Power transmission line temperature detection method, system and device based on data model

Abstract

This invention provides a method for detecting transmission line temperature based on data model prediction, comprising: acquiring transmission line images, including infrared and visible light images; classifying and storing the images according to geographical information acquired at the time of acquisition; aligning the transmission line images; removing background interference using visible light images; extracting features from the infrared images and storing them as a hierarchical three-dimensional database of temperature gradients and geographical information; constructing a model based on the isolated forest algorithm using the hierarchical three-dimensional database; making predictions using the constructed model; and performing residual analysis on the predicted and actual values ​​to predict fault points and trend changes. By constructing a temperature data information model using temperature data, and using isolated forest outlier detection and gradient change patterns to determine the temperature changes of each tower, the method reflects the fault change patterns of the towers. Through data prediction and analysis, it enables early prediction and warning of potential hazards, rapid screening and prediction, and improves implementation efficiency and practical application significance.

Description

Technical Field

[0001] This invention relates to the application of deep learning technology in the field of power transmission technology, and in particular to a method for detecting the temperature of power transmission lines based on data model prediction. Background Technology

[0002] Image recognition technology is a crucial component of deep learning, aiming to enable computers to process large amounts of physical information, replacing manual labor. The image recognition process involves information acquisition, preprocessing, feature extraction and selection, classifier design, and classification decision-making. The safe operation of the power grid is the cornerstone of social life and economic operation. Due to the long distances, large spans, and high detection points of transmission and transformation lines, conventional detection methods are insufficient. Currently, the industry commonly uses drones to capture images of the lines, combined with infrared thermal imaging technology. After data collection, the visible light image is aligned with the infrared image to locate high-temperature areas in the thermal image. Then, the visible light image is aligned to achieve recognition, combined with manual review. In the field of high-voltage transmission lines, because the tower height is over 50 meters, there is less background interference, making target identification relatively easy. However, for low-lying substations, the background is complex, with vehicles on the road, buildings in the background of the poles, and trees causing interference. The complex background of the acquired images makes it difficult to guarantee good recognition accuracy when using traditional image detection methods to detect the temperature of transmission lines. This requires image segmentation, target recognition, and template matching detection.

[0003] Due to limitations in scenarios, practical application conditions, and costs, the commonly used detection method is segmentation and target recognition followed by temperature detection. However, when the background area is complex, the detection of the power line tower area is easily affected by background interference. Background interference filtering is a complex process that requires a large amount of manual annotation, which is time-consuming and costly, becoming a bottleneck restricting power line inspection. Summary of the Invention

[0004] This invention aims to overcome at least one of the defects of the prior art. Addressing the problems in infrared thermal imaging detection of transmission line towers, it proposes to establish a temperature data information model using temperature data, use isolated forest outlier detection and gradient change law to judge the temperature change of each tower, reflect the fault change law of the tower, and through data prediction analysis, can predict and warn of potential hazards in advance, quickly screen and predict, improve implementation efficiency and practical application significance.

[0005] In a first aspect, the present invention provides a method for detecting the temperature of transmission lines based on a data model, comprising:

[0006] Acquire images of power transmission lines, including infrared and visible light images;

[0007] The images of the power transmission lines are classified and stored according to the geographic information acquired at the time of image acquisition.

[0008] Align the images of power transmission lines;

[0009] Using visible light images to remove background effects;

[0010] Features are extracted from infrared images and stored as a hierarchical three-dimensional database containing temperature gradients and geographic information;

[0011] A model is constructed using a hierarchical three-dimensional database based on the isolated forest algorithm;

[0012] Make predictions using the constructed model;

[0013] Residual analysis is performed on the predicted and actual values ​​to predict fault points and trend changes.

[0014] The transmission line images of this invention can be captured by means of drones, and the collected images can be classified and stored according to geographical information. Images of the same location are stored in the same database table, and can be classified and stored according to azimuth features.

[0015] Infrared and visible light images need to be aligned to ensure a one-to-one correspondence, which is beneficial for subsequent data filtering. This ensures that the geometric center of the line is located in the central region of the image, allowing for comparison and filtering based on this feature.

[0016] Since uncertain objects may appear in the background of the tower where the line is located, these randomly acquired targets will interfere with the segmentation of the temperature isotherm region. Therefore, a filtering mechanism strategy will be given for targets that accidentally enter the region based on historical files, adding permanent regions and filtering out accidental temporary regions.

[0017] During the background screening process, since background issues frequently occur on power lines and towers, most background influences can be eliminated through comparison with historical samples. For local samples after line updates and replacements, they can be supplemented by updating files. For occasional and continuous background interference, the patterns of their temporal changes can be summarized, their characteristics extracted, and then used for screening.

[0018] The hierarchical three-dimensional database of this invention is based on the logical relationship between feature training and storage of line towers. After the image is converted into structured data, dynamic features are filtered through historical comparison, fixed features are retained, and key storage and management areas can be determined through time series storage regression. Redundant parts can be processed by pruning.

[0019] The data structure of the hierarchical three-dimensional database is established according to the temperature gradient, and a three-level temperature information database can be dynamically established based on temperature.

[0020] Furthermore, the feature extraction of the infrared image specifically includes:

[0021] The infrared image is corrected and registered, and the temperature value and pixel value are calibrated.

[0022] The temperature information of the infrared image is mapped and converted using the RGB channel pixel values.

[0023] By calibrating the temperature value and pixel value of the reference temperature infrared image, correction and registration can be performed to establish a mathematical relationship between pixel value and temperature value. Then, after the infrared image is classified according to geographical information, the temperature information of the infrared image can be mapped and converted through the RGB three-channel pixel values, and the temperature information of the infrared image can be stored according to the pixel values.

[0024] Furthermore, the hierarchical three-dimensional database storing temperature gradients and geographic information specifically includes:

[0025] By comparing historical data in the database, dynamically occurring features are filtered out, while fixed features are retained.

[0026] Key storage management areas are identified through time-series storage regression.

[0027] Pruning removes redundant parts;

[0028] Layered storage is implemented based on temperature gradients and geographic information.

[0029] Furthermore, the hierarchical storage includes:

[0030] Based on the ambient temperature of the transmission line images, three temperature information databases were established: the normal temperature range, the higher temperature range, and the dangerous range.

[0031] The temperature information databases are all constructed with a two-dimensional data storage relationship based on the storage time and the corresponding geographical information.

[0032] Furthermore, the specific steps of constructing a model based on the isolated forest algorithm using a hierarchical three-dimensional database include:

[0033] Pixel positions and pixel values ​​are extracted from the infrared images based on the three temperature information databases.

[0034] Models were constructed based on the Isolation Forest algorithm on three temperature information databases.

[0035] Furthermore, the prediction using the constructed model includes the following steps:

[0036] a. Randomly select Ψ temperature point samples from the hierarchical 3D database as subsamples and place them in the root node;

[0037] b. Randomly specify a dimension and randomly generate a cut point p in the current node. The cut point p is generated between the maximum and minimum values ​​of the specified dimension in the current node.

[0038] c. Generate a hyperplane with the cutting point p, divide the data space of the current node into two subspaces, and generate nodes for data smaller than p in the specified dimension to the left of the current node, and generate nodes for data larger than p to the right of the current node;

[0039] d. Recursively repeat steps b and c in each child node until there is only one piece of data in the child node.

[0040] The isolated forest algorithm identifies outliers by sparsely distributed points that are far from denser populations, which is time-complex. However, by combining transmission line images with the encouraged forest algorithm, outliers can be effectively identified. This allows for efficient and accurate detection and prediction of transmission line temperatures, reflecting the fault patterns of towers. Through data prediction and analysis, potential hazards can be predicted and alerted in advance, enabling rapid screening and assessment, thus improving implementation efficiency and practical application significance.

[0041] Furthermore, the alignment of the transmission line image specifically includes:

[0042] Alignment is performed by comparing the outer contours.

[0043] Eliminate distractors;

[0044] Filter out power line images that cannot be used for detection.

[0045] Secondly, the present invention provides a transmission line temperature detection system based on a data model, comprising:

[0046] The image acquisition module is used to acquire images of power transmission lines, including infrared images and visible light images;

[0047] The geographic information classification module is used to classify and store geographic information based on the geographic information acquired when the power transmission line images were obtained.

[0048] The image alignment module is used to align images of power transmission lines.

[0049] The removal module is used to remove background effects using visible light images;

[0050] A layered stereo database is used to extract features from infrared images and store them as a layered stereo database containing temperature gradients and geographic information.

[0051] The isolated forest model is constructed using a hierarchical three-dimensional database based on the isolated forest algorithm.

[0052] The prediction module is used to make predictions using the constructed model;

[0053] The residual analysis module is used to perform residual analysis on predicted and actual values ​​to predict fault points and trend changes.

[0054] Thirdly, the present invention also provides a computer device, including a memory and a processor, wherein the memory stores a computer program, characterized in that the processor executes the computer program to implement the data model-based transmission line temperature detection method.

[0055] Fourthly, the present invention also provides a computer-readable storage medium having a computer program stored thereon, characterized in that the computer program, when executed by a processor, implements the data model-based transmission line temperature detection method.

[0056] The data model-based transmission line temperature detection system, equipment, and storage medium of the present invention are based on the same inventive concept as the data model-based transmission line temperature detection method of the present invention, and will not be described again here.

[0057] Compared with the prior art, the beneficial effects of the present invention are as follows:

[0058] (1) Effectively eliminate influencing factors in the line image.

[0059] (2) Effectively identify abnormal values ​​and reflect the fault change pattern of the tower.

[0060] (3) Determine the temperature change of each tower to reflect the fault change pattern of the tower. Through data prediction and analysis, potential hazards can be predicted and alarmed in advance, and rapid screening and prediction can be carried out to improve implementation efficiency and practical application significance. Attached Figure Description

[0061] Figure 1 This is a flowchart illustrating the method for implementing the present invention.

[0062] Figure 2 This is a system composition framework diagram for implementing the present invention.

[0063] Figure 3 This is a system control flowchart for implementing the present invention.

[0064] Figure 4 This is a visible light diagram of the circuit in an embodiment of the present invention.

[0065] Figure 5 This is a schematic diagram of infrared thermal imaging pixel segmentation implemented in this invention. Detailed Implementation

[0066] The accompanying drawings are for illustrative purposes only and should not be construed as limiting the invention. To better illustrate the following embodiments, some parts in the drawings may be omitted, enlarged, or reduced, and do not represent the actual product dimensions; it is understandable to those skilled in the art that some well-known structures and their descriptions may be omitted in the drawings.

[0067] Example 1

[0068] like Figure 1 As shown, this embodiment provides a data model-based method for detecting the temperature of transmission lines, including:

[0069] S1. Acquire images of the transmission line, including infrared and visible light images;

[0070] S2. Classify and store the images of the power transmission lines according to the geographic information acquired at the time of image acquisition;

[0071] S3. Align the transmission line image;

[0072] S4. Use visible light images to remove background effects;

[0073] S5. Extract features from infrared images and store them as a hierarchical three-dimensional database of temperature gradients and geographic information;

[0074] S6. Construct a model based on the isolated forest algorithm using a hierarchical three-dimensional database;

[0075] S7. Make predictions using the constructed model;

[0076] S8. Perform residual analysis on the predicted and actual values ​​to predict fault points and trend changes.

[0077] like Figure 4 As shown, the visible light image obtained in this embodiment may have background interference problems, so it is necessary to remove background influencing factors. This is because uncertain objects may appear in the background of the tower where the line is located. These targets that are accidentally acquired will interfere with the segmentation of the temperature isotherm region. Therefore, a filtering mechanism strategy will be given for targets that accidentally enter the region based on historical files, adding permanent regions and filtering out accidental temporary regions.

[0078] Preferably, the feature extraction of the infrared image specifically includes:

[0079] The infrared image is corrected and registered, and the temperature value and pixel value are calibrated.

[0080] like Figure 5 As shown, the temperature information of the infrared image is mapped and converted through the RGB channel pixel values.

[0081] Preferably, the hierarchical three-dimensional database storing temperature gradients and geographic information specifically includes:

[0082] By comparing historical data in the database, dynamically occurring features are filtered out, while fixed features are retained.

[0083] Key storage management areas are identified through time-series storage regression.

[0084] Pruning removes redundant parts;

[0085] Layered storage is implemented based on temperature gradients and geographic information.

[0086] Preferably, the hierarchical storage includes:

[0087] Based on the ambient temperature of the transmission line images, three temperature information databases were established: the normal temperature range, the higher temperature range, and the dangerous range.

[0088] The temperature information databases are all constructed with a two-dimensional data storage relationship based on the storage time and the corresponding geographical information.

[0089] Preferably, the method of constructing a model based on the isolated forest algorithm using a hierarchical three-dimensional database specifically includes:

[0090] Pixel positions and pixel values ​​are extracted from the infrared images based on the three temperature information databases.

[0091] Models were constructed based on the Isolation Forest algorithm on three temperature information databases.

[0092] Preferably, the prediction using the constructed model includes the following steps:

[0093] a. Randomly select Ψ temperature point samples from the hierarchical 3D database as subsamples and place them in the root node;

[0094] b. Randomly specify a dimension and randomly generate a cut point p in the current node. The cut point p is generated between the maximum and minimum values ​​of the specified dimension in the current node.

[0095] c. Generate a hyperplane with the cutting point p, divide the data space of the current node into two subspaces, and generate nodes for data smaller than p in the specified dimension to the left of the current node, and generate nodes for data larger than p to the right of the current node;

[0096] d. Recursively repeat steps b and c in each child node until there is only one piece of data in the child node.

[0097] Preferably, the alignment of the transmission line image specifically includes:

[0098] Alignment is performed by comparing the outer contours.

[0099] Eliminate distractors;

[0100] Filter out power line images that cannot be used for detection.

[0101] The transmission line images in this embodiment can be captured by means of drones, etc. The collected images are classified and stored according to geographical information. Images of the same location are stored in the same database table, and can be classified and stored according to azimuth features.

[0102] This embodiment utilizes a drone to photograph the area around the power line tower, capturing images of the target tower and pole at standard locations. The drone is equipped with both visible light and infrared thermal imagers to acquire image information. After data aggregation, the data is processed using the method described in this patent, and a prediction result is output and sent to the operator for on-site maintenance. During data collection, the day's weather information is shared and incorporated into the data modeling to eliminate interference factors. Power consumption parameters for different time periods are also included in the data modeling task. An isolated forest algorithm is used to construct a line outlier prediction model to predict temperature trends in advance and provide early warnings of potential problems.

[0103] Infrared and visible light images need to be aligned to ensure a one-to-one correspondence, which is beneficial for subsequent data filtering. This ensures that the geometric center of the line is located in the central region of the image, allowing for comparison and filtering based on this feature.

[0104] During the background screening process, since background issues frequently occur on power lines and towers, most background influences can be eliminated through comparison with historical samples. For local samples after line updates and replacements, they can be supplemented by updating files. For occasional and continuous background interference, the patterns of their temporal changes can be summarized, their characteristics extracted, and then used for screening.

[0105] The hierarchical 3D database in this embodiment is based on the logical relationship between feature training and storage of line towers. After the image is converted into structured data, dynamic features are filtered through historical comparison, while fixed features are retained. Through time series storage regression, key storage and management areas can be determined, and redundant parts can be processed by pruning.

[0106] The data structure of the hierarchical three-dimensional database is established according to the temperature gradient, and a three-level temperature information database can be dynamically established based on temperature.

[0107] In this embodiment, abnormal temperature data will be pushed to the front end by the system for alarm processing. At the same time, over time, the temperature data of each point will accumulate and establish a temperature spectrum change gradient for that point. The gradient change pattern will also reflect whether the temperature at that point is normal and valid. When the gradient changes rapidly, a danger warning will be given.

[0108] Example 2

[0109] Based on the same concept as Embodiment 1, this embodiment provides a data model-based transmission line temperature detection system, including:

[0110] The image acquisition module is used to acquire images of power transmission lines, including infrared images and visible light images;

[0111] The geographic information classification module is used to classify and store geographic information based on the geographic information acquired when the power transmission line images were obtained.

[0112] The image alignment module is used to align images of power transmission lines.

[0113] The removal module is used to remove background effects using visible light images;

[0114] A layered stereo database is used to extract features from infrared images and store them as a layered stereo database containing temperature gradients and geographic information.

[0115] The isolated forest model is constructed using a hierarchical three-dimensional database based on the isolated forest algorithm.

[0116] The prediction module is used to make predictions using the constructed model;

[0117] The residual analysis module is used to perform residual analysis on predicted and actual values ​​to predict fault points and trend changes.

[0118] The data source for this embodiment is obtained by using a drone to photograph the area around the power line tower. The drone, equipped with both visible light and infrared thermal imagers, captures images of the target tower at standard locations. After acquiring the image information, the data is aggregated and processed using the method described in this patent. The resulting predictions are then sent to the operator for on-site maintenance. During data collection, the day's weather information is shared and incorporated into the data modeling to eliminate interference. Power consumption parameters for different time periods are also included in the data modeling task. An isolated forest algorithm is used to construct a line outlier prediction model, predicting temperature trends in advance and providing early warnings of potential problems.

[0119] like Figure 2As shown, the image acquisition module may include a flight control camera device, which performs image data acquisition and infrared image thermal detection.

[0120] The image alignment module aligns the data. Infrared and visible light images need to be aligned to ensure a one-to-one correspondence, which is beneficial for subsequent data filtering. This ensures that the geometric center of the line is located in the central region of the image, allowing for comparison and filtering based on this feature.

[0121] Since visible light images may have background interference issues, it is necessary to remove background influencing factors. This is because uncertain objects may appear in the background of the tower where the line is located. These accidentally acquired targets will interfere with the segmentation of the temperature isotherm region. Therefore, based on historical files, a filtering mechanism strategy will be provided for targets that accidentally enter the region, adding permanent regions and filtering out accidental temporary regions.

[0122] The layered three-dimensional database can be established based on the temperature sensing database, and features of infrared images are extracted and stored as a layered three-dimensional database of temperature gradients and geographic information.

[0123] Based on the temperature sensing database, the data model-based transmission line temperature detection system also includes an image processing unit, a data processing unit, a data storage module, and a risk early warning module.

[0124] like Figure 3 The system control flowchart shown first activates the transmission line early warning system, acquires and stores image data, and uses a drone to photograph the area around the tower at standard locations. The drone, equipped with both visible light and infrared thermal imagers, captures images of the target tower. After acquiring the image information, the data is aggregated and processed using the method described in this patent. The resulting predictions are then sent to the operator for on-site maintenance. During data acquisition, the day's weather information is shared and incorporated into the data modeling to eliminate interference. Power consumption parameters for different time periods are also included in the data modeling task. An isolated forest algorithm is used to construct a line anomaly prediction model to predict temperature trends and provide early warnings of potential problems.

[0125] Image data needs to be aligned. Infrared images and visible light images need to be aligned to ensure a one-to-one correspondence, which is beneficial for subsequent data screening. This ensures that the geometric center of the line is located in the central area of ​​the image, and comparison and screening can be performed based on this feature.

[0126] Infrared image data is processed. After the infrared images are classified according to their positioning attributes, the temperature information of the thermal imaging is mapped and converted through the RGB three-channel pixel values. The temperature information of the infrared thermal image is stored according to the pixel values. Before the mapping conversion, calibration and registration must be completed. The temperature value and pixel value of the reference temperature infrared image are calibrated. After calibration, image mapping is carried out, thus establishing the mathematical relationship between pixel values ​​and temperature values.

[0127] Aggregate and model the data.

[0128] Making predictions using models includes the following steps:

[0129] a. Randomly select Ψ temperature point samples from the hierarchical 3D database as subsamples and place them in the root node;

[0130] b. Randomly specify a dimension and randomly generate a cut point p in the current node. The cut point p is generated between the maximum and minimum values ​​of the specified dimension in the current node.

[0131] c. Generate a hyperplane with the cutting point p, divide the data space of the current node into two subspaces, and generate nodes for data smaller than p in the specified dimension to the left of the current node, and generate nodes for data larger than p to the right of the current node;

[0132] d. Recursively repeat steps b and c in each child node until there is only one piece of data in the child node.

[0133] Finally, the results are pushed out. Temperature anomaly data will be pushed to the front end by the system for alarm processing. At the same time, over time, the temperature data of each point will accumulate and establish a temperature spectrum change gradient for that point. The gradient change pattern will also reflect whether the temperature at that point is normal and valid. When the gradient changes rapidly, a danger warning will be given.

[0134] Obviously, the above embodiments of the present invention are merely examples for clearly illustrating the technical solution of the present invention, and are not intended to limit the specific implementation of the present invention. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of the claims of the present invention should be included within the protection scope of the claims of the present invention.

Claims

1. A method for detecting the temperature of transmission lines based on a data model, characterized in that, include: Acquire images of power transmission lines, including infrared and visible light images; The images of the power transmission lines are classified and stored according to the geographic information acquired at the time of image acquisition. Align the images of power transmission lines; Using visible light images to remove background effects; Features are extracted from infrared images and stored as a hierarchical three-dimensional database of temperature gradients and geographic information. The hierarchical three-dimensional database of temperature gradients and geographic information specifically includes: filtering out dynamically appearing features and retaining fixed features through historical comparison of the database; determining key storage and management areas through time series storage regression; and pruning redundant parts. Hierarchical storage is based on temperature gradients and geographic information; The hierarchical storage includes: establishing three temperature information databases based on the ambient temperature of the transmission line images, namely, a normal temperature range, a higher temperature range, and a dangerous range; each temperature information database constructs a two-dimensional data storage relationship according to the storage time and the corresponding geographical information of the line. The model is constructed using a hierarchical three-dimensional database based on the isolated forest algorithm, specifically including: extracting pixel positions and pixel values ​​from infrared images based on the three temperature information databases respectively; and constructing models based on the isolated forest algorithm based on the three temperature information databases respectively. Make predictions using the constructed model; Residual analysis is performed on the predicted and actual values ​​to predict fault points and trend changes.

2. The method for detecting transmission line temperature based on a data model according to claim 1, characterized in that, The feature extraction from the infrared image specifically includes: The infrared image is corrected and registered, and the temperature value and pixel value are calibrated. The temperature information of the infrared image is mapped and converted using the RGB channel pixel values.

3. The method for detecting transmission line temperature based on a data model according to claim 1, characterized in that, The prediction using the constructed model includes the following steps: a. Randomly select Ψ temperature point samples from the hierarchical 3D database as subsamples and place them in the root node; b. Randomly specify a dimension and randomly generate a cut point p in the current node. The cut point p is generated between the maximum and minimum values ​​of the specified dimension in the current node. c. Generate a hyperplane with the cutting point p, divide the data space of the current node into two subspaces, and generate nodes for data smaller than p in the specified dimension to the left of the current node, and generate nodes for data larger than p to the right of the current node; d. Recursively repeat steps b and c in each child node until there is only one piece of data in the child node.

4. The method for detecting transmission line temperature based on a data model according to claim 1, characterized in that, The alignment of the transmission line image specifically includes: Alignment is performed by comparing the outer contours. Eliminate distractors; Filter out power line images that cannot be used for detection.

5. A data model-based temperature detection system for transmission lines, characterized in that, include: The image acquisition module is used to acquire images of power transmission lines, including infrared images and visible light images; The geographic information classification module is used to classify and store geographic information based on the geographic information acquired when the power transmission line images were obtained. The image alignment module is used to align images of power transmission lines. The removal module is used to remove background effects using visible light images; Using the layered stereo database described in any one of claims 1-4, feature extraction is performed on infrared images and they are stored as a layered stereo database containing temperature gradients and geographic information. The isolated forest model is constructed using a hierarchical three-dimensional database based on the isolated forest algorithm. The prediction module is used to make predictions using the constructed model; The residual analysis module is used to perform residual analysis on predicted and actual values ​​to predict fault points and trend changes.

6. A computer device comprising a memory and a processor, wherein the memory stores a computer program, characterized in that, When the processor executes the computer program, it implements the data model-based transmission line temperature detection method according to any one of claims 1 to 4.

7. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by the processor, it implements the data model-based transmission line temperature detection method according to any one of claims 1 to 4.

Images