Power transformation equipment monitoring method and system based on image super-resolution
By using image super-resolution processing and fusion technology, the problem of insufficient image quality in power equipment monitoring has been solved, enabling more accurate fault identification and early warning level determination, and improving the monitoring effect.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-11-20
- Publication Date
- 2026-04-07
AI Technical Summary
Poor image quality during power equipment monitoring affects the detection and monitoring of equipment defects.
A substation equipment monitoring method based on image super-resolution is adopted. The image is super-resolution processed by interpolation, convolutional neural network, generative adversarial network and Transformer network, and the image is fused. Finally, it is compared with the fault database to determine the fault category and monitoring and early warning level.
It improves image clarity, enhances the monitoring effect of power equipment, and enables more accurate identification of faults and determination of corresponding early warning levels.
Smart Images
Figure CN121810490A_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of power equipment monitoring technology, and more specifically to a power equipment monitoring method based on image super-resolution. Background Technology
[0002] Image resolution refers to the number of pixels in an image. Higher resolution results in better image clarity and detail. Image super-resolution is an image processing technique designed to increase the resolution of an image, thereby increasing the number of pixels and making it visually clearer. Image super-resolution has a wide range of applications, including digital photography, medical imaging, image enhancement, and satellite imagery. In the monitoring of power substation equipment, equipment identification is typically performed using screenshots from video surveillance footage. However, the image quality of these screenshots is often poor, which may affect subsequent equipment defect detection and monitoring. Therefore, a power substation equipment monitoring method based on image super-resolution is needed to perform super-resolution operations on images of average quality, thereby facilitating better monitoring of power substation equipment. Summary of the Invention
[0003] The purpose of this invention is to provide a method and system for monitoring substation equipment based on image super-resolution. This method can perform super-resolution operation on images of general quality to facilitate better monitoring of substation equipment.
[0004] To achieve the above objectives, embodiments of the present invention provide a method for monitoring substation equipment based on image super-resolution, the method comprising: Acquire images of the power equipment; Super-resolution processing is performed on the image using a difference method; The image is reconstructed using a convolutional neural network; The image is then subjected to super-resolution reconstruction using a generative adversarial network. The image is then subjected to super-resolution reconstruction using a Transformer network. The image after super-resolution processing by multiple processing models is acquired, and the image is fused to obtain a fused image; Based on the comparison results between the fused image and the fault database, the corresponding fault category is obtained; Based on the fault category and the warning value mapped to the time and location when the corresponding fault category is obtained, the monitoring and warning level of the response is determined.
[0005] Optionally, super-resolution processing of the image is performed using an interpolation method, including: Acquire an image of the power equipment, use the image as the source image, and select the pixels in the source image as floating-point coordinates; Obtain the nearest 4x4 neighboring pixels around the given floating-point coordinates; The weights of the bicubic interpolation are calculated using a cubic polynomial kernel function. The pixels within the nearest 4x4 neighborhood of the floating-point coordinates are traversed, and a weighted summation using bicubic interpolation is performed based on the obtained weights to obtain the pixel value of the target point mapped by the floating-point coordinates, thereby completing the super-resolution processing of the image.
[0006] Optionally, the image can be super-resolution reconstructed using a convolutional neural network; Images of the power equipment are acquired as training images, and the size of the training images is reduced to 1 / K of the original size; The scaled-down training images are fed into the SRCNN network, and the same channels are rearranged by pixel rearrangement to obtain the output image magnified by K times. The original image is compared with the magnified image as the ground truth, and the SRCNN network is trained using the L2 loss function. After training is completed, the images that need to be super-resolution processed are fed into the SRNCT network to complete the super-resolution processing of the images.
[0007] Optionally, super-resolution reconstruction processing of the image is performed using a generative adversarial network, including: Images of the power equipment are acquired as training images and fed into the SRGAN network for training. The generator of the SRGAN network performs super-resolution on the training images and transmits the results to the discriminator; The discriminator classifies the output images to complete the training of the SRGAN network; During training, the SRGAN network is trained in a supervised manner by using a weighted average of the L2 loss function and the generative adversarial loss. After training is complete, the images that need super-resolution processing are fed into the SRGAN network to complete the super-resolution processing of the images.
[0008] Optionally, the image after super-resolution processing by multiple processing models is acquired, and the image is fused to obtain a fused image, including: Images obtained after super-resolution processing using multiple processing models are aligned with the images containing the same scene. The aligned image is cut into segments to obtain segmented image blocks, and image blocks containing the same scene are aligned. The aligned image blocks are obtained, and the sharpness of each image block is determined by the gradient function method; Based on the sharpness of the aligned image blocks, determine the weight value of each aligned image block; Based on the weight values of the aligned image blocks, a weighted algorithm is used to fuse the aligned image blocks to obtain a fused image block. The merged image blocks are stitched together to obtain the merged image.
[0009] Optionally, based on the comparison results between the fused image and the fault database, the corresponding fault category is obtained, including: The fused image is acquired and converted into a fused feature vector to obtain a vector of the form 1*a; The fault feature vectors in the fault database are vertically grouped, with each group containing multiple fault feature vectors, and each fault feature vector having the form a*1. The fault feature vectors of each group are concatenated into a feature matrix of form n*a; The fused feature vector is multiplied by the feature matrix to obtain the corresponding similarity. Obtain the maximum similarity for each group and determine whether the maximum similarity is greater than a preset threshold; If the value is greater than a preset threshold, the fused image is determined to match the corresponding fault feature vector of the group, and the fault category corresponding to the image is obtained.
[0010] Optionally, based on the comparison results between the fused image and the fault database, the corresponding fault category is obtained, including: If the value is less than a preset threshold, it is determined that no fault was detected in the fused image, and the substation equipment image is reacquired and monitored.
[0011] Optionally, based on the fault category and the warning value mapped to the time and location when the corresponding fault category was acquired, the monitoring and warning level is determined, including: Obtain the fault category and the warning value mapped to the time and location when the corresponding fault category is obtained; Determine the weight of each fault category and its corresponding time and location; The warning index is obtained by weighting and summing the obtained weights and warning values. Based on the warning index, the corresponding monitoring and warning level is determined, which includes no risk, low risk, medium risk, and high risk.
[0012] On the other hand, the present invention also provides a substation equipment monitoring system based on image super-resolution, the substation equipment monitoring system comprising: The image acquisition module is used to acquire images of the substation equipment; The monitoring module is used to execute a substation monitoring method based on image super-resolution as described above, based on the acquired images of the substation equipment.
[0013] Through the above technical solution, the present invention provides a method and system for monitoring substation equipment based on image super-resolution. This involves acquiring images of the substation equipment, performing super-resolution processing on the images using interpolation, and simultaneously performing super-resolution processing using convolutional neural networks, generative adversarial networks, and Transformer networks. After obtaining the super-resolution images through these methods, the images obtained from the multiple processing models are fused to obtain a fused image. These multiple processing models can include the aforementioned interpolation method, convolutional neural networks, generative adversarial networks, and Transformer networks. This image fusion method fully considers the advantages and disadvantages of the super-resolution images processed by each model, resulting in a more uniform image. After obtaining the fused image, it is compared with a fault database to determine the corresponding fault category. Based on the fault category and the corresponding warning value mapped to the time and location, the monitoring and warning level can be determined. This method for monitoring power equipment can perform super-resolution processing on images of general quality to enable better monitoring of power equipment.
[0014] Other features and advantages of the embodiments of the present invention will be described in detail in the following detailed description section. Attached Figure Description
[0015] The accompanying drawings are provided to further illustrate embodiments of the present invention and form part of the specification. They are used together with the following detailed description to explain the embodiments of the present invention, but do not constitute a limitation thereof. In the drawings: Figure 1 This is a flowchart of a method for monitoring substation equipment based on image super-resolution according to an embodiment of the present invention; Figure 2 This is a flowchart illustrating a method for monitoring substation equipment based on image super-resolution according to an embodiment of the present invention, involving super-resolution operation through difference. Figure 3 This is a flowchart illustrating a method for monitoring substation equipment based on image super-resolution according to an embodiment of the present invention, involving super-resolution operation via a convolutional neural network. Figure 4 This is a flowchart illustrating a method for monitoring substation equipment based on image super-resolution according to an embodiment of the present invention, involving super-resolution operation via a generative adversarial network. Figure 5 This is a flowchart of an image fusion method for monitoring substation equipment based on image super-resolution according to an embodiment of the present invention; Figure 6 This is a flowchart illustrating the acquisition of fault categories in a substation monitoring method based on image super-resolution according to an embodiment of the present invention. Figure 7 This is a flowchart illustrating the determination of early warning levels in a method for monitoring substation equipment based on image super-resolution according to an embodiment of the present invention. Detailed Implementation
[0016] The specific embodiments of the present invention will be described in detail below with reference to the accompanying drawings. It should be understood that the specific embodiments described herein are for illustration and explanation only and are not intended to limit the scope of the present invention.
[0017] Figure 1 This is a flowchart of a substation equipment monitoring method based on image super-resolution according to an embodiment of the present invention. In this invention, the substation equipment monitoring process may include: In step S1, an image of the power equipment is acquired.
[0018] In step S2, super-resolution processing is performed on the image using the interpolation method.
[0019] In step S3, super-resolution reconstruction of the image is performed using a convolutional neural network.
[0020] In step S4, super-resolution reconstruction of the image is performed using a generative adversarial network.
[0021] In step S5, the image is reconstructed using a Transformer network.
[0022] In step S6, images after super-resolution processing by multiple processing models are acquired, and the images are fused to obtain a fused image.
[0023] In step S7, the corresponding fault category is obtained based on the comparison results between the fused image and the fault database.
[0024] In step S8, the monitoring and warning level of the response is determined based on the fault category and the warning value mapped to the time and location when the corresponding fault category is obtained.
[0025] In this invention, when monitoring power equipment, images of the equipment can be acquired. These images can then undergo super-resolution processing via interpolation. Simultaneously, super-resolution processing can be performed using convolutional neural networks, generative adversarial networks, and Transformer networks. After obtaining the super-resolution images through these methods, the images obtained from the multiple processing models can be fused to obtain a fused image. These multiple processing models can include interpolation, convolutional neural networks, generative adversarial networks, and Transformer networks. This image fusion method fully considers the advantages and disadvantages of the super-resolution images processed by each model, resulting in a more uniform image. After obtaining the fused image, it can be compared with a fault database. Based on the comparison result, the corresponding fault category can be obtained. After obtaining the fault category, the monitoring and warning level can be determined based on the fault category and the warning value mapped to the corresponding time and location. This method for monitoring power equipment can perform super-resolution processing on images of general quality to enable better monitoring of power equipment.
[0026] In this invention, such as Figure 2 As shown, the process of super-resolution operation through interpolation can include: In step S9, an image of the power equipment is acquired and used as the source image, and the pixels on the source image are selected as floating-point coordinates.
[0027] In step S10, the nearest 4*4 neighboring pixels around the floating-point coordinates are obtained.
[0028] In step S11, the weights of the bicubic interpolation are calculated using a cubic polynomial kernel function.
[0029] In step S12, the nearest 4*4 neighborhood pixels around the floating-point coordinates are traversed, and the weighted sum of bicubic interpolation is performed using the obtained weights to obtain the pixel value of the target point mapped by the floating-point coordinates, thereby completing the super-resolution processing of the image.
[0030] In this invention, after acquiring an image of the power equipment, the image can be used as the source image, and pixels on the source image can be selected as floating-point coordinates. Then, the nearest 4x4 neighborhood pixels around the floating-point coordinates can be obtained. The weights of bicubic interpolation are calculated using a cubic polynomial kernel function. After obtaining the weights, the pixels within the nearest 4x4 neighborhood of the floating-point coordinates can be traversed, and a weighted sum of the bicubic interpolation values is performed using the obtained weights. This yields the pixel value of the target point mapped from the floating-point coordinates, thus completing the super-resolution processing of the image.
[0031] In one embodiment of the present invention, such as Figure 3 As shown, the process of performing super-resolution operations using convolutional neural networks can include: In step S13, an image of the substation equipment is acquired as a training image, and the size of the training image is reduced to 1 / K of the original size.
[0032] In step S14, the scaled-down training image is fed into the SRCNN network, and the same channel is rearranged by pixel rearrangement to obtain the output image magnified by K times.
[0033] In step S15, the original image is compared with the magnified image as the true value, and the SRCNN network is trained using the L2 loss function.
[0034] In step S16, after training is completed, the image that needs to be super-resolution processed is sent into the SRNCT network to complete the super-resolution processing of the image.
[0035] In this invention, during super-resolution reconstruction, images of the substation equipment can be acquired as training images, which can then be reduced to 1 / K of their original size. After reduction, the reduced training images are fed into the SRCNN network, and pixel rearrangement is used to rearrange the same channels, resulting in an output image magnified K times. During training, the original image is compared with the magnified image as the ground truth, and the SRCNN network is trained using the L2 loss function. After training, the images requiring super-resolution processing are fed into the SRNCT network to complete the super-resolution reconstruction.
[0036] In one embodiment of the present invention, such as Figure 4 As shown, the process of performing super-resolution operations using generative adversarial networks can include: In step S17, images of the substation equipment are acquired as training images and fed into the SRGAN network for training.
[0037] In step S18, the generator of the SRGAN network performs super-resolution on the training images and transmits the results to the discriminator.
[0038] In step S19, the discriminator discriminates the output image to complete the training of the SRGAN network.
[0039] In step S20, the SRGAN network is trained under supervision by using a weighted average of the L2 loss function and the adversarial loss during the training process.
[0040] In step S21, after training is completed, the image that needs to be super-resolution processed is fed into the SRGAN network to complete the super-resolution processing of the image.
[0041] In this invention, during super-resolution reconstruction, images of substation equipment can be acquired as training images and fed into an SRGAN network for training. The generator of the SRGAN network performs super-resolution on the training images and sends the results to the discriminator. The discriminator judges the output images to complete the training of the SRGAN network. During training, a weighted average of the L2 loss function and the generative adversarial loss can be used as a determined loss function to supervise the training of the SRGAN network. After training, images requiring super-resolution processing can be fed into the SRGAN network to complete the super-resolution processing.
[0042] In one embodiment of the present invention, such as Figure 5 As shown, the image fusion process may include: In step S22, images after super-resolution processing by multiple processing models are acquired, and images containing the same scene are aligned.
[0043] In step S23, the aligned image is cut to obtain cut image blocks, and image blocks containing the same scene are aligned.
[0044] In step S24, aligned image blocks are obtained, and the sharpness of each image block is determined by the gradient function method.
[0045] In step S25, the weight value of each aligned image block is determined based on the sharpness of the aligned image blocks.
[0046] In step S26, the aligned image blocks are fused using a weighted algorithm based on their weight values to obtain a fused image block.
[0047] In step S27, the fused image blocks are stitched together to obtain a fused image.
[0048] In this invention, after acquiring images processed by multiple processing models for super-resolution, images containing the same scene can be aligned to ensure pixel correspondence within each image. After alignment, the aligned images can be segmented to obtain segmented image blocks, and these image blocks containing the same scene can be aligned again. After obtaining the aligned image blocks, the sharpness of each block can be determined using the gradient function method. Based on the sharpness of the aligned image blocks, a weight value for each aligned image block can be determined. After obtaining the weight values, the aligned image blocks can be fused using a weighted algorithm to obtain a fused image block. The fused image blocks can then be stitched together to obtain a fused image.
[0049] In one embodiment of the present invention, such as Figure 6 As shown, the process for obtaining the fault category may include: In step S28, the fused image is acquired and converted into a fused feature vector to obtain a vector of the form 1*a; In step S29, the fault feature vectors in the fault database are vertically grouped, with each group containing multiple fault feature vectors, and each fault feature vector having the form a*1. In step S30, the fault feature vectors of each group are concatenated into a feature matrix of form n*a; In step S31, the fused feature vector is multiplied by the feature matrix to obtain the corresponding similarity. In step S32, the maximum similarity of each group is obtained, and it is determined whether the maximum similarity is greater than a preset threshold. In step S33, if the value is greater than a preset threshold, the fused image is determined to match the corresponding fault feature vector of the group, and the fault category corresponding to the image is obtained.
[0050] In this invention, after acquiring the fused image, it can be converted into a fused feature vector to obtain a vector of form 1*a. Then, the fault feature vectors in the fault database can be vertically grouped, with each group containing multiple fault feature vectors, each of form a*1. The fault feature vectors of each group are concatenated to form a feature matrix of form n*a. After obtaining the feature matrix, the fused feature vector is multiplied by the feature matrix to obtain the corresponding similarity. After obtaining the similarity of each group, the maximum similarity of each group can be obtained, and it is determined whether the maximum similarity is greater than a preset threshold. If it is greater than the preset threshold, it can be determined that the fused image matches the corresponding fault feature vector of that group, thus obtaining the fault category corresponding to the image.
[0051] In one embodiment of the present invention, such as Figure 6 As shown, the process for obtaining the fault category may include: In step S34, if the value is less than a preset threshold, it is determined that no fault was detected in the fused image, and the substation equipment image is reacquired and monitored.
[0052] In this invention, if the maximum similarity of each group is less than the preset threshold, it can be determined that no fault was detected in the fused image, and then the image of the substation equipment can be reacquired and monitored.
[0053] In one embodiment of the present invention, such as Figure 7 As shown, the process for determining the warning level may include: In step S35, the fault category and the warning value mapped to the time and location when the corresponding fault category is obtained are acquired.
[0054] In step S36, the weights of each fault category and its corresponding time and location are determined.
[0055] In step S37, the obtained weights and warning values are weighted and summed to obtain the warning index.
[0056] In step S38, the corresponding monitoring and early warning level is determined based on the early warning index. The monitoring and early warning levels include no risk, low risk, medium risk, and high risk.
[0057] In this invention, when determining the monitoring and early warning level, the fault category and the corresponding early warning value mapped to the time and location of the fault category can be obtained. Then, the weight of each fault category and its corresponding time and location can be determined. Based on the obtained weights and early warning values, a weighted sum can be performed to obtain the early warning index. Based on the early warning index, the corresponding monitoring and early warning level can be determined. This monitoring and early warning level can include no risk, low risk, medium risk, and high risk.
[0058] On the other hand, the present invention can also provide a substation equipment monitoring system based on image super-resolution, which may include an image acquisition module and a monitoring module. The image acquisition module can be used to acquire images of the substation equipment. The monitoring module can be used to execute a substation equipment monitoring method based on image super-resolution as described above based on the acquired images of the substation equipment.
[0059] Through the above technical solution, the present invention provides a method and system for monitoring substation equipment based on image super-resolution. This involves acquiring images of the substation equipment, then performing super-resolution processing on the images using interpolation. Simultaneously, super-resolution processing can be performed using convolutional neural networks, generative adversarial networks, and Transformer networks. After obtaining the super-resolution images through these methods, the images obtained from the multiple processing models are fused to obtain a fused image. These multiple processing models can be the aforementioned interpolation, convolutional neural networks, generative adversarial networks, and Transformer networks. This image fusion method fully considers the advantages and disadvantages of the super-resolution images processed by each model, resulting in a more uniform image. After obtaining the fused image, it can be compared with a fault database to determine the corresponding fault category. Based on the fault category and the corresponding warning value mapped to the time and location, the monitoring and warning level can be determined. This method for monitoring power equipment can perform super-resolution processing on images of general quality to enable better monitoring of power equipment.
[0060] Those skilled in the art will understand that embodiments of this application can be provided as methods, systems, or computer program products. Therefore, this application can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, this application can take the form of a computer program product embodied on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.
[0061] This application is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of this application. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart... Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.
[0062] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.
[0063] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.
[0064] In a typical configuration, a computing device includes one or more processors (CPU), input / output interfaces, network interfaces, and memory.
[0065] Memory may include non-persistent memory in computer-readable media, such as random access memory (RAM) and / or non-volatile memory, such as read-only memory (ROM) or flash RAM. Memory is an example of computer-readable media.
[0066] Computer-readable media includes both permanent and non-permanent, removable and non-removable media that can store information using any method or technology. Information can be computer-readable instructions, data structures, modules of programs, or other data. Examples of computer storage media include, but are not limited to, phase-change memory (PRAM), static random access memory (SRAM), dynamic random access memory (DRAM), other types of random access memory (RAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), flash memory or other memory technologies, CD-ROM, digital versatile optical disc (DVD) or other optical storage, magnetic tape, magnetic magnetic disk storage or other magnetic storage devices, or any other non-transferable medium that can be used to store information accessible by a computing device. As defined herein, computer-readable media does not include transient computer-readable media, such as modulated data signals and carrier waves.
[0067] It should also be noted that the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such process, method, article, or apparatus. Unless otherwise specified, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes that element.
[0068] The above are merely embodiments of this application and are not intended to limit the scope of this application. Various modifications and variations can be made to this application by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of this application should be included within the scope of the claims of this application.
Claims
1. A method for monitoring substation equipment based on image super-resolution, characterized in that, The method for monitoring power equipment includes: Acquire images of the power equipment; Super-resolution processing is performed on the image using a difference method; The image is reconstructed using a convolutional neural network; The image is then subjected to super-resolution reconstruction using a generative adversarial network. The image is then subjected to super-resolution reconstruction using a Transformer network. The image after super-resolution processing by multiple processing models is acquired, and the image is fused to obtain a fused image; Based on the comparison results between the fused image and the fault database, the corresponding fault category is obtained; Based on the fault category and the warning value mapped to the time and location when the corresponding fault category is obtained, the monitoring and warning level of the response is determined.
2. The method for monitoring power equipment according to claim 1, characterized in that, Super-resolution processing of the image is performed using an interpolation method, including: Acquire an image of the power equipment, use the image as the source image, and select the pixels in the source image as floating-point coordinates; Obtain the nearest 4x4 neighboring pixels around the given floating-point coordinates; The weights of the bicubic interpolation are calculated using a cubic polynomial kernel function. The pixels within the nearest 4x4 neighborhood of the floating-point coordinates are traversed, and a weighted summation using bicubic interpolation is performed based on the obtained weights to obtain the pixel value of the target point mapped by the floating-point coordinates, thereby completing the super-resolution processing of the image.
3. The method for monitoring power equipment according to claim 1, characterized in that, The image is reconstructed using a convolutional neural network; Images of the power equipment are acquired as training images, and the size of the training images is reduced to 1 / K of the original size; The scaled-down training images are fed into the SRCNN network, and the same channels are rearranged by pixel rearrangement to obtain the output image magnified by K times. The original image is compared with the magnified image as the ground truth, and the SRCNN network is trained using the L2 loss function. After training is completed, the images that need to be super-resolution processed are fed into the SRCNN network to complete the super-resolution processing of the images.
4. The method for monitoring power equipment according to claim 1, characterized in that, Super-resolution reconstruction of the image is performed using a generative adversarial network, including: Images of the power equipment are acquired as training images and fed into the SRGAN network for training. The generator of the SRGAN network performs super-resolution on the training images and transmits the results to the discriminator; The discriminator classifies the output images to complete the training of the SRGAN network; During training, the SRGAN network is trained in a supervised manner by using a weighted average of the L2 loss function and the generative adversarial loss. After training is complete, the images that need super-resolution processing are fed into the SRGAN network to complete the super-resolution processing of the images.
5. The method for monitoring power equipment according to claim 1, characterized in that, The process involves acquiring the image after super-resolution processing using multiple processing models, and then fusing the images to obtain a fused image, including: Images obtained after super-resolution processing using multiple processing models are aligned with the images containing the same scene. The aligned image is cut into segments to obtain segmented image blocks, and image blocks containing the same scene are aligned. The aligned image blocks are obtained, and the sharpness of each image block is determined by the gradient function method; Based on the sharpness of the aligned image blocks, determine the weight value of each aligned image block; Based on the weight values of the aligned image blocks, a weighted algorithm is used to fuse the aligned image blocks to obtain a fused image block. The merged image blocks are stitched together to obtain the merged image.
6. The method for monitoring power equipment according to claim 5, characterized in that, Based on the comparison results between the fused image and the fault database, the corresponding fault categories are obtained, including: The fused image is acquired and converted into a fused feature vector to obtain a vector of the form 1*a; The fault feature vectors in the fault database are vertically grouped, with each group containing multiple fault feature vectors, and each fault feature vector having the form a*1. The fault feature vectors of each group are concatenated into a feature matrix of form n*a; The fused feature vector is multiplied by the feature matrix to obtain the corresponding similarity. Obtain the maximum similarity for each group and determine whether the maximum similarity is greater than a preset threshold; If the value is greater than a preset threshold, the fused image is determined to match the corresponding fault feature vector of the group, and the fault category corresponding to the image is obtained.
7. The method for monitoring power equipment according to claim 6, characterized in that, Based on the comparison results between the fused image and the fault database, the corresponding fault categories are obtained, including: If the value is less than a preset threshold, it is determined that no fault was detected in the fused image, and the substation equipment image is reacquired and monitored.
8. The method for monitoring power equipment according to claim 6, characterized in that, Based on the fault category and the warning value mapped to the time and location when the corresponding fault category was acquired, the monitoring and warning level is determined, including: Obtain the fault category and the warning value mapped to the time and location when the corresponding fault category is obtained; Determine the weight of each fault category and its corresponding time and location; The warning index is obtained by weighting and summing the obtained weights and warning values. Based on the warning index, the corresponding monitoring and warning level is determined, which includes no risk, low risk, medium risk, and high risk.
9. A substation equipment monitoring system based on image super-resolution, characterized in that, The power equipment monitoring system includes: The image acquisition module is used to acquire images of the substation equipment; The monitoring module is used to execute a substation monitoring method based on image super-resolution as described in any one of claims 1-8 based on the acquired images of the substation equipment.