A method and system for monitoring local heating of transmission lines

By processing the temperature data and infrared thermal image of the photovoltaic power station, calculating the comprehensive heating index, positioning local high-temperature areas and generating early warning models, the shortcomings of monitoring local high-temperature areas in the existing technology are solved, the early warning and maintenance efficiency are improved, and the The operational safety of photovoltaic power stations.

CN118781095BActive Publication Date: 2024-11-22TIANJIN RAIL TRANSIT URBAN DEV CO LTD

Patent Information

Application Number
CN202411214913.3
Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
Filing Date
2024-09-02
Publication Date
2024-11-22
Estimated Expiration
2044-09-02

AI Technical Summary

Technical Problem

The existing technology is difficult to accurately, timely and comprehensively monitor local high-temperature areas in photovoltaic power plants, resulting in difficulty in early warning and maintenance, affecting the safe and stable operation of photovoltaic power plants.

Method used

By obtaining the temperature data and infrared thermal image of the transmission line, denoising, image enhancement, interpolation processing and data fusion are carried out, comprehensive heating index is calculated, local high-temperature areas are located and quantified, and visual early warning models and differentiated patrol and maintenance are generated. plan.

Benefits of technology

Automatic detection and quantitative evaluation of local high-temperature areas is realized, early warning and maintenance efficiency of heating faults is improved, and operational safety and reliability of photovoltaic power plants are improved.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN118781095B_ABST
    Figure CN118781095B_ABST
Patent Text Reader

Abstract

The present invention relates to the field of image processing technology, and discloses a method and system for monitoring local heating of transmission lines, which are used to improve the efficiency and accuracy of local heating monitoring of transmission lines. The method includes: interpolating the noise reduction temperature data and the enhanced thermal image data to obtain the target temperature matrix and the target thermal image matrix; fusing the temperature matrix and the thermal image matrix to obtain the fused temperature distribution matrix; performing threshold segmentation on the temperature distribution matrix to obtain the local high temperature area, locating and quantifying the local high temperature area to obtain the heating position coordinate set and the heating degree data set; calculating the comprehensive heating index, and determining the local heating warning level according to the comprehensive heating index; matching the local heating warning level with the preset transmission line BIM model to generate a visual warning model, and performing differentiated inspection scheme analysis on the visual warning model according to the local heating warning level to generate a differentiated inspection and maintenance plan.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present invention relates to the field of image processing technology, and in particular to a method and system for monitoring local heating of a power transmission line. Background Art

[0002] With the acceleration of urbanization, subways have become an important means of transportation for people's daily travel. In order to meet the power supply needs of subway operations, photovoltaic power stations are usually built on the top of subway garages. Photovoltaic power stations are composed of a large number of photovoltaic module arrays. Photovoltaic modules are easily affected by natural factors such as wind, sun, and rain in long-term outdoor working environments, resulting in uneven temperature distribution of photovoltaic modules, excessive temperature in local areas, hot spot effects, reduced power generation efficiency of photovoltaic modules, and shortened service life.

[0003] At present, the monitoring of photovoltaic power stations mainly adopts manual inspection and traditional temperature sensor monitoring methods. Manual inspection is inefficient and prone to missing problem areas; the number of traditional temperature sensors is limited, making it difficult to fully monitor the temperature distribution of the photovoltaic array, and the measurement accuracy of the temperature sensor is greatly affected by environmental factors. The above methods cannot timely, accurately and comprehensively detect local high-temperature areas in the photovoltaic array, making it difficult to warn and maintain local heating faults, affecting the safe and stable operation of the photovoltaic power station. Summary of the invention

[0004] In view of this, an embodiment of the present invention provides a method and system for monitoring local heating of a transmission line, which are used to improve the efficiency and accuracy of monitoring local heating of a transmission line.

[0005] The present invention provides a method for monitoring local heating of a power transmission line, comprising: acquiring temperature data and an infrared thermal image of the power transmission line, performing denoising processing on the temperature data to obtain denoised temperature data, and performing image enhancement on the infrared thermal image to obtain enhanced thermal image data; performing interpolation processing on the denoised temperature data and the enhanced thermal image data to obtain a target temperature matrix and a target thermal image matrix; fusing the temperature matrix and the thermal image matrix to obtain a fused temperature distribution matrix; performing threshold segmentation on the temperature distribution matrix to obtain a local high temperature area, and performing positioning and quantification processing on the local high temperature area to obtain a heating position coordinate set and a heating degree data set; calculating a comprehensive heating index according to the heating position coordinate set and the heating degree data set, and determining a local heating warning level according to the comprehensive heating index; matching the local heating warning level with a preset power transmission line BIM model to generate a visual warning model, and performing a differentiated inspection scheme analysis on the visual warning model according to the local heating warning level to generate a differentiated inspection and maintenance plan.

[0006] In the present invention, the steps of acquiring temperature data and infrared thermal images of the power transmission line, denoising the temperature data to obtain denoised temperature data, and enhancing the infrared thermal images to obtain enhanced thermal image data include: performing wavelet transform on the temperature data to obtain high-frequency coefficients corresponding to the temperature data, and performing inverse wavelet transform on the temperature data according to the high-frequency coefficients to obtain the denoised temperature data; performing grayscale stretching on the infrared thermal image to obtain a stretched infrared thermal image; performing histogram equalization on the stretched infrared thermal image to obtain a balanced infrared thermal image. image; performing median filtering on the balanced infrared thermal image to obtain a filtered infrared thermal image; removing the illumination component of the filtered infrared thermal image to obtain a candidate infrared thermal image; performing pixel gradient data analysis on the candidate infrared thermal image to obtain a gradient amplitude map and a gradient direction map; performing non-maximum suppression on the gradient amplitude map to obtain a candidate gradient amplitude map, and performing hysteresis threshold processing on the candidate gradient amplitude map to obtain a target edge map; performing edge sharpening processing on the candidate infrared thermal image based on the gradient direction map and the target edge map to obtain the enhanced thermal image data.

[0007] In the present invention, the step of interpolating the de-noised temperature data and the enhanced thermal image data to obtain a target temperature matrix and a target thermal image matrix includes: obtaining the size of the de-noised temperature data and the size of the enhanced thermal image data, and performing target interpolation size analysis according to the size of the de-noised temperature data and the size of the enhanced thermal image data to obtain a temperature interpolation size and a thermal image interpolation size; performing two-dimensional cubic spline interpolation on the de-noised temperature data according to the temperature interpolation size to obtain interpolated temperature data; performing gradient interpolation on the enhanced thermal image data according to the thermal image interpolation size. The method comprises the steps of: performing bilinear interpolation on the noise-reduced temperature data according to the normalized temperature data to obtain a candidate temperature matrix; performing bicubic interpolation on the enhanced thermal image data according to the normalized thermal image data to obtain a candidate thermal image matrix; performing smoothing on the candidate temperature matrix to obtain the target temperature matrix, and performing Gaussian filtering on the candidate thermal image matrix to obtain the target thermal image matrix.

[0008] In the present invention, the step of fusing the temperature matrix and the thermal image matrix to obtain a fused temperature distribution matrix includes: performing multi-scale decomposition on the temperature matrix and the thermal image matrix respectively to obtain a multi-scale temperature feature matrix and a multi-scale thermal image feature matrix; performing joint sparse coding on the multi-scale temperature feature matrix and the multi-scale thermal image feature matrix to obtain a multi-scale temperature-thermal image feature matrix; performing local linear embedding on the multi-scale temperature-thermal image feature matrix to obtain a low-dimensional temperature-thermal image feature matrix; performing deep feature extraction on the low-dimensional temperature-thermal image feature matrix to obtain a deep temperature-thermal image fusion feature matrix; performing adaptive region selection on the deep temperature-thermal image fusion feature matrix to obtain a candidate fusion region matrix; performing superpixel segmentation on the candidate fusion region matrix to obtain multiple segmentation region matrices; performing global optimization processing on multiple segmentation region matrices to obtain the fused temperature distribution matrix.

[0009] In the present invention, the steps of performing threshold segmentation on the temperature distribution matrix to obtain local high temperature areas, and locating and quantifying the local high temperature areas to obtain a fever position coordinate set and a fever degree data set include: binarizing the temperature distribution matrix to obtain a binary temperature distribution matrix; performing morphological opening operation on the binary temperature distribution matrix to obtain an open operation binary temperature distribution matrix; extracting connected areas from the open operation binary temperature distribution matrix to obtain a high temperature connected area matrix; performing area area and area circumscribed rectangle analysis on the high temperature connected area matrix to obtain a candidate high temperature area matrix set; extracting the centroid coordinates of each high temperature area according to the candidate high temperature area matrix set to obtain an initial fever position coordinate set; performing spatial clustering on the initial fever position coordinate set to merge adjacent fever position coordinates to obtain a fever position coordinate set; extracting the temperature value of each fever position in the temperature distribution matrix according to the clustered fever position coordinate set to obtain a fever position temperature value set; analyzing the fever position temperature value set to obtain an average temperature and a maximum temperature; analyzing the fever degree of each fever position according to the average temperature and the maximum temperature to obtain a fever degree data set.

[0010] In the present invention, the step of calculating a comprehensive fever index according to the fever position coordinate set and the fever degree data set, and determining a local fever warning level according to the comprehensive fever index includes: calculating the spatial aggregation of the fever positions according to the fever position coordinate set to obtain the fever position aggregation; performing a fever degree statistical feature analysis on the fever degree data set to obtain a fever degree statistical feature; performing a weighted summation on the fever position aggregation and the fever degree statistical feature to obtain an initial value of the comprehensive fever index; correcting the initial value of the comprehensive fever index based on a preset spatial distribution influencing factor to obtain a comprehensive fever index; constructing a fever warning level determination condition according to the comprehensive fever index to obtain a multi-level fever warning level determination condition; performing a logical comparison between the comprehensive fever index and the multi-level fever warning level determination condition to obtain an initial result of the local fever warning level; performing a time series analysis on the initial result of the local fever warning level to obtain a time-smoothed local fever warning level; performing a spatial correlation analysis on the time-smoothed local fever warning level to obtain the local fever warning level.

[0011] In the present invention, the local heating warning level is matched with a preset transmission line BIM model to generate a visual warning model, and a differentiated inspection scheme analysis is performed on the visual warning model according to the local heating warning level to generate a differentiated inspection and maintenance plan step, including: color-labeling the preset transmission line BIM model according to the local heating warning level to obtain a color-labeled transmission line BIM model; performing a three-dimensional geometric transformation on the color-labeled transmission line BIM model to obtain a three-dimensional visual warning model; extracting the spatial boundary of the local heating area according to the three-dimensional visual warning model to obtain a local heating area. hot area boundary; gridding the local heating area boundary to obtain the local heating area grid; calculating the grid density of the warning level area based on the local heating area grid to obtain the warning level area grid density distribution; performing inspection resource demand and time window analysis on the warning level area grid density distribution to obtain inspection constraints; generating an initial inspection path based on the inspection constraints, and optimizing the inspection coverage of the initial inspection path to obtain an optimized inspection path; performing differentiated inspection strategy analysis on the optimized inspection path to obtain a differentiated inspection strategy set, and generating the differentiated inspection maintenance plan based on the differentiated inspection strategy set.

[0012] The present invention also provides a transmission line local heating monitoring system, comprising:

[0013] An acquisition module is used to acquire temperature data and infrared thermal images of the power transmission line, perform denoising on the temperature data to obtain denoised temperature data, and perform image enhancement on the infrared thermal images to obtain enhanced thermal image data;

[0014] An interpolation module, used for performing interpolation processing on the noise-reduced temperature data and the enhanced thermal image data to obtain a target temperature matrix and a target thermal image matrix;

[0015] A fusion module, used for fusing the temperature matrix and the thermal image matrix to obtain a fused temperature distribution matrix;

[0016] A segmentation module is used to perform threshold segmentation on the temperature distribution matrix to obtain a local high temperature area, and to locate and quantify the local high temperature area to obtain a fever position coordinate set and a fever degree data set;

[0017] a calculation module, used to calculate a comprehensive fever index according to the fever position coordinate set and the fever degree data set, and determine a local fever warning level according to the comprehensive fever index;

[0018] A matching module is used to match the local heating warning level with a preset transmission line BIM model to generate a visual warning model, and to perform a differentiated inspection scheme analysis on the visual warning model according to the local heating warning level to generate a differentiated inspection and maintenance plan.

[0019] In the technical solution provided by the present invention, by acquiring the temperature data and infrared thermal image of the transmission line, the temperature data is denoised to obtain the denoised temperature data, and the infrared thermal image is enhanced to obtain the enhanced thermal image data, which effectively improves the quality and reliability of the temperature data and the thermal image data. By interpolating the denoised temperature data and the enhanced thermal image data, the target temperature matrix and the target thermal image matrix are obtained, and the standardization and size unification of the temperature and thermal image data are achieved, which lays the foundation for subsequent data fusion. The target temperature matrix and the target thermal image matrix are fused to obtain the fused temperature distribution matrix, which makes full use of the complementary characteristics of the temperature and thermal image data and obtains more comprehensive and accurate temperature distribution information. The fused temperature distribution matrix is ​​threshold segmented to accurately locate the local high temperature area, and the heating position coordinate set and the heating degree data set are obtained through positioning and quantification processing, which realizes the automatic detection and quantitative evaluation of the local high temperature area. According to the heating position coordinate set and the heating degree data set, the comprehensive heating index is calculated, and the local heating warning level is determined, which realizes the comprehensive evaluation and graded warning of the local heating risk. The local heating warning level is matched with the preset transmission line BIM model to generate an intuitive visual warning model, and the visual warning model is analyzed for a differentiated inspection plan according to the warning level to generate a targeted differentiated inspection and maintenance plan, which greatly improves the early warning and maintenance efficiency of local heating failures in the transmission line. The present invention utilizes the multi-source heterogeneous characteristics of temperature data and infrared thermal images, and obtains a high-quality fused temperature distribution matrix through a series of processing such as data denoising, image enhancement, interpolation processing, and data fusion. It further realizes the automatic detection, positioning, evaluation, and early warning of local high-temperature areas through threshold segmentation, positioning quantification, comprehensive index calculation, and early warning classification. Finally, combined with the transmission line BIM model, a visual warning model and a differentiated inspection and maintenance plan are generated, which greatly improves the intelligence level and practicality of local heating monitoring of transmission lines. BRIEF DESCRIPTION OF THE DRAWINGS

[0020] In order to more clearly illustrate the specific implementation methods of the present invention or the technical solutions in the prior art, the drawings required for use in the specific implementation methods or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are some implementation methods of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0021] Figure 1 This is a flow chart of a method for monitoring local heating of a power transmission line in an embodiment of the present invention;

[0022] Figure 2 Schematic diagram of a local heating monitoring system for a power transmission line according to an embodiment of the present invention. DETAILED DESCRIPTION

[0023] The technical solution of the present invention will be described clearly and completely below in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0024] In the description of the present invention, it should be noted that the terms "center", "upper", "lower", "left", "right", "vertical", "horizontal", "inner", "outer", etc., indicating the orientation or positional relationship, are based on the orientation or positional relationship 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 orientation, be constructed and operated in a specific orientation, and therefore cannot be understood as limiting the present invention. In addition, the terms "first", "second", and "third" are used for descriptive purposes only, and cannot be understood as indicating or implying relative importance.

[0025] In addition, the technical features involved in the different embodiments of the present invention described below can be combined with each other as long as they do not conflict with each other.

[0026] For ease of understanding, the specific process of the embodiment of the present invention is described below. Figure 1 , Figure 1 is a flow chart of a method for monitoring local heating of a transmission line according to an embodiment of the present invention. Figure 1 As shown, the following steps are included:

[0027] S101, acquiring temperature data and infrared thermal images of the power transmission line, performing denoising on the temperature data to obtain denoised temperature data, and performing image enhancement on the infrared thermal images to obtain enhanced thermal image data;

[0028] S102, performing interpolation processing on the noise-reduced temperature data and the enhanced thermal image data to obtain a target temperature matrix and a target thermal image matrix;

[0029] S103, fusing the temperature matrix and the thermal image matrix to obtain a fused temperature distribution matrix;

[0030] S104, performing threshold segmentation on the temperature distribution matrix to obtain a local high temperature area, and performing positioning and quantification processing on the local high temperature area to obtain a fever position coordinate set and a fever degree data set;

[0031] S105, calculating a comprehensive fever index according to the fever location coordinate set and the fever degree data set, and determining a local fever warning level according to the comprehensive fever index;

[0032] S106: Match the local heating warning level with the preset transmission line BIM model to generate a visual warning model, and perform differentiated inspection scheme analysis on the visual warning model according to the local heating warning level to generate a differentiated inspection and maintenance plan.

[0033] It should be noted that by laying temperature sensors and infrared thermal imagers on the transmission lines, the temperature data and infrared thermal images of the transmission lines are obtained. Among them, the temperature sensor can use wireless communication to transmit the temperature data to the monitoring center in real time; the infrared thermal imager can use a high-resolution infrared thermal imaging lens to perform a full-range scan of the transmission line to obtain high-quality infrared thermal images. Since the temperature sensor is affected by environmental factors, the collected temperature data may be interfered by noise. In order to improve the reliability of the temperature data, the temperature data needs to be denoised. Common denoising methods include wavelet transform, Kalman filtering, etc. By performing wavelet decomposition on the temperature data, the temperature signal is decomposed into low-frequency components and high-frequency components, and the high-frequency components are thresholded to filter out the high-frequency noise, and then the denoised temperature data is obtained through wavelet reconstruction. For infrared thermal images, due to the complexity of the imaging environment, the image quality may be affected by factors such as uneven lighting and background interference, and the infrared thermal images need to be enhanced. Commonly used image enhancement methods include histogram equalization, median filtering, Retinex algorithm, etc. Histogram equalization can enhance the contrast of the image and highlight the high-temperature area; median filtering can suppress image noise; Retinex algorithm can eliminate shadows and uneven lighting problems in the image to obtain enhanced infrared thermal image data.

[0034] Due to the different locations and numbers of temperature sensors and infrared thermal imagers, the resolution and size of the collected temperature data and infrared thermal image data may be inconsistent. In order to facilitate subsequent data fusion, the denoised temperature data and enhanced thermal image data need to be interpolated to make them have the same size and resolution. Common interpolation methods include bilinear interpolation and bicubic interpolation. By performing bilinear interpolation on the denoised temperature data, the density of temperature data points can be increased while maintaining the continuity of the temperature data; by performing bicubic interpolation on the enhanced thermal image data, the image resolution can be improved while maintaining the image details, and the target temperature matrix and target thermal image matrix can be obtained.

[0035] In order to make full use of the complementary characteristics of temperature data and infrared thermal image data, it is necessary to fuse the target temperature matrix and the target thermal image matrix. Commonly used data fusion methods include weighted average method, principal component analysis method, wavelet transform fusion method, etc. For example, the target temperature matrix and the target thermal image matrix can be subjected to wavelet transform to decompose them into low-frequency components and high-frequency components, and the low-frequency components can be weighted averaged, and the high-frequency components can be processed to take the maximum value, and then reconstructed through inverse wavelet transform to obtain the fused temperature distribution matrix. The fused temperature distribution matrix contains the comprehensive information of temperature data and infrared thermal image data, which can more accurately reflect the temperature distribution of the transmission line.

[0036] In order to locate the local high temperature area of ​​the transmission line, it is necessary to perform threshold segmentation on the fused temperature distribution matrix. By setting the temperature threshold, the temperature distribution matrix is ​​binarized, and the area above the threshold is extracted as the local high temperature area. The extracted local high temperature area is morphologically processed to remove the small area of ​​noise to obtain the real local high temperature area. Then, the center of gravity of the local high temperature area is located and the area is calculated to obtain the position coordinates and area information of each local high temperature area, forming a heating position coordinate set. At the same time, according to the temperature distribution in each local high temperature area, the heating degree indicators such as average temperature and maximum temperature are calculated to form a heating degree data set.

[0037] In order to evaluate the overall heating risk of the transmission line, it is necessary to calculate the comprehensive heating index based on the heating location coordinate set and the heating degree data set. The comprehensive heating index needs to consider multiple factors such as the number, area, and temperature of local high-temperature areas, and can be calculated by weighted summation. For example, the number, area, average temperature and other indicators of local high-temperature areas can be normalized, and then weights are assigned according to the importance of each indicator, and the weighted sum is obtained to obtain the comprehensive heating index. The higher the comprehensive heating index, the greater the overall heating risk of the transmission line. According to the comprehensive heating index, the warning level division threshold can be set to divide the comprehensive heating index into different warning levels such as normal, general, and severe, which is convenient for subsequent inspection and maintenance decisions.

[0038] In order to intuitively display the heating warning information of the transmission line, it is necessary to match the local heating warning level with the preset transmission line BIM model to generate a visual warning model. The BIM model is a three-dimensional digital model of the transmission line, which contains the geometric information, attribute information, etc. of the transmission line. By associating the local heating warning level with the corresponding transmission line components in the BIM model, the local high temperature areas of different warning levels can be intuitively displayed in the BIM model to form a visual warning model. The visual warning model can intuitively display the heating risk distribution of the transmission line through color marking, flashing warnings, etc.

[0039] Based on the visual early warning model, differentiated inspection schemes can be analyzed for transmission lines according to the local heating warning level. For local high-temperature areas with high warning levels, it is necessary to increase the inspection frequency and shorten the inspection cycle; for areas with low warning levels, the inspection frequency can be appropriately reduced and the inspection cycle can be extended. At the same time, the inspection route can be optimized and the inspection efficiency can be improved based on the location, area and other information of the local high-temperature area. Through the analysis of differentiated inspection schemes, targeted differentiated inspection and maintenance plans can be generated, inspection resources can be reasonably allocated, and the operation and maintenance level of transmission lines can be improved.

[0040] For example, the photovoltaic power station on the top of a subway garage has a transmission line length of 2 kilometers and a total of 100 photovoltaic modules connected in series. By laying 50 temperature sensors and 10 infrared thermal imagers on the transmission line, temperature data and infrared thermal images are collected every 10 minutes. The collected temperature data is processed by wavelet denoising to remove the high-frequency noise introduced by environmental interference; the infrared thermal image is subjected to histogram equalization and Retinex enhancement to improve the image contrast and clarity. Then, bilinear interpolation and bicubic interpolation are performed on the denoised temperature data and enhanced thermal image data to make them have the same resolution and size, and a 100x100 target temperature matrix and target thermal image matrix are obtained. The target temperature matrix and the target thermal image matrix are fused by the wavelet transform fusion algorithm to obtain a fused temperature distribution matrix. Threshold segmentation is performed on the fused temperature distribution matrix to extract 5 local high-temperature areas, and the position coordinates and area information of each local high-temperature area are obtained by center of gravity positioning and area calculation to form a heating position coordinate set. At the same time, the average temperature and maximum temperature of each local high-temperature area are calculated to form a heating degree data set. According to the heating position coordinate set and the heating degree data set, the comprehensive heating index is calculated to be 0.65, corresponding to the severe warning level. The severe warning level is matched with the BIM model of the transmission line, and the local high temperature area is marked in red in the BIM model to generate an intuitive visual warning model. Based on the visual warning model, a differentiated inspection plan analysis is conducted on the five local high temperature areas. The area with the highest warning level is inspected once a day, and other areas are inspected once a week. The inspection route is optimized and a differentiated inspection and maintenance plan is generated. Through this method, the local high temperature hazards of the transmission line are discovered and located in a timely manner, and targeted inspection and maintenance measures are formulated, which effectively improves the operational reliability and safety of the transmission line of the photovoltaic power station.

[0041] By executing the above steps, the temperature data and infrared thermal image of the transmission line are obtained, the temperature data is denoised to obtain the denoised temperature data, and the infrared thermal image is enhanced to obtain the enhanced thermal image data, which effectively improves the quality and reliability of the temperature data and thermal image data. The target temperature matrix and the target thermal image matrix are obtained by interpolation processing of the denoised temperature data and the enhanced thermal image data, which realizes the standardization and size unification of the temperature and thermal image data, and lays the foundation for subsequent data fusion. The target temperature matrix and the target thermal image matrix are fused to obtain the fused temperature distribution matrix, which makes full use of the complementary characteristics of the temperature and thermal image data and obtains more comprehensive and accurate temperature distribution information. The fused temperature distribution matrix is ​​threshold segmented to accurately locate the local high temperature area, and the heating position coordinate set and the heating degree data set are obtained through positioning and quantification processing, which realizes the automatic detection and quantitative evaluation of the local high temperature area. According to the heating position coordinate set and the heating degree data set, the comprehensive heating index is calculated, and the local heating warning level is determined, which realizes the comprehensive evaluation and graded warning of the local heating risk. The local heating warning level is matched with the preset transmission line BIM model to generate an intuitive visual warning model, and the visual warning model is analyzed for a differentiated inspection plan according to the warning level to generate a targeted differentiated inspection and maintenance plan, which greatly improves the early warning and maintenance efficiency of local heating failures in the transmission line. The present invention utilizes the multi-source heterogeneous characteristics of temperature data and infrared thermal images, and obtains a high-quality fused temperature distribution matrix through a series of processing such as data denoising, image enhancement, interpolation processing, and data fusion. It further realizes the automatic detection, positioning, evaluation, and early warning of local high-temperature areas through threshold segmentation, positioning quantification, comprehensive index calculation, and early warning classification. Finally, combined with the transmission line BIM model, a visual warning model and a differentiated inspection and maintenance plan are generated, which greatly improves the intelligence level and practicality of local heating monitoring of transmission lines.

[0042] In a specific embodiment, the process of executing step S101 may specifically include the following steps:

[0043] (1) Perform wavelet transform on the temperature data to obtain the high-frequency coefficients corresponding to the temperature data, and perform inverse wavelet transform on the temperature data according to the high-frequency coefficients to obtain the denoised temperature data;

[0044] (2) grayscale stretching is performed on the infrared thermal image to obtain a stretched infrared thermal image;

[0045] (3) Performing histogram equalization processing on the stretched infrared thermal image to obtain a balanced infrared thermal image;

[0046] (4) Performing median filtering on the balanced infrared thermal image to obtain a filtered infrared thermal image;

[0047] (5) removing the illumination component of the filtered infrared thermal image to obtain a candidate infrared thermal image;

[0048] (6) Analyze the pixel gradient data of the candidate infrared thermal image to obtain the gradient amplitude map and gradient direction map;

[0049] (7) Perform non-maximum suppression on the gradient amplitude map to obtain a candidate gradient amplitude map, and perform hysteresis threshold processing on the candidate gradient amplitude map to obtain a target edge map;

[0050] (8) Based on the gradient direction map and the target edge map, the candidate infrared thermal image is sharpened to obtain enhanced thermal image data.

[0051] Specifically, for temperature data, due to the influence of environmental noise, the collected temperature signals often contain high-frequency noise components. Directly using these noisy temperature data for analysis and judgment will affect the accuracy of monitoring. In order to remove high-frequency noise in temperature data, the wavelet transform method can be used. Wavelet transform is a time-frequency analysis tool that can decompose time domain signals into wavelet coefficients of different scales and frequencies. By thresholding the wavelet coefficients, high-frequency noise can be effectively filtered out. The specific steps are to perform wavelet decomposition on the temperature data to obtain wavelet coefficients at different scales, where the high-frequency coefficients correspond to the noise components in the temperature signal. By performing soft threshold or hard threshold processing on the high-frequency coefficients, the high-frequency coefficients less than the threshold are set to zero to remove the high-frequency noise. Then, the processed wavelet coefficients are reconstructed by wavelets to obtain denoised temperature data, which is called denoised temperature data. For example, for the temperature data collected by a temperature sensor on a transmission line, the original data contains obvious high-frequency fluctuations. Through three-layer wavelet decomposition, the high-frequency noise coefficient is extracted, and the part of the high-frequency coefficient that is less than the threshold is set to zero. Then wavelet reconstruction is performed to obtain smooth denoised temperature data, which effectively removes the high-frequency interference in the original data.

[0052] For infrared thermal images, due to the complexity of the imaging environment, the image quality may be affected by many factors, such as uneven lighting, background interference, etc., and a series of image enhancement processing is required for the infrared thermal image to improve the contrast, clarity and signal-to-noise ratio of the image. First, the infrared thermal image is subjected to grayscale stretching processing to map the grayscale value of the image to a larger dynamic range to enhance the contrast of the image. The specific method is to determine the upper and lower thresholds of grayscale stretching according to the grayscale histogram of the infrared thermal image, map the grayscale value below the lower threshold to 0, and map the grayscale value above the upper threshold to 255, and linearly stretch the intermediate grayscale value in proportion to obtain the infrared thermal image after grayscale stretching, which is called a stretched infrared thermal image. After stretching, the grayscale distribution of the infrared thermal image is more balanced, and the contrast between the high temperature area and the low temperature area is more obvious. Then, the stretched infrared thermal image is subjected to histogram equalization processing to further enhance the image contrast and grayscale distribution uniformity. Histogram equalization is an image enhancement method based on probability distribution. By adjusting the grayscale histogram of the image to a uniform distribution, the grayscale distribution of the image is more balanced and the contrast of the low-contrast area is improved. The stretched infrared thermal image is processed by histogram equalization to obtain a balanced infrared thermal image, and the overall contrast and visual effect of the image are further improved.

[0053] In order to remove high-frequency noise and salt and pepper noise in infrared thermal images, it is necessary to perform median filtering on the balanced infrared thermal images. Median filtering is a nonlinear smoothing filtering method that replaces the grayscale value of the center pixel of the window with the median of the pixel grayscale value in the sliding window. It can effectively suppress the impulse noise and salt and pepper noise in the image while maintaining the clarity of the image edge. The balanced infrared thermal image is subjected to median filtering to obtain a filtered infrared thermal image. The high-frequency noise and salt and pepper noise in the image are effectively suppressed, and the image becomes smoother.

[0054] During the imaging process of infrared thermal images, due to the uneven lighting conditions, there may be illumination components in the image, that is, the problem of uneven background brightness. In order to eliminate the impact of uneven lighting on image quality, it is necessary to remove the illumination component from the filtered infrared thermal image. Commonly used illumination component removal methods include Retinex algorithm, homomorphic filtering, etc. Taking the Retinex algorithm as an example, the algorithm assumes that the image is the product of the illumination component and the reflection component. By performing multi-scale Gaussian filtering on the image, the illumination component is estimated, and then the original image is divided by the illumination component to obtain the reflection component, that is, the image with uneven lighting removed. Retinex illumination component removal processing is performed on the filtered infrared thermal image to obtain a candidate infrared thermal image. The uneven lighting problem in the image is improved, and the background brightness is more uniform and consistent.

[0055] In order to further extract the edge and contour information in the infrared thermal image, it is necessary to analyze the pixel gradient data of the candidate infrared thermal image. The gradient is the direction in which the grayscale of the image changes fastest, and the gradient amplitude is larger at the edges and contours. By calculating the gradient amplitude and gradient direction of each pixel in the candidate infrared thermal image, the gradient amplitude map and gradient direction map are obtained. The gradient amplitude map reflects the drastic degree of grayscale change in the image, and the gradient amplitude is larger at the edges and contours; the gradient direction map reflects the direction information of the grayscale change in the image, and the gradient direction at the edges and contours is perpendicular to the edge direction.

[0056] Non-maximum suppression is performed on the gradient amplitude map to refine the edges and remove non-edge pixels. The principle of non-maximum suppression is to compare the gradient amplitude of the pixel point in the gradient direction. If the gradient amplitude of the point is not the local maximum, its gradient amplitude is set to 0, and the pixel point with the largest local gradient amplitude is retained to obtain the candidate gradient amplitude map. Then, the candidate gradient amplitude map is subjected to hysteresis threshold processing, and the gradient amplitude is divided into three categories: strong edge, weak edge and non-edge according to the high threshold and the low threshold. Pixels above the high threshold are considered strong edges, pixels below the low threshold are considered non-edges, and pixels between the two thresholds are considered weak edges. Through hysteresis threshold processing, strong edge pixels are retained, non-edge pixels are removed, and weak edge pixels are connected to strong edge pixels to obtain a complete target edge map.

[0057] Finally, based on the gradient direction map and the target edge map, the candidate infrared thermal image is edge sharpened to obtain enhanced thermal image data. The purpose of edge sharpening is to enhance the edge and contour information in the image and improve the clarity and sharpness of the image. The specific method is to determine the edge direction of each pixel according to the gradient direction map, and then perform weighted superposition on the edge pixels in the target edge map in the edge direction to enhance the grayscale value of the edge pixels and make the edge more prominent. At the same time, according to the edge weighting coefficient, the grayscale value of the pixel at the corresponding position in the candidate infrared thermal image is corrected to make the grayscale change at the edge more obvious, and the enhanced thermal image data after edge sharpening is obtained.

[0058] For example, for an infrared thermal image of a photovoltaic power station transmission line, the original image is dim, with low contrast, uneven illumination and noise interference. By performing grayscale stretching on the original infrared thermal image, the grayscale distribution of the image is more balanced, and the contrast between the high-temperature area and the low-temperature area is enhanced. Then, after histogram equalization, the overall contrast and visual effect of the image are further improved. Through median filtering, the high-frequency noise and salt-and-pepper noise in the image are effectively suppressed. The illumination component of the image is removed by the Retinex algorithm, eliminating the problem of uneven background brightness, and the image background is more consistent. Then, through pixel gradient data analysis, the gradient amplitude map and gradient direction map of the image are obtained, highlighting the edge and contour information in the image. After non-maximum suppression and hysteresis threshold processing, the target edge map of the image is obtained, and the edge pixels in the image are accurately extracted. Finally, based on the gradient direction map and the target edge map, the original infrared thermal image is edge sharpened to make the edges and contours clearer and sharper, and the enhanced thermal image data is obtained. Through a series of image preprocessing and enhancement operations, the quality and usability of infrared thermal images have been significantly improved, laying a good data foundation for subsequent temperature anomaly analysis and fault diagnosis.

[0059] In a specific embodiment, the process of executing step S102 may specifically include the following steps:

[0060] (1) Obtaining the size of the denoised temperature data and the size of the enhanced thermal image data, and performing target interpolation size analysis based on the size of the denoised temperature data and the size of the enhanced thermal image data to obtain the temperature interpolation size and the thermal image interpolation size;

[0061] (2) According to the temperature interpolation size, two-dimensional cubic spline interpolation is performed on the denoised temperature data to obtain interpolated temperature data;

[0062] (3) According to the thermal image interpolation size, gradient interpolation processing is performed on the enhanced thermal image data to obtain interpolated thermal image data;

[0063] (4) performing normalization processing on the interpolated temperature data to obtain normalized temperature data, and performing normalization processing on the interpolated thermal image data to obtain normalized thermal image data;

[0064] (5) Based on the normalized temperature data, bilinear interpolation is performed on the denoised temperature data to obtain a candidate temperature matrix;

[0065] (6) Based on the normalized thermal image data, bicubic interpolation is performed on the enhanced thermal image data to obtain a candidate thermal image matrix;

[0066] (7) The candidate temperature matrix is ​​smoothed to obtain the target temperature matrix, and the candidate thermal image matrix is ​​Gaussian filtered to obtain the target thermal image matrix.

[0067] Specifically, the size of the denoised temperature data and the size of the enhanced thermal image data are obtained. Usually, the temperature data is a one-dimensional time series, while the infrared thermal image is a two-dimensional image matrix, and the size and resolution of the two may be different. In order to determine the target size of the interpolation processing, it is necessary to perform a target interpolation size analysis based on the size of the denoised temperature data and the size of the enhanced thermal image data. By comparing the sizes of the two, a suitable target interpolation size is selected so that the interpolated temperature data and thermal image data have the same number of rows and columns. Generally speaking, the target interpolation size can be selected to be the same as the size of the enhanced thermal image data, so that the spatial resolution of the thermal image can be maintained. According to the results of the target interpolation size analysis, the temperature interpolation size and the thermal image interpolation size are obtained to guide the subsequent temperature data interpolation and thermal image data interpolation, respectively.

[0068] For the denoised temperature data, the two-dimensional cubic spline interpolation method is used to interpolate it according to the temperature interpolation size. Two-dimensional cubic spline interpolation is a commonly used interpolation algorithm that generates new data points by constructing a cubic spline function to smoothly interpolate between known data points. Specifically, based on the timestamp and temperature value of the denoised temperature data, a cubic spline function is constructed in the time dimension, and then interpolation is performed in the spatial dimension according to the temperature interpolation size to obtain interpolated temperature data of the same size as the thermal image data. The interpolated temperature data not only maintains the changing trend of the original temperature data, but also has a higher spatial resolution, which matches the size of the thermal image data.

[0069] For enhanced thermal image data, the gradient interpolation method is used to interpolate it according to the thermal image interpolation size. Gradient interpolation is an interpolation algorithm based on gradient information. By calculating the gradient direction and gradient amplitude of the local area of ​​the image, the structure and edge information of the image are considered in the interpolation process, so as to obtain a smoother and more natural interpolation result. Specifically, the gradient of the enhanced thermal image data is calculated to obtain the gradient amplitude map and the gradient direction map, and then the image is interpolated in the gradient direction according to the thermal image interpolation size to obtain the interpolated thermal image data. Compared with traditional bilinear interpolation or bicubic interpolation, gradient interpolation can better maintain the edge and texture details of the image and avoid blurring or jagged effects after interpolation.

[0070] In order to facilitate subsequent data fusion and analysis, the interpolated temperature data and interpolated thermal image data need to be normalized. The purpose of normalization is to map data of different scales and ranges to the same fixed interval, usually the [0,1] interval, so as to eliminate the dimensional influence between different data and make the data comparable. The interpolated temperature data is normalized, and the temperature value is mapped to the [0, 1] interval to obtain normalized temperature data. Similarly, the interpolated thermal image data is normalized, and the grayscale value is mapped to the [0, 1] interval to obtain normalized thermal image data. After normalization, the value range of temperature data and thermal image data is consistent, which is convenient for subsequent fusion calculations.

[0071] After obtaining the normalized temperature data and the normalized thermal image data, the de-noised temperature data and the enhanced thermal image data are interpolated to obtain the candidate temperature matrix and the candidate thermal image matrix. For the de-noised temperature data, the bilinear interpolation method is used to interpolate it according to the normalized temperature data. Bilinear interpolation is a simple and efficient two-dimensional interpolation algorithm. The value of the interpolation point is obtained by weighted averaging the values ​​of four adjacent pixels. Specifically, according to the normalized temperature data, the de-noised temperature data is bilinearly interpolated to obtain a candidate temperature matrix of the same size as the thermal image data. The candidate temperature matrix not only has the same spatial resolution as the thermal image data, but also the temperature value is consistent with the normalized temperature data.

[0072] For the enhanced thermal image data, the bicubic interpolation method is used to interpolate it according to the normalized thermal image data. Bicubic interpolation is a high-quality two-dimensional interpolation algorithm. It obtains the value of the interpolation point by weighted averaging the values ​​of 16 adjacent pixels. Compared with bilinear interpolation, bicubic interpolation can obtain smoother and more natural interpolation results and avoid the block effect after interpolation. Specifically, according to the normalized thermal image data, bicubic interpolation is performed on the enhanced thermal image data to obtain a candidate thermal image matrix. The candidate thermal image matrix not only has the same spatial resolution as the denoised temperature data, but also has the same grayscale value as the normalized thermal image data, while maintaining the smoothness and continuity of the image.

[0073] Finally, the candidate temperature matrix and the candidate thermal image matrix are post-processed to obtain the final target temperature matrix and the target thermal image matrix. For the candidate temperature matrix, a smoothing method is used to eliminate the spike noise or mutation points that may be introduced during the interpolation process, so that the temperature distribution is smoother and continuous. Common smoothing methods include moving average filtering, Gaussian filtering, etc. By smoothing the candidate temperature matrix, the target temperature matrix is ​​obtained, the temperature distribution is more stable, and the noise is suppressed. For the candidate thermal image matrix, the Gaussian filtering method is used to further eliminate the noise and artifacts that may be introduced during the interpolation process, and improve the smoothness and continuity of the image. Gaussian filtering is a commonly used image smoothing filter. By performing Gaussian weighted averaging on the image, a smoothed image is obtained. Gaussian filtering is performed on the candidate thermal image matrix to obtain the target thermal image matrix. The image is smoother and more natural, and the details are retained.

[0074] For example, for local heating monitoring of a certain transmission line, the size of the original thermal image collected by the infrared thermal imager is 640×480, and the original temperature data collected by the temperature sensor is one sampling point per minute, totaling 1440 data points. After denoising and enhancement processing, enhanced thermal image data with a size of 640×480 and denoised temperature data with a length of 1440 are obtained. In order to achieve data fusion, the denoised temperature data needs to be interpolated to the same size as the enhanced thermal image data. Through the target interpolation size analysis, the temperature interpolation size is determined to be 640×480, that is, 1440 temperature data points are interpolated into a 640×480 two-dimensional matrix. The 2D cubic spline interpolation method is used to interpolate the denoised temperature data according to the temperature interpolation size, and the interpolated temperature data with a size of 640×480 is obtained. For the enhanced thermal image data, since its size is consistent with the target interpolation size, no interpolation processing is required. Next, the interpolated temperature data and enhanced thermal image data are normalized, and their value ranges are mapped to the interval [0, 1] to obtain normalized temperature data and normalized thermal image data. Then, based on the normalized temperature data, the denoised temperature data is bilinearly interpolated to obtain a candidate temperature matrix of size 640×480; based on the normalized thermal image data, the enhanced thermal image data is bicubicly interpolated to obtain a candidate thermal image matrix of size 640×480. Finally, the candidate temperature matrix is ​​smoothed to eliminate the noise introduced in the interpolation process to obtain the target temperature matrix; the candidate thermal image matrix is ​​Gaussian filtered to improve the smoothness and continuity of the image to obtain the target thermal image matrix.

[0075] In a specific embodiment, the process of executing step S103 may specifically include the following steps:

[0076] (1) Perform multi-scale decomposition on the temperature matrix and thermal image matrix respectively to obtain a multi-scale temperature feature matrix and a multi-scale thermal image feature matrix;

[0077] (2) Performing joint sparse coding on the multi-scale temperature feature matrix and the multi-scale thermal image feature matrix to obtain a multi-scale temperature-thermal image feature matrix;

[0078] (3) Perform local linear embedding on the multi-scale temperature-thermal image feature matrix to obtain a low-dimensional temperature-thermal image feature matrix;

[0079] (4) Perform deep feature extraction on the low-dimensional temperature-thermal image feature matrix to obtain a deep temperature-thermal image fusion feature matrix;

[0080] (5) Adaptively select regions for the deep temperature-thermal image fusion feature matrix to obtain a candidate fusion region matrix;

[0081] (6) Perform superpixel segmentation on the candidate fusion region matrix to obtain multiple segmentation region matrices;

[0082] (7) Perform global optimization processing on multiple segmented region matrices to obtain a fused temperature distribution matrix.

[0083] Specifically, the temperature matrix and thermal image matrix are subjected to multi-scale decomposition to extract feature information at different scales. Multi-scale decomposition is a commonly used signal processing technology. By performing multi-resolution analysis on the signal, approximate coefficients and detail coefficients at different scales are obtained to reflect the characteristics of the signal in different frequency bands. For the temperature matrix, multi-scale decomposition methods such as wavelet transform are used to decompose the temperature matrix into approximate coefficient matrices and detail coefficient matrices at different scales to obtain a multi-scale temperature feature matrix, which contains the feature information of the temperature signal at different frequency bands. Similarly, the thermal image matrix is ​​subjected to multi-scale decomposition to obtain a multi-scale thermal image feature matrix, which contains the feature information of the thermal image at different frequency bands. After multi-scale decomposition, the temperature features and thermal image features are represented more comprehensively and meticulously.

[0084] In order to mine the correlation between temperature features and thermal imaging features, it is necessary to perform joint sparse coding on the multi-scale temperature feature matrix and the multi-scale thermal imaging feature matrix. Joint sparse coding is a multimodal data fusion method that maps data of different modalities to the same sparse coding space and learns their common sparse representation, thereby realizing data association and fusion. Specifically, the multi-scale temperature feature matrix and the multi-scale thermal imaging feature matrix are used as input, and a common over-complete dictionary and sparse coding coefficients are learned through joint sparse coding algorithms such as dictionary learning and sparse autoencoders to obtain a multi-scale temperature-thermal imaging feature matrix. This matrix not only retains the essential information of temperature features and thermal imaging features, but also mines the correlation between them, realizing the preliminary fusion of temperature data and thermal imaging data.

[0085] However, the dimensionality of the multi-scale temperature-thermal imaging feature matrix may be high, containing a large amount of redundant and noisy information. In order to further extract compact and effective feature representations, it is necessary to reduce the dimensionality of the multi-scale temperature-thermal imaging feature matrix. Local linear embedding (LLE) is a dimensionality reduction algorithm for manifold learning. It maps high-dimensional data to a low-dimensional space by maintaining the linear reconstruction relationship of the local neighborhood structure while retaining the intrinsic manifold structure of the data. The multi-scale temperature-thermal imaging feature matrix is ​​locally linearly embedded to learn its intrinsic low-dimensional manifold representation and obtain a low-dimensional temperature-thermal imaging feature matrix. The low-dimensional feature matrix not only reduces the redundancy of the data, but also retains the essential information of the temperature features and thermal imaging features, making subsequent feature extraction and fusion more efficient and accurate.

[0086] After obtaining the low-dimensional temperature-thermal image feature matrix, the deep learning method is further used to extract deep features from it. Deep learning is a multi-level feature learning method that automatically learns the hierarchical feature representation of data by building a multi-layer neural network. For deep feature extraction of the low-dimensional temperature-thermal image feature matrix, deep learning models such as convolutional neural network (CNN) and long short-term memory network (LSTM) can be used. Through multi-layer nonlinear transformation and feature combination, the high-level semantic representation of temperature features and thermal image features is learned to obtain a deep temperature-thermal image fusion feature matrix. The deep fusion feature matrix contains the deep and abstract feature representation of temperature data and thermal image data, which can more accurately and comprehensively characterize the characteristic pattern of the local heating area.

[0087] In order to further highlight the feature representation of local fever areas, it is necessary to perform adaptive region selection on the deep temperature-thermal image fusion feature matrix. Adaptive region selection is a region extraction method based on saliency and prior knowledge. By calculating the saliency measure of each region in the feature matrix and combining prior knowledge and constraints, the most important and relevant regions for local fever judgment are automatically selected. Specifically, a saliency analysis is performed on the deep temperature-thermal image fusion feature matrix, and the saliency score of each region is calculated. Then, based on the saliency score and prior knowledge, such as the shape, size, and position of the fever area, candidate fusion regions are adaptively selected to obtain a candidate fusion region matrix. The candidate fusion region matrix contains the most important and relevant regions for local fever judgment, providing key focus areas for subsequent refined analysis and fever degree assessment.

[0088] Perform superpixel segmentation on the candidate fusion region matrix and divide it into multiple locally consistent regions with complete edges. Superpixel segmentation is an image segmentation method based on graph theory. It achieves over-segmentation and local representation of the image by dividing the image into multiple locally consistent small regions (superpixels) with complete edges. Superpixel segmentation algorithms such as SLIC and GraphCut can be used to perform superpixel segmentation on the candidate fusion region matrix. According to the similarity of the color, texture, temperature and other features of the pixels in the region, the candidate fusion region matrix is ​​divided into multiple locally consistent segmentation region matrices with complete edges. Each segmentation region matrix represents a locally consistent fever sub-region within the candidate fusion region, providing a refined analysis unit for subsequent quantification and early warning analysis of the fever degree.

[0089] Finally, multiple segmented area matrices are globally optimized to obtain the final fused temperature distribution matrix. Global optimization is an optimization method based on minimization of energy functions. By constructing a global energy function and considering the spatial correlation and consistency between segmented areas, multiple segmented area matrices are optimized as a whole to obtain a globally optimal temperature distribution representation. Specifically, a global energy function is constructed, including a data term and a smoothing term. The data term measures the consistency of the temperature distribution in each segmented area, and the smoothing term measures the continuity of the temperature distribution between adjacent segmented areas. By minimizing the global energy function, the optimal fused temperature distribution matrix is ​​obtained, which realizes the deep fusion of temperature data and thermal image data, and obtains a globally consistent, continuous and smooth temperature distribution representation.

[0090] For example, for local heating monitoring of a certain transmission line, the thermal images collected by the infrared thermal imager and the temperature data collected by the temperature sensor are subjected to the previous preprocessing steps such as denoising, enhancement, and interpolation to obtain the target temperature matrix and target thermal image matrix of size 640×480. First, the temperature matrix and thermal image matrix are subjected to wavelet transform respectively to obtain the multi-scale temperature feature matrix and multi-scale thermal image feature matrix under the three-layer wavelet decomposition, which contain the feature representation of temperature signals and thermal images at different frequency scales. Then, the multi-scale temperature feature matrix and the multi-scale thermal image feature matrix are jointly sparsely encoded, and a 256-dimensional overcomplete dictionary and sparse coding coefficient matrix are obtained through the dictionary learning algorithm, realizing the preliminary fusion of temperature features and thermal image features. Next, the multi-scale temperature-thermal image feature matrix is ​​locally linearly embedded to reduce its dimension to 64 dimensions, and a low-dimensional temperature-thermal image feature matrix is ​​obtained, which retains the essential information of temperature features and thermal image features. The low-dimensional temperature-thermal image feature matrix is ​​subjected to deep feature extraction of convolutional neural network. Through three convolutional layers and two fully connected layers, a 256-dimensional deep temperature-thermal image fusion feature matrix is ​​obtained, and the deep semantic representation of temperature data and thermal image data is learned. Then, adaptive region selection is performed on the deep temperature-thermal image fusion feature matrix. According to the significance score of the region and prior knowledge, 64 candidate fusion regions are selected to obtain the candidate fusion region matrix. SLIC superpixel segmentation is performed on the candidate fusion region matrix, which is divided into 128 locally consistent segmentation region matrices, each of which corresponds to a local heating sub-region. Finally, a global energy function is constructed, and the 128 segmentation region matrices are globally optimized to obtain a fused temperature distribution matrix, which realizes the deep fusion of temperature data and thermal image data and obtains a globally consistent, continuous and smooth temperature distribution representation. Through a series of data fusion and feature extraction steps such as multi-scale decomposition, joint sparse coding, local linear embedding, deep feature extraction, adaptive region selection, superpixel segmentation and global optimization, the complementary characteristics of temperature data and thermal image data are fully utilized to obtain a refined representation of local fever areas, laying a solid data foundation for subsequent quantification of fever levels and early warning analysis.

[0091] In a specific embodiment, the process of executing step S104 may specifically include the following steps:

[0092] (1) Binarize the temperature distribution matrix to obtain a binary temperature distribution matrix;

[0093] (2) Performing morphological opening operation on the binary temperature distribution matrix to obtain an open binary temperature distribution matrix;

[0094] (3) Performing an open operation on the binary temperature distribution matrix to extract connected regions and obtain a high-temperature connected region matrix;

[0095] (4) Analyze the area and circumscribed rectangle of the high-temperature connected region matrix to obtain a set of candidate high-temperature region matrices;

[0096] (5) Extracting the centroid coordinates of each high-temperature region according to the candidate high-temperature region matrix set to obtain the initial heating position coordinate set;

[0097] (6) Performing spatial clustering on the initial fever location coordinate set, merging adjacent fever location coordinates, and obtaining a fever location coordinate set;

[0098] (7) According to the clustered fever position coordinate set, the temperature value of each fever position is extracted from the temperature distribution matrix to obtain a fever position temperature value set;

[0099] (8) Analyze the temperature value set of the heating position to obtain the average temperature and the maximum temperature;

[0100] (9) Based on the average temperature and the maximum temperature, the heating degree of each heating location is analyzed to obtain a heating degree data set.

[0101] Specifically, a series of image processing and data analysis operations are performed on the fused temperature distribution matrix. First, the fused temperature distribution matrix is ​​binarized to convert the temperature value into a binary image and highlight the high temperature area. Specifically, a temperature threshold is set, and the temperature values ​​above the threshold are set to 1, and the temperature values ​​below the threshold are set to 0 to obtain a binary temperature distribution matrix. The binarization process converts the fused temperature distribution matrix into a black and white image, in which the white area represents the high temperature area and the black area represents the normal temperature area, which facilitates subsequent morphological operations and area extraction.

[0102] In order to eliminate isolated noise points and small areas in the binary temperature distribution matrix, a morphological opening operation is required. Morphological opening operation is an image morphological operation. By corroding and dilating the image, the noise points and small areas in the image are removed to obtain a smoother and more complete area. The morphological opening operation is performed on the binary temperature distribution matrix. First, a structural element is used to corrode it to eliminate isolated white pixels and small white areas, and then the same structural element is used to dilate it to restore the edges of the corroded white areas. Through the morphological opening operation, the binary temperature distribution matrix with open operation is obtained, in which the white areas represent relatively complete and stable high-temperature areas, laying the foundation for the subsequent connected area extraction.

[0103] After obtaining the binary temperature distribution matrix of the opening operation, it is necessary to extract the connected regions, and form the adjacent white pixels into connected regions to represent a complete high temperature region. Connected region extraction is an image segmentation algorithm that scans each pixel in the image and divides it into different regions according to the connectivity between pixels. The connected region extraction of the binary temperature distribution matrix of the opening operation can be performed by using a graph-based depth-first search (DFS) algorithm or a region-growing-based algorithm. Taking the DFS algorithm as an example, starting from the upper left corner of the image, each pixel is scanned row by row. When a white pixel is encountered, it is marked as a new connected region, and then the adjacent white pixels in the four directions of up, down, left, and right are recursively searched, and they are added to the same connected region until all adjacent white pixels are visited. Through the connected region extraction, the high temperature connected region matrix is ​​obtained, in which each connected region represents a complete high temperature region, which provides the basis for region division for the subsequent regional area and circumscribed rectangle analysis.

[0104] The high-temperature connected region matrix is ​​analyzed for regional area and regional bounding rectangle, and the area size of each high-temperature region and the position and size of the minimum bounding rectangle are calculated. The regional area reflects the size and range of the high-temperature region, and the regional bounding rectangle reflects the position and shape of the high-temperature region. For each connected region, the number of white pixels inside it is counted, which is the area size of the region; then, the maximum and minimum horizontal and vertical coordinates of all white pixels in the region are found to form the upper left and lower right corner coordinates of the minimum bounding rectangle of the region, and the width and height of the bounding rectangle are calculated. Through the analysis of regional area and regional bounding rectangle, a set of candidate high-temperature region matrices is obtained, in which each matrix represents a candidate high-temperature region, and contains attribute information such as the area, position and size of the bounding rectangle of the region, which provides an important regional attribute reference for the subsequent extraction of heating position coordinates and temperature value analysis.

[0105] According to the candidate high-temperature region matrix set, the centroid coordinates of each high-temperature region are extracted as the heating position coordinates of the region. The centroid coordinates are the arithmetic mean of all pixel coordinates in the high-temperature region, reflecting the geometric center position of the high-temperature region. For each matrix in the candidate high-temperature region matrix set, the arithmetic mean of the horizontal and vertical coordinates of all white pixels is calculated to obtain the centroid coordinates of the region and form an initial heating position coordinate set. The initial heating position coordinate set represents the geometric center position of each high-temperature region, but there may be some adjacent and overlapping heating position coordinates that need to be further merged and optimized.

[0106] In order to merge adjacent and overlapping fever location coordinates, it is necessary to perform spatial clustering analysis on the initial fever location coordinate set. Spatial clustering is a data analysis method that divides adjacent and similar objects in space into different categories. Common clustering algorithms such as K-means clustering and DBSCAN clustering can be used to perform spatial clustering on the initial fever location coordinate set. Taking DBSCAN clustering as an example, according to the Euclidean distance between the fever location coordinates, the coordinate points with a distance less than a certain threshold are divided into the same category to form a cluster cluster, which represents a merged fever location. Through spatial clustering analysis, the adjacent and overlapping fever location coordinates are merged into one fever location to obtain a fever location coordinate set, in which each coordinate represents an independent and non-overlapping fever location, providing an accurate location reference for subsequent temperature value extraction and fever degree analysis.

[0107] According to the clustered fever position coordinate set, the temperature value of each fever position is extracted from the original fused temperature distribution matrix to obtain the fever position temperature value set. The fever position temperature value reflects the actual temperature level at the position and is an important basis for evaluating the degree of fever. For each coordinate in the fever position coordinate set, the pixel position corresponding to the coordinate is found in the fused temperature distribution matrix, and the temperature value of the pixel is extracted to form the fever position temperature value set. The fever position temperature value set contains the actual temperature readings of each fever position, providing quantitative temperature data support for subsequent fever degree analysis.

[0108] The temperature value set of the fever location is statistically analyzed to obtain the average temperature and maximum temperature of each fever location. The average temperature reflects the overall temperature level of the fever location, and the maximum temperature reflects the local peak temperature of the fever location. For each subset of temperature values ​​in the fever location temperature value set, the arithmetic mean and maximum value are calculated to obtain the average temperature and maximum temperature of the fever location, respectively. By calculating the average temperature and the maximum temperature, the temperature statistical characteristics of each fever location are obtained, which provides a quantitative temperature index for the subsequent fever degree analysis.

[0109] Finally, according to the average temperature and the maximum temperature, the fever degree of each fever location is analyzed and evaluated to obtain the fever degree data set. The fever degree reflects the temperature abnormality level and danger level of the fever location, and is the direct basis for early warning and maintenance decisions. According to the average temperature and the maximum temperature, the quantitative standards and thresholds of the fever degree are set, and the fever locations are divided into different fever degree levels, such as normal, mild, moderate, severe, etc. For each fever location, the fever degree level to which it belongs is determined according to its average temperature and the maximum temperature, and the fever degree data of the location is obtained to form a fever degree data set. The fever degree data set quantifies the temperature abnormality level and danger level of each fever location, providing a direct decision-making basis for subsequent early warning level judgment and maintenance strategy formulation.

[0110] For example, for local heating monitoring of a certain transmission line, a 640×480 fused temperature distribution matrix is ​​obtained through data fusion, in which the temperature value of each pixel ranges from 0 to 100 degrees Celsius. First, the temperature threshold is set to 60 degrees Celsius, and the fused temperature distribution matrix is ​​binarized to obtain a binary temperature distribution matrix, in which pixels with temperatures above 60 degrees Celsius are white and pixels with temperatures below 60 degrees Celsius are black. Then, the binary temperature distribution matrix is ​​morphologically opened, using a 5×5 rectangular structure element, first performing an erosion operation and then an expansion operation to obtain an open binary temperature distribution matrix, removing isolated noise points and small areas. Next, the connected region extraction is performed on the open binary temperature distribution matrix, and an algorithm based on region growing is used to obtain 10 high-temperature connected regions to form a high-temperature connected region matrix. The high-temperature connected region matrix is ​​analyzed for regional area and bounding rectangle, and the attributes such as the area, position and size of the bounding rectangle of each region are obtained to form a candidate high-temperature region matrix set. According to the candidate high temperature area matrix set, the centroid coordinates of each area are extracted to obtain 10 initial fever position coordinates. The initial fever position coordinates are subjected to DBSCAN cluster analysis, and the coordinates with a distance of less than 20 pixels are merged into the same class, and finally 5 clustered fever position coordinates are obtained. According to the clustered fever position coordinates, the temperature values ​​of the corresponding positions are extracted in the fusion temperature distribution matrix to obtain the temperature value sets of 5 fever positions. Statistical analysis is performed on each temperature value set, and the average temperature and the maximum temperature are calculated to obtain the temperature statistical characteristics of each fever position. Finally, according to the temperature statistical characteristics, the quantitative standard of fever degree is set, and the fever position with an average temperature higher than 80 degrees Celsius and a maximum temperature higher than 90 degrees Celsius is judged as severe fever, the fever position with an average temperature higher than 70 degrees Celsius but a maximum temperature lower than 90 degrees Celsius is judged as moderate fever, and the fever position with an average temperature higher than 60 degrees Celsius but lower than 70 degrees Celsius is judged as mild fever. The fever degree data of each fever position is obtained to form a fever degree data set. Through a series of operations such as binarization, morphological opening, connected region extraction, regional area and circumscribed rectangle analysis, fever location coordinate extraction, spatial clustering, temperature value extraction, temperature statistical analysis and fever degree analysis, the local high temperature areas are accurately extracted and located from the fused temperature distribution matrix, and the fever degree of each high temperature area is quantitatively analyzed, laying a data foundation for subsequent early warning and maintenance decisions.

[0111] In a specific embodiment, the process of executing step S105 may specifically include the following steps:

[0112] (1) Calculate the spatial aggregation of the fever position according to the fever position coordinate set to obtain the fever position aggregation;

[0113] (2) Analyze the fever degree statistical characteristics of the fever degree data set to obtain the fever degree statistical characteristics;

[0114] (3) Perform weighted summation of fever location aggregation and fever degree statistical characteristics to obtain the initial value of the comprehensive fever index;

[0115] (4) Based on the preset spatial distribution influencing factor, the initial value of the comprehensive fever index is corrected to obtain the comprehensive fever index;

[0116] (5) Constructing fever warning level determination conditions based on the comprehensive fever index to obtain multi-level fever warning level determination conditions;

[0117] (6) Logically compare the comprehensive fever index with the multi-level fever warning level judgment conditions to obtain the initial result of the local fever warning level;

[0118] (7) Perform time series analysis on the initial results of the local fever warning level to obtain the time-smoothed local fever warning level;

[0119] (8) Perform spatial correlation analysis on the time-smoothed local fever warning level to obtain the local fever warning level.

[0120] Specifically, according to the set of fever location coordinates, the spatial aggregation of fever locations is calculated to obtain the fever location aggregation index. Spatial aggregation reflects the density of the spatial distribution of fever locations and is an important factor in assessing fever risk. Spatial statistical methods such as kernel density estimation (KDE) can be used to calculate spatial aggregation. Taking KDE as an example, for each fever location coordinate, the number of other fever locations within a certain radius around it is calculated to obtain the local density value of the location. Then, the local density values ​​of all fever locations are normalized to obtain the fever location aggregation index, which ranges from 0 to 1. The larger the value, the denser the distribution of the fever locations in space and the higher the aggregation.

[0121] Next, the fever degree data set is analyzed for fever degree statistical characteristics to obtain fever degree statistical characteristics. The fever degree statistical characteristics reflect the overall distribution of temperature anomaly levels and dangerousness at fever locations. Descriptive statistical analysis is performed on the fever degree data set to calculate the maximum, minimum, mean, standard deviation, median and other statistical quantities of the fever degree to obtain the basic statistical characteristics of the fever degree. At the same time, a histogram analysis is performed on the fever degree data to statistically analyze the frequency distribution of different fever degree levels to obtain the frequency distribution characteristics of the fever degree. The fever degree statistical characteristics depict the distribution of the fever degree of the transmission line as a whole, providing an important basis for the comprehensive assessment of fever risks.

[0122] After obtaining the statistical characteristics of the heat location concentration and the degree of heat, they need to be weighted and summed to obtain the initial value of the comprehensive heat index. The initial value of the comprehensive heat index is the weighted average of the statistical characteristics of the heat location concentration and the degree of heat, which reflects the overall heat risk level of the transmission line. According to the importance of the statistical characteristics of the heat location concentration and the degree of heat, their weight coefficients are set, multiplied by the corresponding index values ​​respectively, and then summed to obtain the initial value of the comprehensive heat index. The setting of the weight coefficient needs to be combined with the actual operation of the transmission line and expert experience. Generally speaking, the weight of the statistical characteristics of the degree of heat should be greater than the weight of the heat location concentration, because the degree of heat directly reflects the severity of the temperature anomaly and has a greater impact on the safe operation of the transmission line.

[0123] However, the initial value of the comprehensive heating index only considers the statistical characteristics of the aggregation degree and degree of heating of the heating position, and does not consider the distribution of the heating position in the space of the transmission line. In order to more comprehensively evaluate the heating risk of the transmission line, it is necessary to introduce the spatial distribution influencing factor, correct the initial value of the comprehensive heating index, and obtain the final comprehensive heating index. The spatial distribution influencing factor reflects the importance and influence of the heating position in different parts and sections of the transmission line. According to the spatial structure and power load distribution of the transmission line, the spatial distribution influencing factors of different parts and sections are preset, and the value range is 0~1. The larger the value, the higher the importance of the part or section, and the greater the impact of the heating risk. For each heating position, according to its location and section, multiply it by the corresponding spatial distribution influencing factor to obtain the correction coefficient of the position. Then, the initial value of the comprehensive heating index is multiplied by the average value of the correction coefficients of all heating positions to obtain the corrected comprehensive heating index, which ranges from 0~100. The larger the value, the higher the overall heating risk of the transmission line.

[0124] The comprehensive fever index quantifies the overall fever risk level of the transmission line and provides a quantitative basis for the judgment of the warning level. In order to convert the comprehensive fever index into an intuitive warning level, it is necessary to construct the fever warning level judgment condition based on the comprehensive fever index to obtain the multi-level fever warning level judgment condition. The fever warning level judgment condition is the threshold interval of the comprehensive fever index, which divides the comprehensive fever index into different warning levels such as normal, warning, and severe. According to the safe operation requirements of the transmission line and historical data analysis, the threshold interval of the comprehensive fever index is set to construct the multi-level fever warning level judgment condition. For example, the comprehensive fever index between 0 and 30 is the normal level, between 30 and 60 is the warning level, and between 60 and 100 is the severe level. The multi-level fever warning level judgment condition provides a quantitative standard for the judgment of the local fever warning level.

[0125] The calculated comprehensive fever index is logically compared with the multi-level fever warning level judgment conditions to determine the warning level to which the comprehensive fever index belongs, and the initial result of the local fever warning level is obtained. The initial result of the local fever warning level reflects the fever risk level of the transmission line at the current moment, but there may be short-term fluctuations and isolated outliers. In order to obtain a more stable and reliable warning level, it is necessary to perform time series analysis on the initial results of the local fever warning level to obtain a time-smoothed local fever warning level. Time series analysis takes into account the continuity and smoothness of the warning level in time. By performing moving average or exponential smoothing on the initial results of the warning level over a period of time, short-term fluctuations and outliers are filtered out to obtain a smoothed warning level sequence, which reflects the overall trend and change law of the warning level.

[0126] Finally, the spatial correlation analysis of the local fever warning level after time smoothing was performed to obtain the final local fever warning level. The spatial correlation analysis takes into account the spatial correlation and consistency of the warning level. By analyzing the correlation between the warning levels of adjacent fever locations, the local fever warning level is spatially corrected and optimized. Specifically, for each fever location, the warning levels of other fever locations within a certain range around it are analyzed, and their mode or weighted average is calculated as the reference warning level of the location. Then, the time-smoothed warning level of the location is compared with the reference warning level. If the difference is large, the warning level of the location is adjusted to make it more consistent with the warning level of the surrounding locations. Through spatial correlation analysis, a more spatially consistent and reliable local fever warning level is obtained, which comprehensively considers the correlation characteristics of the warning level in time and space.

[0127] For example, for local heating monitoring of a certain transmission line, the coordinate sets of 10 heating locations and the corresponding fever degree data sets were obtained through the previous analysis. First, based on the fever location coordinate set, the kernel density estimation method was used to calculate the number of other fever locations within 20 meters around each fever location, and the local density values ​​of the 10 fever locations were obtained. Then, normalization was performed to obtain the fever location aggregation index, with a maximum value of 0.85, indicating that the spatial aggregation degree of the local heating locations of the transmission line was high. Next, the fever degree data set was statistically analyzed, and the maximum value of the fever degree was calculated to be 95 degrees Celsius, the minimum value was 65 degrees Celsius, the average value was 80 degrees Celsius, the standard deviation was 10 degrees Celsius, and the median was 82 degrees Celsius, obtaining the basic statistical characteristics of the fever degree. At the same time, the fever degree data was analyzed by histogram to obtain the frequency distribution characteristics of the fever degree, among which the frequency of severe fever (greater than 90 degrees Celsius) was 20%, the frequency of moderate fever (70-90 degrees Celsius) was 60%, and the frequency of mild fever (60-70 degrees Celsius) was 20%.

[0128] Then, the weight coefficients of the statistical characteristics of the heat location concentration and heat degree were set to 0.3 and 0.7 respectively, and they were weighted and summed to obtain the initial value of the comprehensive heat index of 82.5. According to the spatial structure of the transmission line and the distribution of power load, the spatial distribution influencing factors of different parts and sections are preset, among which the influencing factors of important parts (such as substations, junction boxes, etc.) are 1.0, the influencing factors of general parts are 0.8, and the influencing factors of no-load sections are 0.6. For each heat location, multiply it by the corresponding spatial distribution influencing factor to obtain the correction coefficient, and then correct the initial value of the comprehensive heat index to obtain a final comprehensive heat index of 78.2, indicating that the overall heat risk of the transmission line is high.

[0129] According to the comprehensive fever index, a multi-level fever warning level judgment condition is constructed, and the comprehensive fever index is set between 0 and 30 as the normal level, between 30 and 60 as the warning level, and between 60 and 100 as the severe level. The comprehensive fever index of 78.2 is compared with the judgment condition, and the initial result of the local fever warning level of the transmission line is obtained as the severe level. The initial result of the warning level is analyzed by time series, and the exponential smoothing method is used. The smoothing coefficient is set to 0.6 to obtain the smoothed warning level sequence in the past hour. It is found that the warning level shows an overall upward trend, from the warning level to the severe level. Finally, the spatial correlation analysis of the smoothed warning level sequence is performed, and it is found that the warning level of important parts of the transmission line (such as substations) is generally higher than that of general parts, and the warning levels of adjacent positions are relatively consistent. Based on this, the local fever warning level is spatially corrected, and the final local fever warning level is obtained as the severe level.

[0130] In a specific embodiment, the process of executing step S106 may specifically include the following steps:

[0131] (1) According to the local heating warning level, the preset transmission line BIM model is color-coded to obtain the color-coded transmission line BIM model;

[0132] (2) Perform three-dimensional geometric transformation on the color-annotated transmission line BIM model to obtain a three-dimensional visual warning model;

[0133] (3) According to the three-dimensional visualization early warning model, the spatial boundary of the local fever area is extracted to obtain the boundary of the local fever area;

[0134] (4) Divide the boundary of the local heating area into grids to obtain the local heating area grid;

[0135] (5) Calculate the grid density of the warning level area based on the local heating area grid to obtain the grid density distribution of the warning level area;

[0136] (6) Analyze the inspection resource demand and time window of the grid density distribution in the warning level area to obtain the inspection constraint conditions;

[0137] (7) Generate an initial inspection path based on the inspection constraints, and optimize the inspection coverage of the initial inspection path to obtain an optimized inspection path;

[0138] (8) Perform differentiated inspection strategy analysis on the optimized inspection path to obtain a differentiated inspection strategy set, and generate a differentiated inspection maintenance plan based on the differentiated inspection strategy set.

[0139] It should be noted that according to the local heating warning level, the preset transmission line BIM model is color-labeled to generate a color-labeled transmission line BIM model. The transmission line BIM model is a three-dimensional digital model of the transmission line, which contains the geometric information, topological structure, electrical parameters and other attribute information of the transmission line. The local heating warning level is mapped to different colors, such as red for severe warning, orange for warning, and green for normal state. Then, according to the coordinates of the heating position, the corresponding position in the transmission line BIM model is color-labeled to intuitively display the spatial distribution of local heating areas with different warning levels.

[0140] Next, the color-annotated transmission line BIM model is transformed into a three-dimensional geometric model to obtain a three-dimensional visual warning model. The three-dimensional geometric transformation includes operations such as rotation, translation, and scaling. The purpose is to adjust the transmission line BIM model to a suitable perspective and scale so that the spatial position and distribution characteristics of the local heating area can be observed more clearly. Through the three-dimensional geometric transformation, an intuitive and realistic three-dimensional visual warning model is generated, which intuitively displays the overall structure of the transmission line and the spatial relationship between the local heating area, providing an intuitive reference for subsequent inspection path planning and maintenance decisions.

[0141] After obtaining the three-dimensional visualization warning model, it is necessary to further extract the spatial boundaries of the local heating area to obtain the boundaries of the local heating area. Spatial boundary extraction can use three-dimensional voxel division, surface reconstruction and other algorithms to divide the three-dimensional visualization warning model into non-intersecting local heating areas and extract the spatial boundary surface of each area. Specifically, the three-dimensional visualization warning model is voxelized and discretized into regular cubic units. Then, according to the color labeling information, the voxels are regionally grown and merged to obtain a set of non-intersecting local heating areas. For each local heating area, a surface reconstruction algorithm, such as the Marching Cubes algorithm, is used to extract its spatial boundary surface to obtain a set of closed triangular mesh patches, which represent the spatial range and shape of the local heating area.

[0142] The extracted local heating area boundary is meshed and discretized into regular grid units to obtain the local heating area grid. The meshing can use spatial data structures such as quadtree and octree to recursively divide the local heating area boundary into grid units of different granularities to generate a multi-resolution grid model. The purpose of meshing is to facilitate the subsequent inspection path planning and maintenance task allocation. The local heating area is discretized into grid units that can be processed independently, and each grid unit represents a basic unit of inspection or maintenance. At the same time, the meshing can also consider the geometric characteristics and warning level distribution of the local heating area. A finer mesh is used in areas with higher warning levels, and a coarser mesh is used in areas with lower warning levels to achieve adaptive mesh generation.

[0143] According to the grid of the local heating area, the grid density of the areas with different warning levels is calculated to obtain the grid density distribution of the warning level areas. The grid density reflects the number and distribution characteristics of the grid units in the areas with different warning levels, and is an important basis for formulating differentiated inspection strategies. Specifically, the grid of the local heating area is statistically analyzed, the number of grid units in each warning level area is calculated, and then divided by the total area of ​​the area to obtain the grid density of the warning level area. The higher the grid density, the more serious the local heating problem in the warning level area, requiring more frequent and fine-grained inspections and maintenance. The grid density distribution intuitively shows the differences in inspection and maintenance requirements in areas with different warning levels, and provides quantitative indicators for subsequent inspection path optimization and maintenance strategy formulation.

[0144] After obtaining the grid density distribution of the warning level area, it is necessary to further analyze the inspection resource demand and time window limit to obtain the inspection constraint conditions. Inspection resource demand refers to the types and quantities of resources such as manpower, equipment, and transportation required to complete the inspection task of the local heating area. The time window limit refers to the time range and duration allowed for inspection in each local heating area. By analyzing the grid density distribution of the warning level area, combining historical inspection data and expert experience, the inspection resource demand and time window limit of different warning level areas are estimated to generate inspection constraint conditions. For example, for areas with higher warning levels, more inspection personnel and equipment are allocated, and more frequent and longer inspection time windows are arranged; for areas with lower warning levels, fewer inspection resources are allocated, and less frequent and shorter inspection time windows are arranged. Inspection constraint conditions provide a realistic basis and restriction conditions for subsequent inspection route planning.

[0145] According to the inspection constraints, the initial inspection path is generated, that is, a patrol route covering all local heating areas is planned for each inspector. The initial inspection path generation can adopt graph theory algorithms, heuristic search algorithms, etc., such as the shortest path algorithm, the traveling salesman problem solving algorithm, etc. Taking the shortest path algorithm as an example, the center point of the local heating area grid is used as the inspection node, and the edge between adjacent grids is used as the inspection path. A weighted directed graph model is constructed, and then the Dijkstra algorithm or A* algorithm is used to solve the shortest inspection path covering all inspection nodes as the initial inspection path. The initial inspection path meets the inspection constraints, but there may be problems such as repeated coverage and detours, which need to be further optimized.

[0146] The inspection coverage rate of the initial inspection path is optimized to improve the coverage efficiency and adequacy of the inspection path for the local heating area. Inspection coverage rate optimization can adopt heuristic optimization algorithms, such as genetic algorithm, ant colony algorithm, simulated annealing algorithm, etc. By defining appropriate optimization objective functions and search strategies, the initial inspection path is iteratively optimized to obtain an inspection path with higher coverage and better path. For example, a genetic algorithm is used to encode the initial inspection path into a gene sequence, and a set of candidate inspection paths are randomly generated as the initial population. Then, through genetic operations such as selection, crossover, and mutation, the inspection path is continuously iterated and optimized until the coverage threshold is met or the maximum number of iterations is reached, and the optimized inspection path is obtained. The optimized inspection path takes into account both coverage efficiency and path rationality, reduces repeated coverage and detours, and improves inspection efficiency and quality.

[0147] Finally, the optimized inspection paths are analyzed for differentiated inspection strategies. Different inspection methods, inspection contents, inspection frequencies, etc. are formulated according to the characteristics and needs of different warning level areas to obtain differentiated inspection strategy sets. Differentiated inspection strategies refer to the use of different inspection methods and standards for different warning level areas to achieve refined and personalized inspection management. For example, for key areas with higher warning levels, a combination of manual inspection and drone inspection is used, with inspections conducted twice a week, focusing on the temperature change trend and equipment operation status of the heating location; for general areas with lower warning levels, manual inspections are mainly used, with inspections conducted once a month, focusing on the appearance of the equipment and the line channel environment. The differentiated inspection strategy set provides targeted inspection plans for each warning level area, improving the effectiveness and refinement of inspections.

[0148] Based on the differentiated inspection strategy set and the optimized inspection path, a differentiated inspection and maintenance plan is generated to clarify the inspection tasks, inspection routes, inspection personnel, etc. within each inspection time window, forming a complete and executable inspection and maintenance plan. The differentiated inspection and maintenance plan comprehensively considers factors such as the local fever warning level, inspection resource restrictions, and inspection time windows, reasonably arranges inspection tasks and inspection paths, maximizes the use of limited inspection resources, and improves the inspection and maintenance efficiency and reliability of transmission lines.

[0149] For example, for local heating monitoring of a certain transmission line, through the previous analysis, a set of local heating warning levels were obtained, including 2 severe warning areas, 3 warning areas, and 5 normal areas. According to the warning level, the transmission line BIM model is color-coded, with 2 severe warning areas marked as red, 3 warning areas marked as orange, and 5 normal areas marked as green, generating a color-coded transmission line BIM model. Then, the color-coded BIM model is rotated and scaled in three dimensions to adjust to the appropriate viewing angle and scale to generate a three-dimensional visual warning model.

[0150] Next, the three-dimensional visual warning model was voxelized to extract the spatial boundaries of 2 severe warning areas, 3 warning areas, and 5 normal areas, and 10 local fever area boundaries were obtained. Each local fever area boundary was divided into octrees with a recursive depth of 3 to obtain a local fever area grid, with each grid unit having a size of 5 m × 5 m × 5 m. According to the local fever area grid, the number of grid units in each warning level area was counted, and the grid density was calculated to obtain the grid density distribution of the warning level area, where the grid density of the severe warning area was 8 / 100 square meters, the grid density of the warning area was 6 / 100 square meters, and the grid density of the normal area was 2 / 100 square meters.

[0151] According to the grid density distribution of the warning level area, the inspection resource demand and time window limit are analyzed, and the inspection constraint conditions are formulated. For the severe warning area, two inspection personnel are arranged, each equipped with an infrared thermometer, and the inspection is carried out twice a week, and each inspection lasts no less than 2 hours; for the alert warning area, one inspection personnel is arranged, each equipped with an infrared thermometer, and the inspection is carried out once a week, and each inspection lasts no less than 1 hour; for the normal area, one inspection personnel is arranged, and the inspection is carried out once a month, and each inspection lasts no less than 30 minutes.

[0152] According to the inspection constraints, the shortest path algorithm was used to generate the initial inspection path, and 3 inspection paths were obtained, covering 2 severe warning areas, 3 warning areas and 5 normal areas respectively. Then, the genetic algorithm was used to optimize the initial inspection path, with a population size of 50, an evolutionary generation of 100, a crossover probability of 0.8, and a mutation probability of 0.1. The optimized inspection path was obtained, and the inspection coverage rate was increased by 15%, and the total path length was shortened by 10%.

[0153] Finally, the optimized inspection paths are analyzed for differentiated inspection strategies, and differentiated inspection strategy sets are formulated. For severe warning areas, a combination of manual inspection and drone inspection is used to focus on checking the temperature change trend and equipment operation status of the heating location, record abnormal conditions, and report them in a timely manner; for warning areas, manual inspection is mainly used, focusing on checking the appearance of the equipment and the line channel environment, recording abnormal conditions, and regularly summarizing and analyzing; for normal areas, manual inspection is mainly used, focusing on checking channel hygiene and safety hazards, recording inspection results, and regularly summarizing and evaluating.

[0154] Based on the differentiated inspection strategy set and the optimized inspection path, a differentiated inspection and maintenance plan for the next month is generated. The plan includes: every Monday and Thursday from 9:00 to 11:00 am, two inspectors will inspect the two severe warning areas along the optimized inspection paths 1 and 2 respectively, focusing on checking the temperature change trend and equipment operation status of the heating position, recording abnormal conditions, and reporting them in a timely manner; every Wednesday from 2:00 to 4:00 pm, one inspector will inspect the three warning areas along the optimized inspection path 3, focusing on checking the equipment appearance and line channel environment, recording abnormal conditions, and regularly summarizing and analyzing; every first Friday of each month from 10:00 to 12:00 am, one inspector will inspect the five normal areas along the optimized inspection path 4, focusing on checking channel hygiene and safety hazards, recording inspection results, and regularly summarizing and evaluating.

[0155] The embodiment of the present invention further provides a transmission line local heating monitoring system, and the transmission line local heating monitoring system specifically includes:

[0156] The acquisition module 201 is used to acquire the temperature data and infrared thermal image of the power transmission line, perform denoising on the temperature data to obtain denoised temperature data, and perform image enhancement on the infrared thermal image to obtain enhanced thermal image data;

[0157] An interpolation module 202 is used to perform interpolation processing on the noise-reduced temperature data and the enhanced thermal image data to obtain a target temperature matrix and a target thermal image matrix;

[0158] A fusion module 203 is used to fuse the temperature matrix and the thermal image matrix to obtain a fused temperature distribution matrix;

[0159] The segmentation module 204 is used to perform threshold segmentation on the temperature distribution matrix to obtain a local high temperature area, and to locate and quantify the local high temperature area to obtain a heating position coordinate set and a heating degree data set;

[0160] A calculation module 205, configured to calculate a comprehensive fever index according to the fever position coordinate set and the fever degree data set, and determine a local fever warning level according to the comprehensive fever index;

[0161] The matching module 206 is used to match the local heating warning level with the preset transmission line BIM model to generate a visual warning model, and perform a differentiated inspection scheme analysis on the visual warning model according to the local heating warning level to generate a differentiated inspection and maintenance plan.

[0162] Through the collaborative work of the above modules, the temperature data and infrared thermal images of the transmission line are obtained, the temperature data is denoised to obtain the denoised temperature data, and the infrared thermal images are enhanced to obtain the enhanced thermal image data, which effectively improves the quality and reliability of the temperature data and thermal image data. The target temperature matrix and the target thermal image matrix are obtained by interpolation processing of the denoised temperature data and the enhanced thermal image data, which realizes the standardization and size unification of the temperature and thermal image data, and lays the foundation for subsequent data fusion. The target temperature matrix and the target thermal image matrix are fused to obtain the fused temperature distribution matrix, which makes full use of the complementary characteristics of temperature and thermal image data and obtains more comprehensive and accurate temperature distribution information. The fused temperature distribution matrix is ​​threshold segmented to accurately locate the local high temperature area, and the heating position coordinate set and the heating degree data set are obtained through positioning and quantification processing, which realizes the automatic detection and quantitative evaluation of the local high temperature area. According to the heating position coordinate set and the heating degree data set, the comprehensive heating index is calculated, and the local heating warning level is determined, which realizes the comprehensive evaluation and graded warning of the local heating risk. The local heating warning level is matched with the preset transmission line BIM model to generate an intuitive visual warning model, and the visual warning model is analyzed for a differentiated inspection plan according to the warning level to generate a targeted differentiated inspection and maintenance plan, which greatly improves the early warning and maintenance efficiency of local heating failures in the transmission line. The present invention utilizes the multi-source heterogeneous characteristics of temperature data and infrared thermal images, and obtains a high-quality fused temperature distribution matrix through a series of processing such as data denoising, image enhancement, interpolation processing, and data fusion. It further realizes the automatic detection, positioning, evaluation, and early warning of local high-temperature areas through threshold segmentation, positioning quantification, comprehensive index calculation, and early warning classification. Finally, combined with the transmission line BIM model, a visual warning model and a differentiated inspection and maintenance plan are generated, which greatly improves the intelligence level and practicality of local heating monitoring of transmission lines.

[0163] The above embodiments are only used to illustrate the technical solutions of the present invention rather than to limit the same. Although the present invention has been described in detail with reference to the embodiments, a person skilled in the art should understand that the specific implementation modes of the present invention can still be modified or replaced by equivalents, and any modification or equivalent replacement that does not depart from the spirit and scope of the present invention should be included in the scope of the claims of the present invention.

Claims

1. A method for monitoring local heating of a power transmission line, characterized in that: include: Acquire temperature data and infrared thermal images of the power transmission line, perform denoising on the temperature data to obtain denoised temperature data, and perform image enhancement on the infrared thermal images to obtain enhanced thermal image data; Performing interpolation processing on the noise-reduced temperature data and the enhanced thermal image data to obtain a target temperature matrix and a target thermal image matrix; fusing the temperature matrix and the thermal image matrix to obtain a fused temperature distribution matrix; Performing threshold segmentation on the temperature distribution matrix to obtain a local high temperature area, and performing positioning and quantification processing on the local high temperature area to obtain a fever position coordinate set and a fever degree data set; According to the fever position coordinate set and the fever degree data set, a comprehensive fever index is calculated, and a local fever warning level is determined according to the comprehensive fever index. The step of calculating the comprehensive fever index according to the fever position coordinate set and the fever degree data set, and determining the local fever warning level according to the comprehensive fever index comprises: calculating the spatial aggregation of the fever position according to the fever position coordinate set to obtain the fever position aggregation; performing a fever degree statistical feature analysis on the fever degree data set to obtain a fever degree statistical feature; A weighted sum is performed on the fever position aggregation degree and the fever degree statistical characteristics to obtain an initial value of a comprehensive fever index; based on a preset spatial distribution influencing factor, the initial value of the comprehensive fever index is corrected to obtain a comprehensive fever index; a fever warning level determination condition is constructed according to the comprehensive fever index to obtain a multi-level fever warning level determination condition; a logical comparison is performed between the comprehensive fever index and the multi-level fever warning level determination condition to obtain an initial result of a local fever warning level; a time series analysis is performed on the initial result of the local fever warning level to obtain a time-smoothed local fever warning level; a spatial correlation analysis is performed on the time-smoothed local fever warning level to obtain the local fever warning level; The spatial distribution impact factor is used to reflect the importance and influence of the heating position in different parts and sections of the transmission line. According to the spatial structure and power load distribution of the transmission line, the spatial distribution impact factors of different parts and sections are preset, and the value range is 0~1. The larger the value, the higher the importance of the corresponding part or section, and the greater the impact of the heating risk. The local heating warning level is matched with a preset transmission line BIM model to generate a visual warning model, and a differentiated inspection scheme analysis is performed on the visual warning model according to the local heating warning level to generate a differentiated inspection and maintenance plan.

2. The method for monitoring local heating of a power transmission line according to claim 1, characterized in that: The steps of acquiring temperature data and infrared thermal images of the power transmission line, performing denoising on the temperature data to obtain denoised temperature data, and performing image enhancement on the infrared thermal images to obtain enhanced thermal image data include: Performing a wavelet transform on the temperature data to obtain high-frequency coefficients corresponding to the temperature data, and performing an inverse wavelet transform on the temperature data according to the high-frequency coefficients to obtain the denoised temperature data; Performing grayscale stretching processing on the infrared thermal image to obtain a stretched infrared thermal image; performing histogram equalization processing on the stretched infrared thermal image to obtain a balanced infrared thermal image; Performing median filtering on the balanced infrared thermal image to obtain a filtered infrared thermal image; removing illumination components from the filtered infrared thermal image to obtain a candidate infrared thermal image; Performing pixel gradient data analysis on the candidate infrared thermal image to obtain a gradient amplitude map and a gradient direction map; Performing non-maximum suppression on the gradient amplitude map to obtain a candidate gradient amplitude map, and performing hysteresis threshold processing on the candidate gradient amplitude map to obtain a target edge map; The candidate infrared thermal image is subjected to edge sharpening processing based on the gradient direction map and the target edge map to obtain the enhanced thermal image data.

3. The method for monitoring local heating of a power transmission line according to claim 1, characterized in that: The step of interpolating the noise-reduced temperature data and the enhanced thermal image data to obtain a target temperature matrix and a target thermal image matrix includes: Acquire the size of the noise-reduced temperature data and the size of the enhanced thermal image data, and perform target interpolation size analysis according to the size of the noise-reduced temperature data and the size of the enhanced thermal image data to obtain a temperature interpolation size and a thermal image interpolation size; According to the temperature interpolation size, performing two-dimensional cubic spline interpolation on the noise-reduced temperature data to obtain interpolated temperature data; According to the thermal image interpolation size, the enhanced thermal image data is subjected to gradient interpolation processing to obtain interpolated thermal image data; Normalizing the interpolated temperature data to obtain normalized temperature data, and normalizing the interpolated thermal image data to obtain normalized thermal image data; According to the normalized temperature data, bilinear interpolation is performed on the noise-reduced temperature data to obtain a candidate temperature matrix; According to the normalized thermal image data, bicubic interpolation is performed on the enhanced thermal image data to obtain a candidate thermal image matrix; The candidate temperature matrix is ​​smoothed to obtain the target temperature matrix, and the candidate thermal image matrix is ​​Gaussian filtered to obtain the target thermal image matrix.

4. The method for monitoring local heating of a power transmission line according to claim 1, characterized in that: The step of fusing the temperature matrix and the thermal image matrix to obtain a fused temperature distribution matrix includes: Performing multi-scale decomposition on the temperature matrix and the thermal image matrix respectively to obtain a multi-scale temperature feature matrix and a multi-scale thermal image feature matrix; Performing joint sparse coding on the multi-scale temperature feature matrix and the multi-scale thermal imaging feature matrix to obtain a multi-scale temperature-thermal imaging feature matrix; Performing local linear embedding on the multi-scale temperature-thermal image feature matrix to obtain a low-dimensional temperature-thermal image feature matrix; Performing deep feature extraction on the low-dimensional temperature-thermal image feature matrix to obtain a deep temperature-thermal image fusion feature matrix; Performing adaptive region selection on the deep temperature-thermal image fusion feature matrix to obtain a candidate fusion region matrix; Performing superpixel segmentation on the candidate fusion region matrix to obtain a plurality of segmentation region matrices; A global optimization process is performed on the plurality of segmented region matrices to obtain the fusion temperature distribution matrix.

5. The method for monitoring local heating of a power transmission line according to claim 1, characterized in that: The step of performing threshold segmentation on the temperature distribution matrix to obtain a local high temperature area, and performing positioning and quantification processing on the local high temperature area to obtain a heating position coordinate set and a heating degree data set includes: Binarizing the temperature distribution matrix to obtain a binary temperature distribution matrix; Performing a morphological opening operation on the binary temperature distribution matrix to obtain an open binary temperature distribution matrix; Performing connected region extraction on the opening binary temperature distribution matrix to obtain a high-temperature connected region matrix; Performing region area and region circumscribed rectangle analysis on the high temperature connected region matrix to obtain a candidate high temperature region matrix set; Extracting the centroid coordinates of each high temperature region according to the candidate high temperature region matrix set to obtain an initial heating position coordinate set; Performing spatial clustering on the initial fever location coordinate set, merging adjacent fever location coordinates, and obtaining a fever location coordinate set; According to the clustered heating position coordinate set, the temperature value of each heating position is extracted in the temperature distribution matrix to obtain a heating position temperature value set; The temperature value set of the heating position is analyzed to obtain the average temperature and the maximum temperature; based on the average temperature and the maximum temperature, the heating degree of each heating position is analyzed to obtain a heating degree data set.

6. The method for monitoring local heating of a power transmission line according to claim 1, characterized in that: The steps of matching the local heating warning level with a preset transmission line BIM model to generate a visual warning model, and performing a differentiated inspection scheme analysis on the visual warning model according to the local heating warning level to generate a differentiated inspection and maintenance plan include: According to the local heating warning level, color-labeling the preset transmission line BIM model to obtain the color-labeled transmission line BIM model; Performing a three-dimensional geometric transformation on the color-annotated transmission line BIM model to obtain a three-dimensional visual early warning model; According to the three-dimensional visualization early warning model, extract the spatial boundary of the local heating area to obtain the boundary of the local heating area; Dividing the boundary of the local heating area into a grid to obtain a local heating area grid; calculating the grid density of the warning level area according to the local heating area grid to obtain the grid density distribution of the warning level area; Perform inspection resource demand and time window analysis on the grid density distribution in the warning level area to obtain inspection constraint conditions; According to the inspection constraint conditions, an initial inspection path is generated, and the inspection coverage rate of the initial inspection path is optimized to obtain an optimized inspection path; A differentiated inspection strategy analysis is performed on the optimized inspection path to obtain a differentiated inspection strategy set, and the differentiated inspection maintenance plan is generated according to the differentiated inspection strategy set.

7. A transmission line local heating monitoring system, used to execute the transmission line local heating monitoring method according to any one of claims 1 to 6, characterized in that: include: An acquisition module is used to acquire temperature data and infrared thermal images of the power transmission line, perform denoising on the temperature data to obtain denoised temperature data, and perform image enhancement on the infrared thermal images to obtain enhanced thermal image data; An interpolation module, used for performing interpolation processing on the noise-reduced temperature data and the enhanced thermal image data to obtain a target temperature matrix and a target thermal image matrix; A fusion module, used for fusing the temperature matrix and the thermal image matrix to obtain a fused temperature distribution matrix; A segmentation module is used to perform threshold segmentation on the temperature distribution matrix to obtain a local high temperature area, and to locate and quantify the local high temperature area to obtain a fever position coordinate set and a fever degree data set; a calculation module, for calculating a comprehensive fever index according to the fever position coordinate set and the fever degree data set, and determining a local fever warning level according to the comprehensive fever index, wherein the steps of calculating the comprehensive fever index according to the fever position coordinate set and the fever degree data set, and determining the local fever warning level according to the comprehensive fever index, include: calculating the spatial concentration of the fever position according to the fever position coordinate set to obtain the fever position concentration; performing a fever degree statistical feature analysis on the fever degree data set to obtain a fever degree statistical feature; performing a weighted summation on the fever position concentration and the fever degree statistical feature to obtain an initial value of the comprehensive fever index; modifying the initial value of the comprehensive fever index based on a preset spatial distribution influencing factor to obtain a comprehensive fever index; constructing a fever warning level determination condition according to the comprehensive fever index to obtain a multi-level fever warning level determination condition; performing a logical comparison between the comprehensive fever index and the multi-level fever warning level determination condition to obtain an initial result of the local fever warning level; performing a time series analysis on the initial result of the local fever warning level to obtain a time-smoothed local fever warning level; performing a spatial correlation analysis on the time-smoothed local fever warning level to obtain the local fever warning level; The spatial distribution impact factor is used to reflect the importance and influence of the heating position in different parts and sections of the transmission line. According to the spatial structure and power load distribution of the transmission line, the spatial distribution impact factors of different parts and sections are preset, and the value range is 0~1. The larger the value, the higher the importance of the corresponding part or section, and the greater the impact of the heating risk. A matching module is used to match the local heating warning level with a preset transmission line BIM model to generate a visual warning model, and to perform a differentiated inspection scheme analysis on the visual warning model according to the local heating warning level to generate a differentiated inspection and maintenance plan.

Citation Information

Patent Citations

  • Ultraviolet, infrared and visible imaging-based power inspection online monitoring system

    CN110672980A

  • Autonomous inspection and hot spot identification method and system based on photovoltaic power station UAV

    CN111931565A

  • Photovoltaic power station online operation and maintenance method and system based on infrared and visible light images

    CN112465738A

  • Early warning method and device for early fire of cable channel, early warning monitoring platform and storage medium

    CN113689651A

  • Power grid monitoring method based on multi-type image recognition fusion

    CN117830870A

Cited By

  • Visual power transmission line multi-dimensional risk detection system and method

    CN120373849A

  • A Visualized Multi-Dimensional Risk Detection System and Method for Transmission Lines

    CN120373849B