Method and system for quantitative evaluation of insulation defects of high-voltage equipment based on ultraviolet imaging

Abstract

The application discloses a high-voltage equipment insulation defect quantitative evaluation method and system based on ultraviolet imaging, relates to the technical field of electrical equipment defect evaluation, and collects ultraviolet image data in the detection area of the high-voltage equipment through an ultraviolet imager, carries out pretreatment on the collected ultraviolet image data through an image pretreatment mode after the collection is completed, carries out accurate processing on the pretreated ultraviolet image data through a light spot segmentation, feature extraction and photon counting calibration mode after the pretreatment is completed, obtains processed ultraviolet image data, carries out fusion on the processed ultraviolet image data through an image fusion mode after the processing is completed, analyzes the fused ultraviolet image data through a defect analysis mode, finally carries out quantitative evaluation on the analyzed ultraviolet image data through a deep learning mode, and generates a corresponding evaluation report according to the quantitative evaluation result, thereby improving the accuracy of the intelligent evaluation of the ultraviolet image of the high-voltage equipment.

Description

Technical Field

[0001] This invention relates to the field of electrical equipment defect assessment technology, specifically to a quantitative assessment method and system for insulation defects in high-voltage equipment based on ultraviolet imaging. Background Technology

[0002] Corona discharge in power transmission lines refers to the discharge phenomenon that occurs when the electric field strength around a conductor exceeds 20-30 kV / cm under a certain voltage, causing discontinuous ionization of the air. This phenomenon is very common in high-voltage power transmission. On the one hand, corona discharge causes energy loss, generates audible noise and electromagnetic interference; on the other hand, harmful substances such as nitrogen oxides and ozone produced by corona discharge can corrode equipment, thus exacerbating the corona discharge phenomenon and creating a vicious cycle, which poses a serious threat to the safe operation of transmission lines. Furthermore, high-voltage electrical equipment (transformers, GIS, insulators, bushings, etc.) is prone to partial discharge due to insulation aging, contamination, and moisture during long-term operation, which can then develop into insulation breakdown accidents.

[0003] Existing technology, such as the invention patent application with publication number CN111951247A, discloses a method and system for ultraviolet image diagnosis of electrical equipment. The method includes: identifying photon values ​​and gain values, the names and identification box positions of each component, and the position of light spots from the ultraviolet image of the electrical equipment; and diagnosing the defect level and defect type of the electrical equipment based on the photon values ​​and gain values, the names and identification box positions of each component, and the position of the light spots. This invention achieves intelligent, accurate, and convenient diagnosis of ultraviolet inspection images of electrical equipment.

[0004] As can be seen from the above solutions, the current ultraviolet image diagnostic recognition accuracy is insufficient, and there is a lack of quantitative assessment process for electrical equipment defects, resulting in insufficient ability to judge equipment defects and certain limitations. Summary of the Invention

[0005] The purpose of this invention is to provide a quantitative assessment method and system for insulation defects in high-voltage equipment based on ultraviolet imaging, which solves the problems existing in the background art.

[0006] To solve the above-mentioned technical problems, the present invention adopts the following technical solution: The present invention provides a quantitative assessment method for insulation defects in high-voltage equipment based on ultraviolet imaging, specifically including the following steps: S1. Set the detection area for high-voltage electrical equipment and collect ultraviolet image data within the detection area based on the ultraviolet imager; S2. The acquired ultraviolet image data is preprocessed using image preprocessing methods to obtain preprocessed ultraviolet image data. S3. Based on the preprocessed ultraviolet image data, the preprocessed ultraviolet image data is precisely processed through spot segmentation, feature extraction and photon counting calibration to obtain the processed ultraviolet image data. S4. The processed ultraviolet image data is fused using an image fusion method to obtain fused ultraviolet image data; S5. Analyze the fused ultraviolet image data using defect analysis to obtain the analyzed ultraviolet image data; S6. Quantitatively evaluate the analyzed ultraviolet image data using deep learning, and generate a corresponding evaluation report based on the quantitative evaluation results.

[0007] Preferably, the step of setting the high-voltage electrical equipment detection area and acquiring ultraviolet image data within the high-voltage equipment detection area based on an ultraviolet imager includes the following steps: Install ultraviolet imagers around the designated high-voltage electrical equipment testing area; Set the wavelength and acquisition period of the ultraviolet imager, and acquire ultraviolet image data of the high-voltage equipment detection area based on the set wavelength and acquisition period.

[0008] Preferably, the step of preprocessing the acquired ultraviolet image data to obtain preprocessed ultraviolet image data includes the following steps: S21. The collected ultraviolet image data is converted to grayscale using a weighted average method to obtain grayscale ultraviolet image data. S22. Denoise the grayscale ultraviolet image data by Gaussian filtering to obtain the denoised ultraviolet image data. S23. The denoised ultraviolet image data is processed by median filtering to obtain preprocessed ultraviolet image data. Define a sliding window containing the denoised UV image data at the corresponding coordinates and the surrounding neighborhood of the denoised UV image data at the corresponding coordinates; For each pixel in the sliding window, M pixels are successively taken out from the sliding window based on the size of the sliding window. The pixel values ​​of these M pixels are then sorted by size, and the pixel value of the pixel whose index is the center point is taken as the filter output. The filtered outputs of all pixels in the denoised ultraviolet image data are summarized to obtain the preprocessed ultraviolet image data.

[0009] Preferably, the process of precisely processing the preprocessed ultraviolet image data using spot segmentation, feature extraction, and photon counting calibration to obtain processed ultraviolet image data includes the following steps: S31. The preprocessed ultraviolet image data is segmented by spot segmentation to obtain segmented ultraviolet image data. S32. Based on the segmented ultraviolet image data, the Canny algorithm is used to set high and low thresholds, calculate image gradients, locate image edges, remove isolated weak edges, retain continuous and complete target contour boundaries, and summarize the target image data. S33. The target image data is precisely processed by feature extraction and photon counting calibration to obtain the processed ultraviolet image data.

[0010] Preferably, the step of segmenting the preprocessed ultraviolet image data using a spot segmentation method to obtain the segmented ultraviolet image data includes the following steps: Based on the grayscale values ​​of the preprocessed ultraviolet image data, multiple thresholds are used. The algorithm is used to segment the image and obtain the segmented ultraviolet image data. The gray levels of the preprocessed ultraviolet image data are set to... The number of pixels with a grayscale value of l is The total number of pixels is ; Furthermore, the grayscale probability of each grayscale value in the preprocessed ultraviolet image data is calculated. ,in, This represents the number of pixels with a grayscale value of k. This represents the probability of a pixel having a grayscale value of k. Set threshold The image pixels are divided into two categories, and grayscale values ​​are set accordingly. As background, For the goal; Calculate the class weights of pixels in the two image classes respectively; The average gray level of pixels in the two types of images is calculated based on the class weights of the pixels in the two types of images. Calculate the inter-class variance based on the average gray level of pixels in the two types of images and the class weights. Set the threshold corresponding to the maximum inter-class variance. for The algorithm establishes a segmentation threshold and then segments the preprocessed ultraviolet image data based on the segmentation threshold, setting gray values ​​greater than or equal to the segmentation threshold to 255 and gray values ​​less than the segmentation threshold to 0. The segmentation results are summarized to obtain the segmented ultraviolet image data; The obtained segmentation threshold is set as the threshold input for the Canny algorithm.

[0011] Preferably, the precise processing of the target image data through feature extraction and photon counting calibration to obtain the processed ultraviolet image data includes the following steps: S331. Perform feature extraction on the target image data to obtain the target image data after feature extraction. S332. Perform count calibration on the target image data after feature extraction using a photon counting calibration method to obtain calibrated target image data; S333, Set the calibrated target image data to the processed ultraviolet image data.

[0012] Preferably, the step of fusing the processed ultraviolet image data through image fusion to obtain the fused ultraviolet image data includes the following steps: Set the weights for the two sets of images to be fused, and perform image fusion on the two sets of images to be fused based on pixel-weighted averaging. The formula for image fusion is shown below: ; in, Represents the fused image Pixel value at that location, Representing an image middle Pixel value at that location, Representing an image middle Pixels at that location, These represent the weights set for the corresponding image types.

[0013] Preferably, the step of analyzing the fused ultraviolet image data using defect analysis to obtain the analyzed ultraviolet image data includes the following steps: Ultraviolet image data of high-voltage equipment under fault conditions are collected, and features are extracted from the fused ultraviolet image data and the ultraviolet image data of high-voltage equipment under fault conditions through a convolutional neural network. Image features were extracted from the fused ultraviolet image data and the ultraviolet image data under high-voltage equipment fault conditions using a convolutional neural network. The acquired ultraviolet image data is input into a convolutional neural network; A convolutional neural network is defined as including: an input layer, convolutional layers, pooling layers, and fully connected layers; The acquired ultraviolet image data is divided into image data blocks of the same size, and the divided image data blocks are input into the input layer of the neural network. After receiving the image data block, the input layer transmits the received image data block to the convolutional layer, which extracts local features from the image data through convolution operations. The formula for the convolution operation is as follows: ; in, This represents the input image data block. This represents the weights of the corresponding convolutional layer. Indicates the bias value. Indicates output features; After the convolutional layer extracts local features from the image data, the pooling layer downsamples the features to reduce the data dimensionality. At the same time, the image data is continuously extracted through convolution and pooling until the feature extraction converges, the convolution stops, and the extracted features are summarized and input into the fully connected layer. The fully connected layer integrates the extracted features and outputs a feature array; The output feature array is set to include the image features from the fused ultraviolet image data and the ultraviolet image data under high-voltage equipment fault conditions; The image features of the fused ultraviolet image data and the ultraviolet image data under high-voltage equipment fault conditions are compared using a similarity comparison function; A similarity comparison threshold is set. When the similarity between the fused ultraviolet image data and the ultraviolet image data under high-voltage equipment failure state is greater than the set similarity comparison threshold, it indicates that the fused ultraviolet image data has defects; otherwise, there are no defects. The fused ultraviolet image data is divided based on the similarity comparison results, and the analyzed ultraviolet image data is then summarized.

[0014] Preferably, the step of quantitatively evaluating the analyzed ultraviolet image data using deep learning and generating a corresponding evaluation report based on the quantitative evaluation results includes the following steps: S61. The analyzed ultraviolet image data is predicted and evaluated by LSTM time series modeling to obtain the predicted ultraviolet image data. The analyzed ultraviolet image data is sorted according to the time series to obtain the sorted ultraviolet image data; The sorted ultraviolet image data is divided into training and validation sets in equal proportions. An LSTM network model is constructed, and the sorted ultraviolet image data in the training set is input into the constructed LSTM network model. The ultraviolet image data predicted at the next time step is obtained by constructing the input gate, output gate and forget gate of the LSTM network model. The sorted UV image data is used as the input gate to the LSTM network, and the sorted UV image data from the validation set is used as the candidate values ​​to input to the LSTM network. The input at the current time step and the ultraviolet image data from the previous time step are processed and updated using a forget gate; The ultraviolet image data processed by the forget gate is output through the output gate to obtain the predicted ultraviolet image data for the next time step; Iteratively use the LSTM network to determine and obtain the trained LSTM network; Set an error threshold between the UV image data predicted by the LSTM network and the sorted UV image data in the validation set. When the error between the UV image data predicted by the LSTM network and the sorted UV image data in the validation set is within the set error threshold range, the iteration stops, and the trained LSTM network is obtained. An LSTM network that meets the error threshold range is set as the state prediction model, and the predicted ultraviolet image data for the next time step is output. S62. Quantitatively evaluate the predicted ultraviolet image data using a graded quantification method; Discharge intensity grading standards are set based on high-voltage equipment industry standards. Based on the established discharge intensity grading standard, the predicted ultraviolet image data is quantitatively evaluated, and corresponding measures are implemented according to the assessed defect level, and a corresponding evaluation report is generated.

[0015] This invention also provides a quantitative assessment system for insulation defects in high-voltage equipment based on ultraviolet imaging, which is used to realize a quantitative assessment method for insulation defects in high-voltage equipment based on ultraviolet imaging. The system includes: a data acquisition module, an image preprocessing module, a spot detection and segmentation module, a photon counting module, an image fusion module, an intelligent diagnosis module, and a report generation module. The data acquisition module is used to control the ultraviolet imager to acquire ultraviolet imaging images of the high-voltage equipment; The image preprocessing module is used to preprocess the acquired ultraviolet imaging images of the high-voltage equipment to obtain preprocessed ultraviolet image data. The spot detection and segmentation module is used to perform spot detection and segmentation on the preprocessed ultraviolet image data. The photon counting module is used to count and calibrate the segmented ultraviolet image data to obtain the processed ultraviolet image data; The image fusion module is used to fuse the processed ultraviolet image data to obtain fused ultraviolet image data; The intelligent diagnostic module is used to perform defect analysis on the fused ultraviolet image data to obtain the analyzed ultraviolet image data; The report generation module is used to quantitatively evaluate the analyzed ultraviolet image data and generate a report.

[0016] The beneficial effects of this invention are as follows: This invention acquires ultraviolet image data within the detection area of ​​high-voltage equipment using an ultraviolet imager. After acquisition, the acquired ultraviolet image data is preprocessed using image preprocessing. Following preprocessing, the preprocessed ultraviolet image data is precisely processed using spot segmentation, feature extraction, and photon counting calibration to obtain processed ultraviolet image data. After processing, the processed ultraviolet image data is fused using image fusion, and the fused ultraviolet image data is analyzed using defect analysis. Finally, the analyzed ultraviolet image data is quantitatively evaluated using deep learning, and a corresponding evaluation report is generated based on the quantitative evaluation results, thereby improving the accuracy of intelligent evaluation of ultraviolet images of high-voltage equipment. Attached Figure Description

[0017] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0018] Figure 1 This is a schematic diagram of the process for quantitative assessment of insulation defects in high-voltage equipment using ultraviolet imaging, as described in this invention. Detailed Implementation

[0019] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0020] In a specific embodiment of the present invention, Reference Figure 1 As shown, this invention provides a method for quantitative assessment of insulation defects in high-voltage equipment based on ultraviolet imaging, comprising the following steps: S1. Set the detection area for high-voltage electrical equipment and collect ultraviolet image data within the detection area based on the ultraviolet imager; S2. The acquired ultraviolet image data is preprocessed using image preprocessing methods to obtain preprocessed ultraviolet image data. S3. Based on the preprocessed ultraviolet image data, the preprocessed ultraviolet image data is precisely processed through spot segmentation, feature extraction and photon counting calibration to obtain the processed ultraviolet image data. S4. The processed ultraviolet image data is fused using an image fusion method to obtain fused ultraviolet image data; S5. Analyze the fused ultraviolet image data using defect analysis to obtain the analyzed ultraviolet image data; S6. Quantitatively evaluate the analyzed ultraviolet image data using deep learning, and generate a corresponding evaluation report based on the quantitative evaluation results; Furthermore, referring to Figure 1 As shown, setting up a detection area for high-voltage electrical equipment and acquiring ultraviolet image data within that area using an ultraviolet imager includes the following steps: Install ultraviolet imagers around the designated high-voltage electrical equipment testing area; Furthermore, the wavelength and acquisition period of the ultraviolet imager are set, and ultraviolet image data of the high-voltage equipment detection area are acquired based on the set wavelength and acquisition period; Furthermore, referring to Figure 1 As shown, the acquired ultraviolet image data is preprocessed using image preprocessing methods to obtain preprocessed ultraviolet image data, including the following steps: S21. The collected ultraviolet image data is converted to grayscale using a weighted average method to obtain grayscale ultraviolet image data. ; in, Represents image coordinates Grayscale ultraviolet image data, Represents image coordinates The red component in Represents image coordinates The green component in Represents image coordinates The blue component in; S22. Denoise the grayscale ultraviolet image data by Gaussian filtering to obtain the denoised ultraviolet image data. The grayscale ultraviolet image data is traversed, and uniform noise reduction is performed on each grayscale ultraviolet image data based on the Gaussian function; The formula for the Gaussian function is as follows: ; in, For variance, The base of natural numbers, Represents image coordinates The denoised ultraviolet image data; further, after uniform denoising, the denoised ultraviolet image data is obtained by summarizing. S23. The denoised ultraviolet image data is processed by median filtering to obtain preprocessed ultraviolet image data. Define a sliding window containing the denoised UV image data at the corresponding coordinates and the surrounding neighborhood of the denoised UV image data at the corresponding coordinates; Furthermore, for each pixel in the sliding window, M pixels are successively taken out from the sliding window based on the size of the sliding window. The pixel values ​​of these M pixels are then sorted by size, and the pixel value of the pixel whose index is the center point is taken as the filter output. Furthermore, the filtered outputs of all pixels in the denoised ultraviolet image data are summarized to obtain the preprocessed ultraviolet image data; Furthermore, referring to Figure 1 As shown, based on the preprocessed ultraviolet image data, the preprocessed ultraviolet image data is precisely processed through spot segmentation, feature extraction, and photon counting calibration to obtain the processed ultraviolet image data, including the following steps: S31. The preprocessed ultraviolet image data is segmented by spot segmentation to obtain segmented ultraviolet image data. Based on the grayscale values ​​of the preprocessed ultraviolet image data, multiple thresholds are used. The algorithm is used to segment the image and obtain the segmented ultraviolet image data. The gray levels of the preprocessed ultraviolet image data are set to... The number of pixels with a grayscale value of l is The total number of pixels is ; Furthermore, the grayscale probability of each grayscale value in the preprocessed ultraviolet image data is calculated. ,in, This represents the number of pixels with a grayscale value of k. This represents the probability of a pixel having a grayscale value of k. Furthermore, set a threshold. The image pixels are divided into two categories, and grayscale values ​​are set accordingly. As background, For the goal; Furthermore, the class weights of the two types of image pixels are calculated separately; ; in, Let represent the class weights of the two types of image pixels, and . ; Furthermore, the average gray level of the pixels in the two classes of images is calculated based on the class weights of the pixels in the two classes of images; ; in, The average gray level of pixels in the two types of images; Furthermore, the inter-class variance is calculated based on the average gray level of the pixels in the two types of images and the class weights; ; in, Represents the variance between classes; Furthermore, a threshold is set corresponding to the maximum inter-class variance. for The algorithm establishes a segmentation threshold and then segments the preprocessed ultraviolet image data based on the segmentation threshold, setting gray values ​​greater than or equal to the segmentation threshold to 255 and gray values ​​less than the segmentation threshold to 0. Furthermore, the segmentation results are summarized to obtain the segmented ultraviolet image data; Furthermore, the obtained segmentation threshold is set as the threshold input for the Canny algorithm; S32. Based on the segmented ultraviolet image data, the Canny algorithm is used to set high and low thresholds, calculate image gradients, locate image edges, remove isolated weak edges, retain continuous and complete target contour boundaries, and summarize the target image data. The formula for calculating the image gradient is shown below: , ; , ; in These are Sobel operators in the x and y directions, respectively, with a fixed 3×3 weight matrix. The segmented ultraviolet image data, These are the gradient components of the image in the x and y directions, respectively. The magnitude of the image gradient. The direction of the image gradient; S33. The target image data is precisely processed by feature extraction and photon counting calibration to obtain the processed ultraviolet image data. S331. Perform feature extraction on the target image data to obtain the target image data after feature extraction. Extract parameter features from target image data using data computation methods; Record the size of the acquired ultraviolet image data, and randomly select two points from the acquired ultraviolet image data, while measuring the distance between the selected points in the actual scene; Furthermore, the distance ratio between the actual scene and the ultraviolet image data is calculated using the distance formula between two points; Furthermore, based on the distance ratio between the actual scene and the ultraviolet image data, the actual distance parameter corresponding to each pixel in the ultraviolet image data is calculated; Furthermore, the area of ​​the light spot and the centroid coordinate parameters in the target image data are calculated based on the actual distance parameters corresponding to each pixel in the ultraviolet image data; The light spot corresponding to the target image data is set to be approximately a circle; Calculate the centroid coordinates: Collect the coordinates of three non-collinear pixels on the contour of the target image data, and solve the centroid coordinates by combining the collected pixel coordinates with the standard equation of the circle. The formula for calculating the spot area is as follows: ; in, Indicates the area of ​​the light spot. This indicates the diameter of the light spot corresponding to the target image data; By summarizing the spot area and centroid coordinate parameters in the target image data, the parameter characteristics in the target image data are obtained; Furthermore, the parameter features in the target image data are summarized to obtain the target image data after feature extraction; S332. Perform count calibration on the target image data after feature extraction using a photon counting calibration method to obtain calibrated target image data; The photon count rate = the number of effective photon pulses per unit time; The number of effective photon pulses per unit time is the number of light spots in the target image data; Record the number of light spots per unit time in the target image data after feature extraction, and adjust the number of light spots per unit time according to the distance correction model; ; in, To determine the number of light spots in the corrected target image data, The number of light spots in the original target image data. To detect distance, These are the polynomial fitting coefficients. Number of terms in a polynomial; Furthermore, by summing the number of light spots in the corrected target image data and the target image data after corresponding feature extraction, the calibrated target image data is obtained; S333, Set the calibrated target image data to the processed ultraviolet image data; Furthermore, referring to Figure 1 As shown, the image fusion process is used to fuse the processed ultraviolet image data to obtain the fused ultraviolet image data, which includes the following steps: Set the weights for the two sets of images to be fused, and perform image fusion on the two sets of images to be fused based on pixel-weighted averaging. The formula for image fusion is shown below: ; in, Represents the fused image Pixel value at that location, Representing an image middle Pixel value at that location, Representing an image middle Pixels at that location, These represent the weights set for the corresponding image types; Furthermore, referring to Figure 1 As shown, the fused ultraviolet image data is analyzed using a defect analysis method. The analyzed ultraviolet image data includes the following steps: Ultraviolet image data of high-voltage equipment under fault conditions are collected, and features are extracted from the fused ultraviolet image data and the ultraviolet image data of high-voltage equipment under fault conditions through a convolutional neural network. Image features were extracted from the fused ultraviolet image data and the ultraviolet image data under high-voltage equipment fault conditions using a convolutional neural network. The acquired ultraviolet image data is input into a convolutional neural network; A convolutional neural network is defined as including: an input layer, convolutional layers, pooling layers, and fully connected layers; The acquired ultraviolet image data is divided into image data blocks of the same size, and the divided image data blocks are input into the input layer of the neural network. After receiving the image data block, the input layer transmits the received image data block to the convolutional layer, which extracts local features from the image data through convolution operations. The formula for the convolution operation is as follows: ; in, This represents the input image data block. This represents the weights of the corresponding convolutional layer. Indicates the bias value. Indicates output features; After the convolutional layer extracts local features from the image data, the pooling layer downsamples the features to reduce the data dimensionality. At the same time, the image data is continuously extracted through convolution and pooling until the feature extraction converges, the convolution stops, and the extracted features are summarized and input into the fully connected layer. The fully connected layer integrates the extracted features and outputs a feature array; The output feature array is set to include the image features from the fused ultraviolet image data and the ultraviolet image data under high-voltage equipment fault conditions; Furthermore, the image features between the fused ultraviolet image data and the ultraviolet image data under high-voltage equipment fault conditions are compared using a similarity comparison function; Furthermore, a similarity comparison threshold is set. When the similarity between the fused ultraviolet image data and the ultraviolet image data under high-voltage equipment failure state is greater than the set similarity comparison threshold, it indicates that the fused ultraviolet image data has defects; otherwise, there are no defects. Furthermore, the fused ultraviolet image data is divided according to the similarity comparison results, and the analyzed ultraviolet image data is obtained by summarizing the results. Furthermore, referring to Figure 1 As shown, the quantitative evaluation of the analyzed ultraviolet image data using deep learning, and the generation of a corresponding evaluation report based on the quantitative evaluation results, includes the following steps: S61. The analyzed ultraviolet image data is predicted and evaluated by LSTM time series modeling to obtain the predicted ultraviolet image data. The analyzed ultraviolet image data is sorted according to the time series to obtain the sorted ultraviolet image data; Furthermore, the sorted ultraviolet image data is divided proportionally into a training set and a validation set; An LSTM network model is constructed, and the sorted ultraviolet image data in the training set is input into the constructed LSTM network model. The ultraviolet image data predicted at the next time step is obtained by constructing the input gate, output gate and forget gate of the LSTM network model. The sorted UV image data is used as the input gate to the LSTM network, and the sorted UV image data from the validation set is used as the candidate values ​​to input to the LSTM network. The input at the current time step and the ultraviolet image data from the previous time step are processed and updated using a forget gate; The ultraviolet image data processed by the forget gate is output through the output gate to obtain the predicted ultraviolet image data for the next time step; Iteratively use the LSTM network to determine and obtain the trained LSTM network; Set an error threshold between the UV image data predicted by the LSTM network and the sorted UV image data in the validation set. When the error between the UV image data predicted by the LSTM network and the sorted UV image data in the validation set is within the set error threshold range, the iteration stops, and the trained LSTM network is obtained. An LSTM network that meets the error threshold range is set as the state prediction model, and the predicted ultraviolet image data for the next time step is output. S62. Quantitatively evaluate the predicted ultraviolet image data using a graded quantification method; Discharge intensity grading standards are set based on high-voltage equipment industry standards. ; Furthermore, based on the established discharge intensity grading standard, the predicted ultraviolet image data is quantitatively evaluated, and corresponding measures are implemented according to the assessed defect level, and a corresponding evaluation report is generated. In one specific embodiment, the high-voltage equipment insulation defect quantitative assessment system based on ultraviolet imaging is used to implement a method for quantitative assessment of high-voltage equipment insulation defects based on ultraviolet imaging. The system includes: a data acquisition module, an image preprocessing module, a spot detection and segmentation module, a photon counting module, an image fusion module, an intelligent diagnosis module, and a report generation module. The data acquisition module is used to control the ultraviolet imager to acquire ultraviolet imaging images of the high-voltage equipment; The image preprocessing module is used to preprocess the acquired ultraviolet imaging images of the high-voltage equipment to obtain preprocessed ultraviolet image data. The spot detection and segmentation module is used to perform spot detection and segmentation on the preprocessed ultraviolet image data. The photon counting module is used to count and calibrate the segmented ultraviolet image data to obtain processed ultraviolet image data; The image fusion module is used to fuse the processed ultraviolet image data to obtain fused ultraviolet image data; The intelligent diagnostic module is used to perform defect analysis on the fused ultraviolet image data to obtain the analyzed ultraviolet image data; The report generation module is used to quantitatively evaluate the analyzed ultraviolet image data and generate a report.

[0021] It should be noted that the above content is merely an example and illustration of the concept of the present invention. Those skilled in the art can make various modifications or additions to the specific embodiments described or use similar methods to replace them, as long as they do not deviate from the concept of the invention or exceed the scope defined by the present invention, they should all fall within the protection scope of the present invention.

Claims

1. A quantitative assessment method for insulation defects in high-voltage equipment based on ultraviolet imaging, characterized in that, Includes the following steps: S1. Set the detection area for high-voltage electrical equipment and collect ultraviolet image data within the detection area based on the ultraviolet imager; S2. The acquired ultraviolet image data is preprocessed using image preprocessing methods to obtain preprocessed ultraviolet image data. S3. Based on the preprocessed ultraviolet image data, the preprocessed ultraviolet image data is precisely processed through spot segmentation, feature extraction and photon counting calibration to obtain the processed ultraviolet image data. S4. The processed ultraviolet image data is fused using an image fusion method to obtain fused ultraviolet image data; S5. Analyze the fused ultraviolet image data using defect analysis to obtain the analyzed ultraviolet image data; S6. Quantitatively evaluate the analyzed ultraviolet image data using deep learning, and generate a corresponding evaluation report based on the quantitative evaluation results.

2. The method for quantitative assessment of insulation defects in high-voltage equipment based on ultraviolet imaging according to claim 1, characterized in that, The process of setting up a high-voltage electrical equipment detection area and acquiring ultraviolet image data within that area using an ultraviolet imager includes the following steps: Install ultraviolet imagers around the designated high-voltage electrical equipment testing area; Set the wavelength and acquisition period of the ultraviolet imager, and acquire ultraviolet image data of the high-voltage equipment detection area based on the set wavelength and acquisition period.

3. The method for quantitative assessment of insulation defects in high-voltage equipment based on ultraviolet imaging according to claim 1, characterized in that, The process of preprocessing the acquired ultraviolet image data to obtain preprocessed ultraviolet image data includes the following steps: S21. The collected ultraviolet image data is converted to grayscale using a weighted average method to obtain grayscale ultraviolet image data. S22. Denoise the grayscale ultraviolet image data by Gaussian filtering to obtain the denoised ultraviolet image data. S23. The denoised ultraviolet image data is processed by median filtering to obtain preprocessed ultraviolet image data. Define a sliding window containing the denoised ultraviolet image data at the corresponding coordinates and the surrounding neighborhood of the denoised ultraviolet image data at the corresponding coordinates; For each pixel in the sliding window, M pixels are successively taken out from the sliding window based on the size of the sliding window. The pixel values ​​of these M pixels are then sorted by size, and the pixel value of the pixel whose index is the center point is taken as the filter output. The filtered outputs of all pixels in the denoised ultraviolet image data are summarized to obtain the preprocessed ultraviolet image data.

4. The method for quantitative assessment of insulation defects in high-voltage equipment based on ultraviolet imaging according to claim 1, characterized in that, The preprocessed ultraviolet image data is precisely processed through spot segmentation, feature extraction, and photon counting calibration to obtain the processed ultraviolet image data, including the following steps: S31. The preprocessed ultraviolet image data is segmented by spot segmentation to obtain segmented ultraviolet image data. S32. Based on the segmented ultraviolet image data, the Canny algorithm is used to set high and low thresholds, calculate image gradients, locate image edges, remove isolated weak edges, retain continuous and complete target contour boundaries, and summarize the target image data. S33. The target image data is precisely processed by feature extraction and photon counting calibration to obtain the processed ultraviolet image data.

5. The quantitative assessment method for insulation defects in high-voltage equipment based on ultraviolet imaging according to claim 4, characterized in that, The step of segmenting the preprocessed ultraviolet image data using a spot segmentation method to obtain the segmented ultraviolet image data includes the following steps: Based on the grayscale values ​​of the preprocessed ultraviolet image data, multiple thresholds are used. The algorithm is used to segment the image and obtain the segmented ultraviolet image data. The gray levels of the preprocessed ultraviolet image data are set to... The number of pixels with a grayscale value of l is The total number of pixels is ; Furthermore, the grayscale probability of each grayscale value in the preprocessed ultraviolet image data is calculated. ,in, This represents the number of pixels with a grayscale value of k. This represents the probability of a pixel having a grayscale value of k. Set threshold The image pixels are divided into two categories, and grayscale values ​​are set accordingly. As background, For the goal; Calculate the class weights of pixels in the two image classes respectively; The average gray level of pixels in the two types of images is calculated based on the class weights of the pixels in the two types of images. Calculate the inter-class variance based on the average gray level of pixels in the two types of images and the class weights. Set the threshold corresponding to the maximum inter-class variance. for The algorithm establishes a segmentation threshold and then segments the preprocessed ultraviolet image data based on the segmentation threshold, setting gray values ​​greater than or equal to the segmentation threshold to 255 and gray values ​​less than the segmentation threshold to 0. The segmentation results are summarized to obtain the segmented ultraviolet image data; The obtained segmentation threshold is set as the threshold input for the Canny algorithm.

6. The quantitative assessment method for insulation defects in high-voltage equipment based on ultraviolet imaging according to claim 4, characterized in that, The process of precisely processing the target image data through feature extraction and photon counting calibration to obtain the processed ultraviolet image data includes the following steps: S331. Perform feature extraction on the target image data to obtain the target image data after feature extraction. S332. Perform count calibration on the target image data after feature extraction using a photon counting calibration method to obtain calibrated target image data; S333, Set the calibrated target image data to the processed ultraviolet image data.

7. The method for quantitative assessment of insulation defects in high-voltage equipment based on ultraviolet imaging according to claim 1, characterized in that, The process of fusing the processed ultraviolet image data using image fusion to obtain the fused ultraviolet image data includes the following steps: Set the weights for the two sets of images to be fused, and perform image fusion on the two sets of images to be fused based on pixel-weighted averaging. The formula for image fusion is shown below: ； in, Represents the fused image Pixel value at that location, Representing an image middle Pixel value at that location, Representing an image middle Pixels at that location, These represent the weights set for the corresponding image types.

8. The method for quantitative assessment of insulation defects in high-voltage equipment based on ultraviolet imaging according to claim 1, characterized in that, The step of analyzing the fused ultraviolet image data using defect analysis to obtain the analyzed ultraviolet image data includes the following steps: Ultraviolet image data of high-voltage equipment under fault conditions are collected, and features are extracted from the fused ultraviolet image data and the ultraviolet image data of high-voltage equipment under fault conditions through a convolutional neural network. Image features were extracted from the fused ultraviolet image data and the ultraviolet image data under high-voltage equipment fault conditions using a convolutional neural network. The acquired ultraviolet image data is input into a convolutional neural network; A convolutional neural network is defined as including: an input layer, convolutional layers, pooling layers, and fully connected layers; The acquired ultraviolet image data is divided into image data blocks of the same size, and the divided image data blocks are input into the input layer of the neural network. After receiving the image data block, the input layer transmits the received image data block to the convolutional layer, which extracts local features from the image data through convolution operations. The formula for the convolution operation is as follows: ； in, This represents the input image data block. This represents the weights of the corresponding convolutional layer. Indicates the bias value. Indicates output features; After the convolutional layer extracts local features from the image data, the pooling layer downsamples the features to reduce the data dimensionality. At the same time, the image data is continuously extracted through convolution and pooling until the feature extraction converges, the convolution stops, and the extracted features are summarized and input into the fully connected layer. The fully connected layer integrates the extracted features and outputs a feature array; The output feature array is set to include the image features from the fused ultraviolet image data and the ultraviolet image data under high-voltage equipment fault conditions; The image features of the fused ultraviolet image data and the ultraviolet image data under high-voltage equipment fault conditions are compared using a similarity comparison function; A similarity comparison threshold is set. When the similarity between the fused ultraviolet image data and the ultraviolet image data under high-voltage equipment failure state is greater than the set similarity comparison threshold, it indicates that the fused ultraviolet image data has defects; otherwise, there are no defects. The fused ultraviolet image data is divided based on the similarity comparison results, and the analyzed ultraviolet image data is then summarized.

9. The method for quantitative assessment of insulation defects in high-voltage equipment based on ultraviolet imaging according to claim 1, characterized in that, The quantitative evaluation of the analyzed ultraviolet image data using deep learning, and the generation of a corresponding evaluation report based on the quantitative evaluation results, includes the following steps: S61. The analyzed ultraviolet image data is predicted and evaluated by LSTM time series modeling to obtain the predicted ultraviolet image data. The analyzed ultraviolet image data is sorted according to the time series to obtain the sorted ultraviolet image data; The sorted ultraviolet image data is divided into training and validation sets in equal proportions. An LSTM network model is constructed, and the sorted ultraviolet image data in the training set is input into the constructed LSTM network model. The ultraviolet image data predicted at the next time step is obtained by constructing the input gate, output gate and forget gate of the LSTM network model. The sorted ultraviolet image data is used as the input gate to the LSTM network, and the sorted ultraviolet image data in the validation set is used as the candidate values ​​to be input into the LSTM network. The input at the current time step and the ultraviolet image data from the previous time step are processed and updated using a forget gate; The ultraviolet image data processed by the forget gate is output through the output gate to obtain the predicted ultraviolet image data for the next time step; Iteratively use the LSTM network to determine and obtain the trained LSTM network; Set an error threshold between the UV image data predicted by the LSTM network and the sorted UV image data in the validation set. When the error between the UV image data predicted by the LSTM network and the sorted UV image data in the validation set is within the set error threshold range, the iteration stops, and the trained LSTM network is obtained. An LSTM network that meets the error threshold range is set as the state prediction model, and the predicted ultraviolet image data for the next time step is output. S62. Quantitatively evaluate the predicted ultraviolet image data using a graded quantification method; Discharge intensity grading standards are set based on high-voltage equipment industry standards. Based on the established discharge intensity grading standard, the predicted ultraviolet image data is quantitatively evaluated, and corresponding measures are implemented according to the assessed defect level, and a corresponding evaluation report is generated.

10. A system for implementing the quantitative assessment method for insulation defects in high-voltage equipment based on ultraviolet imaging as described in claim 1, characterized in that, include: The system includes a data acquisition module, an image preprocessing module, a spot detection and segmentation module, a photon counting module, an image fusion module, an intelligent diagnosis module, and a report generation module. The data acquisition module is used to control the ultraviolet imager to acquire ultraviolet imaging images of the high-voltage equipment; The image preprocessing module is used to preprocess the acquired ultraviolet imaging images of the high-voltage equipment to obtain preprocessed ultraviolet image data. The spot detection and segmentation module is used to perform spot detection and segmentation on the preprocessed ultraviolet image data. The photon counting module is used to count and calibrate the segmented ultraviolet image data to obtain processed ultraviolet image data; The image fusion module is used to fuse the processed ultraviolet image data to obtain fused ultraviolet image data; The intelligent diagnostic module is used to perform defect analysis on the fused ultraviolet image data to obtain the analyzed ultraviolet image data; The report generation module is used to quantitatively evaluate the analyzed ultraviolet image data and generate a report.

Images