channel c. In this way, we can get a C-dimensional vector , which represents the global information of each channel.

[0012] Excitation stage: Use the fully connected layer to generate the weight coefficient of each channel, then normalize the weight value through the Sigmoid activation function, and finally adjust the weight of each channel of the input feature map. This enables the model to automatically focus on feature channels with important diagnostic information, such as temperature gradients and texture features.

[0013] First, the global descriptor z is reduced in dimension (through an activation function such as ReLU) through a fully connected layer. Suppose it is reduced to an intermediate dimension (r is a scaling factor, usually set to 16 or 32): , in, is the reduced-dimensional weight matrix, is the bias term, is the intermediate activation vector.

[0014] Then, another fully connected layer is used to up-convert the dimension back to channel C, and the result is normalized to the range [0, 1] through the Sigmoid activation function: , in, is the dimension-enhanced weight matrix, is the bias term, It is a C-dimensional vector representing the weight coefficient of each channel.

[0015] The weight coefficients generated by the Excitation stage above It will act on the input feature map. Assuming the input feature map is X, the weight coefficient Multiply the feature map of each channel to get the weighted feature map : , in, is the original feature map at position and the value on channel c, is the channel weight obtained in the Excitation stage.

[0016] As a preferred solution of the capacitive device fault diagnosis method of the improved YOLO model described in the present invention, wherein: the image passes through 4 CBS modules, and the feature extraction process is optimized by introducing the Transformer module, the Transformer module is combined with the traditional convolutional layer, and the position encoding is used to enhance the model's sensitivity to local features while retaining its advantages in global context modeling. Through the self-attention mechanism, the model can establish long-distance dependencies between different positions of the input image, especially when the relationship between the capacitive device area and other parts is more complex, the Transformer can effectively extract these global information. The Transformer's multi-head self-attention mechanism enables the model to focus more on the heat map information of the target area, thereby improving the accuracy of image recognition; and through the ELAN module, the input and output of the convolutional layer are connected to generate more gradient flows. The feature map passes through the Head part, and the feature maps of different levels are fused and the structure is predicted. The specific operations are as follows: C5 passes through the SPPCSP module, and the number of channels is reduced to 512; it is fused with the C4 and C3 feature maps in top-down order to obtain the P3, P4, and P5 feature maps; it is fused with the P4 and P5 feature maps in bottom-up order; the fused feature map is mapped to different target variables through the convolution layer, including the category of capacitive devices, bounding box coordinates, and the probability of object existence; finally, the infrared feature map enters the Head module to convert the feature map into a target detection result.

[0017] As a preferred solution of the capacitive device fault diagnosis method of the improved YOLO model described in the present invention, the YOLO model requires a database to be trained, and the infrared images of the capacitive devices obtained based on network search are used to obtain image data for training the model. At the same time, infrared images are taken from various sides and multiple angles of various capacitive devices by an infrared imager to establish a data set for subsequent training of the model.

[0018] As a preferred solution of the capacitive device fault diagnosis method of the improved YOLO model described in the present invention, the YOLO model requires a database to be trained, and the infrared images of the capacitive devices obtained based on network search are used to obtain image data for training the model. At the same time, infrared images are taken from various sides and multiple angles of various capacitive devices by an infrared imager to establish a data set for subsequent training of the model.

[0019] As a preferred solution of the capacitive device fault diagnosis method of the improved YOLO model described in the present invention, the FCM mean clustering method has the following objective function constraints: , Where J represents the clustering error, C1 represents the number of clusters, and n represents the number of data points. Indicates the degree of membership of data point j to category i, represents the feature vector of data point j, p represents the fuzzy factor, It represents the number of categories. After clustering is completed, it is necessary to analyze the characteristics and fault modes represented by each cluster center in the clustering results. Each cluster center corresponds to the characteristic mode in the infrared image of the substation capacitive equipment.

[0020] After FCM clustering is completed, it is necessary to first analyze the characteristics and possible fault modes represented by each cluster center in the clustering results. Each cluster center corresponds to the characteristic pattern in the infrared image of the substation capacitive equipment, which can reflect the different states of the equipment in operation.

[0021] Based on the clustering results, the characteristics of each cluster are mapped to the known capacitive equipment failure modes. Based on the working principle and empirical knowledge of the substation capacitive equipment, determine which clusters represent the fault area. Among them, the high temperature area (the area with higher temperature in the clustering result) represents an overheating fault, and the change in the hot spot area indicates that there is a fault in a component inside the equipment (such as capacitor failure, cable overheating, poor contact of the switchgear, etc.). Using the membership information of each sample in the FCM clustering result, the potential fault area can be located. After clustering, the area with higher membership usually indicates that the area has stronger fault characteristics and may be the hot spot area of ​​the equipment.

[0022] As a preferred solution of the capacitive device fault diagnosis method of the improved YOLO model described in the present invention, the specific calculation formula for evaluating the clustering result using the F-measure evaluation index is as follows: , , , , Among them, P represents precision, R represents recall, AP represents average precision, rAP represents mean average precision, TP represents true positive, FP represents false positive, TN represents true negative, FN represents false negative, and t represents the number of categories in the data set.

[0023] Beneficial effects of the present invention: The present invention adopts a median filtering algorithm to filter the obtained infrared image. On the basis of preserving the basic information of the image, it can effectively suppress image noise and improve the accuracy of image recognition. The present invention adopts the YOLOv7 model to identify and extract targets from filtered images. The algorithm helps the model learn a more complete feature representation by adding an auxiliary Head during the training process, and quickly and efficiently identifies and extracts equipment in real time. The present invention uses the FCM mean clustering method to cluster infrared images processed by the YOLOv7 algorithm. This method improves the stability of clustering by effectively allocating boundary points between clusters and providing an excellent representation of coverage data. By using the membership information of each sample in the FCM clustering results, the potential fault area can be located, and the nature of the fault can be further determined in combination with the working conditions of the equipment, thereby effectively improving the fault detection and maintenance efficiency of substation capacitive equipment. BRIEF DESCRIPTION OF THE DRAWINGS

[0024] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings required for use in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other accompanying drawings can be obtained based on these accompanying drawings without paying creative work.

[0025] Figure 1 A flowchart of a capacitive device fault diagnosis method based on an improved YOLO model is provided according to an embodiment of the present invention.

[0026] Figure 2 A flowchart of an FCM mean clustering method for a capacitive device fault diagnosis method based on an improved YOLO model is provided in accordance with an embodiment of the present invention. DETAILED DESCRIPTION

[0027] In order to make the above-mentioned purposes, features and advantages of the present invention more obvious and easy to understand, the specific implementation methods of the present invention are described in detail below in conjunction with the drawings of the specification. Obviously, the described embodiments are part of the embodiments of the present invention, but not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary persons in the art without creative work should fall within the scope of protection of the present invention.

[0028] In the following description, many specific details are set forth to facilitate a full understanding of the present invention, but the present invention may also be implemented in other ways different from those described herein, and those skilled in the art may make similar generalizations without violating the connotation of the present invention. Therefore, the present invention is not limited to the specific embodiments disclosed below.

[0029] Secondly, the term "one embodiment" or "embodiment" as used herein refers to a specific feature, structure, or characteristic that may be included in at least one implementation of the present invention. The term "in one embodiment" that appears in different places in this specification does not necessarily refer to the same embodiment, nor does it refer to a separate or selective embodiment that is mutually exclusive with other embodiments.

[0030] The present invention is described in detail with reference to schematic diagrams. When describing the embodiments of the present invention, for the sake of convenience, the cross-sectional diagrams showing the device structure will not be partially enlarged according to the general scale, and the schematic diagrams are only examples, which should not limit the scope of protection of the present invention. In addition, in actual production, the three-dimensional dimensions of length, width and depth should be included.

[0031] At the same time, in the description of the present invention, it should be noted that the directions or positional relationships indicated by the terms "upper, lower, inner and outer" are based on the directions or positional relationships shown in the drawings, and are only for the convenience of describing the present invention and simplifying the description, rather than indicating or implying that the device or element referred to must have a specific direction, be constructed and operated in a specific direction, and therefore cannot be understood as limiting the present invention. In addition, the terms "first, second or third" are only used for descriptive purposes and cannot be understood as indicating or implying relative importance.

[0032] In the present invention, unless otherwise clearly specified and limited, the terms "install, connect, connect" should be understood in a broad sense, for example: it can be a fixed connection, a detachable connection or an integral connection; it can also be a mechanical connection, an electrical connection or a direct connection, or it can be indirectly connected through an intermediate medium, or it can be the internal communication of two components. For ordinary technicians in this field, the specific meanings of the above terms in the present invention can be understood according to specific circumstances.

[0033] Example 1 Reference Figure 1-Figure 2 , which is the first embodiment of the present invention, and provides a capacitive device fault diagnosis method based on an improved YOLO model, comprising: S1: Acquire infrared images of capacitive devices.

[0034] Specifically, an inspection robot equipped with an infrared imager is used to photograph the capacitive device along a set route and generate an infrared image thereof.

[0035] S2: Preprocess the infrared image and use the median filter algorithm to remove noise.

[0036] Specifically, in order to save computing resources, for each pixel in the infrared image, a 5×5 neighborhood window is selected; all pixels in the selected 5×5 neighborhood are arranged in ascending order according to the pixel value to generate a monotonous two-dimensional data sequence; the middle value is taken from the sorted pixel values, and the median value is used to replace the original pixel value of the current pixel, and the noise points in the original image are replaced by the median of the surrounding pixels, thereby achieving the purpose of smoothing the image; the above operations are repeated until the neighborhood window completes the traversal of all pixels in the image.

[0037] Specifically, the expression is: , in, Indicates the selected middle value, Med indicates the median operation, Represents the pixel value of the box diagram, N represents a natural number, m represents the window length, where m is 3, and v represents the number of samples contained on both sides of the window.

[0038] S3: Apply the YOLOv7 model to identify and extract devices in the preprocessed infrared image.

[0039] Specifically, the input image is scaled using the resize_image function to adjust the input image to 640×640 size; the adjusted image is input to the backbone network for further processing; after the convolution layer of the traditional backbone network, the SE module is added. This module consists of two main parts: Squeeze stage: The spatial information of the feature map is compressed into channel-level global information through global average pooling. The purpose of the squeeze stage is to compress the information of the spatial dimension into the information of the channel dimension. Given an input feature map , where H and W are the spatial dimensions of the feature map, and C is the number of channels. The global average pooling operation GAP converts the The dimensions are compressed to a scalar: , in, is the input feature map X at position and the value on channel c, is the global descriptor of channel c. In this way, we can get a C-dimensional vector , which represents the global information of each channel.

[0040] Excitation stage: Use the fully connected layer to generate the weight coefficient of each channel, then normalize the weight value through the Sigmoid activation function, and finally adjust the weight of each channel of the input feature map. This enables the model to automatically focus on feature channels with important diagnostic information, such as temperature gradients and texture features.

[0041] First, the global descriptor z is reduced in dimension (through an activation function such as ReLU) through a fully connected layer. Suppose it is reduced to an intermediate dimension (r is a scaling factor, usually set to 16 or 32): , in, is the reduced-dimensional weight matrix, is the bias term, is the intermediate activation vector.

[0042] Then, another fully connected layer is used to up-convert the dimension back to channel C, and the result is normalized to the range [0, 1] through the Sigmoid activation function: , in, is the dimension-enhanced weight matrix, is the bias term, It is a C-dimensional vector representing the weight coefficient of each channel.

[0043] The weight coefficients generated by the Excitation stage above It will act on the input feature map. Assuming the input feature map is X, the weight coefficient Multiply the feature map of each channel to get the weighted feature map : , in, is the original feature map at position and the value on channel c, is the channel weight obtained in the Excitation stage.

[0044] The image passes through 4 CBS modules, and the feature extraction process is optimized by introducing the Transformer module. The Transformer module is combined with the traditional convolutional layer, and position encoding is used to enhance the model's sensitivity to local features while retaining its advantage in global context modeling. Through the self-attention mechanism, the model can establish long-distance dependencies between different positions of the input image, especially when the relationship between the capacitive device area and other parts is more complex, the Transformer can effectively extract this global information. The Transformer's multi-head self-attention mechanism allows the model to focus more on the heat map information of the target area, thereby improving the accuracy of image recognition; and through the ELAN module, the input and output of the convolutional layer are connected to generate more gradient flows.

[0045] Specifically, the feature map passes through the Head part, the feature maps of different levels are fused, and the structure is predicted. The specific operations are as follows: C5 passes through the SPPCSP module, and the number of channels is reduced to 512; it is fused with the C4 and C3 feature maps in top-down order to obtain the P3, P4, and P5 feature maps; it is fused with the P4 and P5 feature maps in bottom-up order; the fused feature maps are mapped to different target variables through the convolution layer, including the category of capacitive devices, bounding box coordinates, and the probability of object existence.

[0046] Furthermore, the infrared feature map enters the Head module, and the feature map is converted into a target detection result. The specific operations are as follows: through the anchor box mechanism, the predicted box position and size of the capacitive device in the infrared image are partially defined; using the classification head, the category of the capacitive device in the image is identified by analyzing the feature map; using the regression head, the offset of the predicted box is predicted to more accurately locate the target object.

[0047] S4: Apply the FCM mean clustering method to cluster the infrared images processed by the YOLOv7 model.

[0048] Specifically, the FCM mean clustering method is used, such as Figure 2 As shown in the figure, the feature images extracted by the YOLO model are classified: the cluster center is calculated according to the current membership (the cluster center is the weighted average of the data points, and the weight is expressed by the membership); the membership is updated according to the current cluster center; the above operation is repeated until the objective function meets the conditions.

[0049] Specifically, the calculation formula of the cluster center C(k,d) is: , in, Indicates the membership of sample j to category i, It represents the value of data point i in feature dimension d, g represents the fuzzy factor, which is usually a real number greater than 1, and is 1.5 here.

[0050] The updated membership calculation formula is: , Among them, X(i) represents the feature vector of data point i, and C(k) represents the feature vector of cluster center k.

[0051] The objective function constraints are as follows: , Where J represents the clustering error, C1 represents the number of clusters, and n represents the number of data points. Indicates the degree of membership of data point j to category i, represents the feature vector of data point j, p represents the fuzzy factor, It represents the number of categories. After clustering is completed, it is necessary to analyze the characteristics and fault modes represented by each cluster center in the clustering results. Each cluster center corresponds to the characteristic mode in the infrared image of the substation capacitive equipment.

[0052] Specifically, after the FCM clustering is completed, it is first necessary to analyze the characteristics and fault modes represented by each cluster center in the clustering results. Each cluster center corresponds to the characteristic pattern in the infrared image of the substation capacitive equipment, which can reflect the different states of the equipment in operation. According to the clustering results, the characteristics of each cluster are mapped with the known fault modes of the capacitive equipment. According to the working principle and empirical knowledge of the substation capacitive equipment, it is judged which clusters may represent the fault area. Among them, the high temperature area (the area with higher temperature in the clustering result) represents the overheating fault, and the change of the hot spot area indicates that there is a fault in a certain component inside the equipment (such as capacitor failure, cable overheating, poor contact of switchgear, etc.). Using the membership information of each sample in the FCM clustering result, the potential fault area can be located. After clustering, the area with higher membership usually indicates that the area has stronger fault characteristics and may be the hot spot area of ​​the equipment.

[0053] S5: Introduce evaluation indicators to evaluate the clustering results.

[0054] Specifically, the clustering results are evaluated by selecting P (Precision), R (Recall), AP (Average Precision) and mAP (mean Average Precision) as references in the evaluation indicators. The specific calculation formula is as follows: , , , , Among them, P represents precision, R represents recall, AP represents average precision, rAP represents mean average precision, TP represents true positive, FP represents false positive, TN represents true negative, FN represents false negative, and t represents the number of categories in the data set.

[0055] The terms used to describe the positional relationship in the drawings are only used for illustrative purposes and should not be construed as limitations on this patent.

[0056] Obviously, the above embodiments of the present invention are only examples for clearly explaining the present invention, and are not intended to limit the implementation methods of the present invention. For those skilled in the art, other different forms of changes or modifications can be made based on the above description. It is not necessary and impossible to list all the implementation methods here. Any modifications, equivalent substitutions and improvements made within the spirit and principles of the present invention should be included in the protection scope of the claims of the present invention.

Claims

1. A capacitive device fault diagnosis method based on an improved YOLO model, characterized in that: include, Collect infrared images of capacitive devices; Preprocess the infrared image and use the median filter algorithm to remove noise; Apply the YOLOv7 model to identify and extract the preprocessed infrared image devices; The FCM mean clustering method is used to cluster the infrared images processed by the YOLOv7 model; The F-measure evaluation indicator is introduced to evaluate the clustering results.

2. A capacitive device fault diagnosis method based on an improved YOLO model as claimed in claim 1, characterized in that: By using an inspection robot equipped with an infrared thermal imager, capacitive devices are automatically scanned and data collected.

3. A capacitive device fault diagnosis method based on an improved YOLO model as claimed in claim 2, characterized in that: The median filter algorithm performs preliminary noise reduction on the image, and the specific operations are as follows: According to the shape of the picture taken by the infrared imager, a square filter window is selected; Select a 3×3 filter window and sort all pixel values ​​within the window to generate a monotonous two-dimensional data sequence; Take the middle value from the sorted pixel values ​​and replace the number of pixels in the current image with the middle value. The expression is: ， in, Indicates the selected middle value, Med indicates the median operation, Represents the pixel value of the box diagram, N represents a natural number, m represents the window length, and v represents the number of samples contained on both sides of the window.

4. A capacitive device fault diagnosis method based on an improved YOLO model as claimed in claim 3, characterized in that: The YOLOv7 model uses the resize_image function to scale the input image and adjust the input image to 640×640 size.

5. A capacitive device fault diagnosis method based on an improved YOLO model as claimed in claim 4, characterized in that: The adjusted image is input into the backbone network for processing; after the convolution layer of the backbone network, the SE module is added; In the Squeeze stage: the spatial information of the feature map is compressed into channel-level global information through global average pooling. Given an input feature map , where H and W are the spatial dimensions of the feature map, C is the number of channels, and the global average pooling operation GAP converts the The dimensions are compressed to a scalar: ， in, is the input feature map X at position and the value on channel c, is the global descriptor of channel c, and at the same time obtains a C-dimensional vector , represents the global information of each channel.

6. A capacitive device fault diagnosis method based on an improved YOLO model as claimed in claim 5, characterized in that: In the Excitation stage, a fully connected layer is used to generate the weight coefficient of each channel, the weight value is normalized by the Sigmoid activation function, and each channel of the input feature map is weighted and adjusted: ， ， ， in, , is the reduced-dimensional weight matrix, , is the bias term, is the intermediate activation vector, Is a C-dimensional vector, representing the weight coefficient of each channel, is the original feature map at position and the value on channel c, It is the channel weight obtained in the Excitation stage.

7. The capacitive device fault diagnosis method of the improved YOLO model as claimed in claim 6, characterized in that: After the image passes through four CBS modules, the self-attention mechanism of the Transformer module is introduced to optimize the feature extraction process. The Transformer module is combined with the convolutional layer, and the dependency relationship is established between different positions of the input image through the self-attention mechanism. The input and output of the convolutional layer are connected through the ELAN module.

8. The capacitive device fault diagnosis method of the improved YOLO model as claimed in claim 7, characterized in that: The feature map passes through the Head part, and the feature maps of different levels are fused and the structure is predicted. The specific operations are as follows: The C5 feature map is processed by the SPPCSP module, and the number of channels is reduced to 512. Then, the C5 feature map is fused with the C4 and C3 feature maps in a top-down order. At the same time, the feature maps are gradually fused from high level to low level to form feature maps of three different levels: P3, P4, and P5. In the bottom-up order, the P4 and P5 feature maps are fused again to ensure the complementarity of multi-level features; After the feature maps are fused, they are mapped to different target variables through convolutional layers, including the category of capacitive devices, bounding box coordinates, and the probability of object existence; Finally, the infrared feature map is input into the Head module, which converts all feature maps into the final target detection results.

9. The capacitive device fault diagnosis method of the improved YOLO model as claimed in claim 8, characterized in that: The YOLOv7 model requires a database for training. Based on the infrared images of capacitive devices obtained through network search, image data for training the model is obtained. At the same time, infrared images of various capacitive devices are obtained through infrared imagers to establish a data set for model training.

10. The capacitive device fault diagnosis method of the improved YOLO model according to claim 9, characterized in that: The objective function constraints of the FCM mean clustering method are as follows: ， Where J represents the clustering error, C1 represents the number of clusters, and n represents the number of data points. Indicates the degree of membership of data point j to category i, represents the feature vector of data point j, p represents the fuzzy factor, Represents the number of categories. After clustering is completed, it is necessary to analyze the characteristics and fault modes represented by each cluster center in the clustering results. Each cluster center corresponds to the characteristic mode in the infrared image of the substation capacitive equipment; Based on the clustering results, the characteristics of each cluster are mapped with the known capacitive device failure mode, and the fault area is determined; Among them, the area with temperature higher than the set threshold in the clustering result is the high temperature area, which represents overheating failure, and the change of the hot spot area indicates that there is a failure in the internal components of the equipment; The membership information of each sample in the FCM clustering result is used to locate the potential fault area.

11. The capacitive device fault diagnosis method of the improved YOLO model according to claim 10, characterized in that: The clustering results are evaluated, and the specific calculation formula for the evaluation is as follows: ， ， ， ， Among them, P represents precision, R represents recall, AP represents average precision, rAP represents mean average precision, TP represents true positive, FP represents false positive, TN represents true negative, FN represents false negative, and t represents the number of categories in the data set.

Images