Power transmission channel large-scale inspection method adopting high-resolution satellite remote sensing data

Abstract

The invention belongs to the technical field of power transmission line operation and maintenance, particularly relates to a power transmission channel large-scale inspection method adopting high-resolution satellite remote sensing data, and aims to solve the technical problems of low inspection efficiency, limited coverage and poor hidden danger recognition accuracy in the prior art. S2, power transmission channel hidden danger classification, hidden danger model construction and automatic hidden danger interpretation; s3, power transmission channel hidden danger collaborative correction and quality inspection acceptance; s4, power transmission channel hidden danger automatic drawing and list display are carried out, and a hidden danger analysis report is generated; the method overcomes the limitation that existing manual inspection and unmanned aerial vehicle inspection are limited by terrain and climate conditions, and is suitable for large-range and multi-frequency inspection working scenes.

Description

Technical Field

[0001] The present invention belongs to the technical field of operation and maintenance of transmission lines. Background Art

[0002] In recent years, with the continuous development of society, both industrial electricity consumption and residential electricity consumption have witnessed rapid growth, and transmission channels have been continuously constructed. However, since most of the transmission channel lines are erected in natural environments such as fields, hidden dangers will be generated in this environment, which will affect transmission lines, towers, etc. Over time, various damages will be caused to the power equipment in the channel. Therefore, the inspection of hidden dangers in transmission channels has always been an important task to ensure the continuous supply of electric energy and protect the safe operation of transmission lines.

[0003] Existing inspections of transmission channels usually rely on drones to take images. Due to the complex operation environment and being easily affected by terrain and climate conditions, if there is no good shooting angle, situations such as missed reports and false alarms will occur. Moreover, the inspection efficiency is low, especially significant when inspecting large-scale areas, and there is a lack of efficient and accurate hidden danger identification technology for large-scale transmission channels. Summary of the Invention

[0004] For this reason, the present invention proposes a large-scale inspection method for transmission channels using high-resolution satellite remote sensing data, aiming to solve the technical problems of low inspection efficiency, limited coverage, and poor accuracy of hidden danger identification in the prior art through satellite remote sensing images and automated hidden danger interpretation. The specific steps are as follows:

[0005] As Figure 1 shown, S1, satellite data shooting of transmission channels and raster data slicing;

[0006] S2, classification of hidden dangers in transmission channels, construction of hidden danger models, and automated hidden danger interpretation;

[0007] S2.2, obtaining the optimal hidden danger model through the construction and training of the hidden danger model:

[0008] The construction of the hidden danger model uses a hybrid architecture of CNN (Convolutional Neural Network) and GNN (Graph Neural Network) as the basic framework. Among them, CNN is used to extract local spatial features, and GNN constructs a network topology structure with two-way information flow, and realizes cross-layer feature interaction through dynamic connection weights between nodes; a hierarchical attention mechanism is embedded in the network topology structure of GNN, including a spatial attention unit and a channel attention unit, and the optimal hidden danger model is obtained through dataset training;

[0009] S2.3. Automatically interpret potential hazards using the optimal potential hazard model obtained in S2.2: Input the high-resolution remote sensing images of the power transmission channels obtained in S1 into the optimal potential hazard model. The output data is the raster results of various potential hazard classifications. Then, convert the raster results into vectors, remove the background part of the vector prediction results, and only retain the extraction results of the target elements. Calculate the area of each feature patch, and finally complete the optimization operation according to actual requirements.

[0010] S3. Collaboratively correct and quality inspect and accept potential hazards in the power transmission channels; Create a fishing net according to the image range, and delete incorrect potential hazard patches, supplement missing potential hazards, and correct the patch boundaries for the extracted potential hazard vectors according to the fishing net. Re-inspect the results after collaborative correction according to the fishing net until there are no incorrect patches or patches with inaccurate boundaries, and the quality inspection and acceptance are completed.

[0011] S4. Automatically generate maps and display lists of potential hazards in the power transmission channels, and generate a potential hazard analysis report.

[0012] Through spatial overlay analysis of geographical data, overlay the vector of the finally qualified potential hazard patches in S3 on the satellite images of the power transmission channels, and use data-driven analysis tools to automatically generate a potential hazard distribution map of the power transmission channels. At the same time, use the nearest neighbor analysis of geographical information technology to generate a list of potential hazards in the power transmission channels, which can accurately locate potential hazards; Generate a detailed potential hazard analysis report in combination with the potential hazard distribution map and the list of potential hazards.

[0013] Technical effects:

[0014] 1. The present invention proposes a large-scale inspection method for power transmission channels using high-resolution satellite remote sensing data. This method overcomes the limitations of existing manual inspections and drone inspections restricted by terrain and climate conditions, and can quickly and comprehensively identify and analyze potential hazards in large-scale areas.

[0015] 2. The present invention adopts a potential hazard model that combines deep learning and traditional machine learning algorithms. By automatically interpreting potential hazards, it significantly improves the accuracy of potential hazard identification, reduces false alarms and missed detections; Through the collaborative correction and quality inspection and acceptance mechanisms, the results of potential hazard identification are further optimized, ensuring the data quality and reliability during the inspection process.

[0016] 3. Using geographical information technology and data-driven analysis tools, automatically generate a potential hazard distribution map, a list of potential hazards, and a detailed potential hazard analysis report, which not only improves the inspection efficiency, provides intelligent and visual decision-making support for operation and maintenance personnel, but also provides a sustainable and intelligent solution for the power industry.

[0017] Using the existing method of drone inspection, the average inspection distance is 30 kilometers per day, but a large number of drone batteries need to be prepared. It takes 106 days to inspect all lines. At the same time, considering the complex terrain and environment around the transmission corridor, when vehicles cannot reach, it takes more time, and it is more difficult to ensure the safety of the drone operators. When using the present invention for large-scale inspection work, it takes 7 days for satellite data shooting and rapid processing of the transmission corridor, 1 day for classification and interpretation of potential hazards in the transmission corridor, and the collaborative correction and quality inspection and acceptance of potential hazards in the transmission corridor is 50 kilometers per day on average. It takes a total of 72 days to inspect all lines. Moreover, the inspection work is less affected by external conditions such as terrain. After accurately positioning the potential hazards, the operation and maintenance personnel only need to conduct a small-scale on-site verification work on the places with a higher degree of potential hazard. Therefore, the present invention is applicable to the inspection work scenarios with a large range and multiple frequencies. BRIEF DESCRIPTION OF THE DRAWINGS

[0018] Figure 1 It is a block diagram of the overall process of the present invention.

[0019] Figure 2 It is a block diagram of the collaborative correction and quality inspection and acceptance of potential hazards in the transmission corridor in S3.

[0020] Figure 3 It is a distribution map of potential hazards in the transmission corridor output by S4. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0021] Next, the technical solutions in the embodiments of the present invention will be clearly and completely described with reference to the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all the embodiments. By using the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative efforts shall fall within the protection scope of the present invention.

[0022] The specific steps in this embodiment are as follows:

[0023] S1. Satellite data shooting of the transmission corridor and raster data slicing;

[0024] Further, the specific content of S1 is as follows:

[0025] S1.1. By using the geographical data spatial topology analysis method, a buffer zone of the transmission corridor is constructed, and the buffer zone range can be set at 15m or 100m;

[0026] S1.2. Using constellation high-resolution series satellites, sub-meter high-resolution remote sensing images within the buffer zone of the transmission corridor are obtained;

[0027] S1.3. Preprocess the obtained remote sensing images. The preprocessing includes geometric correction, radiometric correction, and atmospheric correction; and slice the raster data, which can reduce the volume of entity data and improve the data processing efficiency;

[0028] S2. Classify potential hazards in the transmission corridor, construct a hazard model, and perform automated hazard interpretation;

[0029] Further, the specific content of S2 is as follows:

[0030] S2.1. Classify potential hazards in the transmission corridor into agricultural greenhouses, plastic films, color steel tiles, dust-proof nets, buildings, water bodies, highways, railways, construction sites, mined areas, forestlands, and cultivated lands;

[0031] S2.2. Obtain the optimal hazard model through the construction and training of the hazard model:

[0032] The construction of the hazard model uses a hybrid architecture of CNN (Convolutional Neural Network) and GNN (Graph Neural Network) as the basic framework. Among them, CNN is used to extract local spatial features, and GNN constructs a mesh topology structure with two-way information flow, and realizes cross-layer feature interaction through the dynamic connection weights between nodes; a hierarchical attention mechanism is embedded in the mesh topology structure of GNN, including a spatial attention unit and a channel attention unit, which can dynamically weight the spatial dimension and channel dimension of the feature map respectively, effectively focusing on the key feature regions; and obtain the optimal hazard model through dataset training;

[0033] S2.3. Perform automated hazard interpretation through the optimal hazard model obtained in S2.2:

[0034] Input the high-resolution remote sensing image of the transmission corridor obtained in S1 into the optimal hazard model. In this embodiment, the output data is the raster results of 12 hazard classifications, then convert the raster results into vectors, and remove the background part of the vector prediction results, only retaining the extraction results of the target elements; calculate the area of each element patch, and finally perform operations such as removing fragmented patches and filling holes by setting an area threshold according to actual needs, and optimize the connectivity of the extraction results to ensure the integrity of the final results.

[0035] Further, the specific training of the hazard model in S2.2 is as follows:

[0036] S2.2.1. Dataset construction: The data is the sub-meter high-resolution remote sensing image obtained in S1.2. Construct a dataset based on the hazard classification in S2.1 to ensure that the collected data covers different seasons, weather conditions, and time periods, thereby increasing the robustness of the hazard model; the data annotation information includes hazard type, location, and boundary information; divide the dataset into a training set, a validation set, and a test set according to a ratio of 7:2:1, which are used for model training, parameter tuning, and evaluation respectively;

[0037] S2.2.2. Model Training: Initialize the weights of the model using random initialization; use the cross-entropy loss method as the loss function; select the stochastic gradient descent method as the optimizer, and dynamically adjust the learning rate by monitoring the model performance to improve the training effect;

[0038] S2.2.3. Model Evaluation: In this embodiment, the hidden danger model is evaluated by evaluation indicators including accuracy, recall rate, and F1 score. At the same time, cross-validation is used to avoid overfitting or underfitting of the model;

[0039] S3. Collaborative Correction and Quality Inspection and Acceptance of Hidden Dangers in Transmission Channels;

[0040] Collaboratively correct the hidden dangers automatically extracted from the transmission channels in S2.3. Input the high-resolution satellite image slices of the transmission channels and the hidden danger vector results into the remote sensing interpretation platform. Create a fishing net according to the image range. In this embodiment, the fishing net is 100m×100m. Delete the wrong hidden danger patches, supplement the missing hidden dangers, and correct the patch boundaries according to the fishing net; Re-inspect the results after collaborative correction according to the fishing net until there are no wrong patches or patches with inaccurate boundaries, and the quality inspection and acceptance are completed; As Figure 2 shown, accumulate hidden danger samples through collaborative correction, continuously optimize the hidden danger recognition algorithm, complete the iteration of the optimal hidden danger model, and continuously improve the accuracy and reliability of hidden danger recognition.

[0041] S4. Automatically Generate Maps and List Displays of Hidden Dangers in Transmission Channels, and Generate Hidden Danger Analysis Reports;

[0042] Through geospatial overlay analysis, overlay the vector of the finally qualified hidden danger patches in S3 on the satellite image of the transmission channel, and use data-driven analysis tools to automatically generate a distribution map of hidden dangers in the transmission channel, as Figure 3 shown. At the same time, use the nearest neighbor analysis of geographic information technology to generate a list of hidden dangers in the transmission channel, which can accurately locate the hidden dangers.

[0043] Furthermore, the distribution map of hidden dangers in the transmission channel includes images, hidden dangers, transmission lines, compass, scale, legend, and image time; The list of hidden dangers includes patch number, patch type, longitude, latitude, area, nearest tower number, and horizontal distance information from the tower; The content of the hidden danger analysis report includes basic information of the transmission line, basic information of the satellite image, quantity and type of hidden dangers, distribution of hidden dangers, and positional relationship between hidden dangers and transmission lines; Based on the type and location of hidden dangers in the transmission channel and combined with the scale of the transmission line, make a clear plan for the inspection priority. After this invention, the distribution map of hidden dangers in the transmission channel (.jpg), the list of hidden dangers (.excel), and the hidden danger analysis report (.pdf) are output, and the result form can be customized according to needs to meet different management requirements.

[0044] The content not described in detail in this specification belongs to the prior art well-known to those skilled in the art. At the same time, for those of ordinary skill in the art, there will be changes in the specific implementation manners and application scopes according to the idea of the present invention. In summary, the content of this specification should not be construed as a limitation on the present invention.

Claims

1. A large-scale inspection method for power transmission channels using high-resolution satellite remote sensing data, characterized in that: The steps include: S1, satellite data capture and raster data slicing of power transmission channels; S2. Classification of hidden dangers in power transmission channels, construction of hidden danger models and automated hidden danger interpretation; S2.

2. Obtain the optimal hidden danger model through hidden danger model construction and training: The hidden danger model is constructed using a hybrid architecture of CNN and GNN as the basic framework, where CNN is used to extract local spatial features, and GNN constructs a mesh topology with bidirectional information flow, and realizes cross-layer feature interaction through dynamic connection weights between nodes; a hierarchical attention mechanism is embedded in the mesh topology of GNN, including spatial attention units and channel attention units, and the optimal hidden danger model is obtained through data set training; S2.3, automatic hazard interpretation is performed through the optimal hazard model obtained in S2.2: the high-resolution transmission channel remote sensing image obtained in S1 is input into the optimal hazard model, and the output data is the raster results of various hazard classifications. The raster results are then converted into vectors, and the background part of the vector prediction results is removed, and only the extraction results of the target elements are retained; the area of ​​each element patch is calculated, and finally the optimization operation is completed according to actual needs; S3. Coordinated correction and quality inspection and acceptance of hidden dangers in power transmission channels; S4. The hidden dangers of the transmission channel are automatically mapped and listed, and a hidden danger analysis report is generated.

2. The large-scale inspection method for power transmission channels using high-resolution satellite remote sensing data according to claim 1 is characterized in that: The S1 is specifically: S1.

1. Construct a transmission channel buffer zone through the spatial topological structure analysis method of geographic data; S1.

2. Use the constellation's high-resolution satellites to obtain sub-meter high-resolution remote sensing images within the transmission channel buffer zone; S1.

3. Perform preprocessing operations on the acquired remote sensing images, including geometric correction, radiation correction and atmospheric correction; and slice the raster data to reduce the volume of physical data and improve data processing efficiency.

3. The large-scale inspection method for power transmission channels using high-resolution satellite remote sensing data according to claim 1 is characterized in that: S2.

1. Hidden dangers in power transmission channels are classified into agricultural greenhouses, ground films, color steel tiles, dust screens, buildings, water bodies, roads, railways, construction sites, mining areas, forests, and cultivated land.

4. The large-scale inspection method for power transmission channels using high-resolution satellite remote sensing data according to claim 1 is characterized in that: The hidden danger model training in S2.2 is specifically as follows: S2.2.

1. Dataset construction: The data is the sub-meter high-resolution remote sensing image obtained in S1.

2. The data set is constructed based on the hidden danger classification in S2.

1. The data annotation information includes the hidden danger type, location and boundary information. The data set is divided into training set, validation set and test set in a ratio of 7:2:1, which are used for model training, parameter adjustment and evaluation respectively. S2.2.2, Model training: Use random initialization to initialize the model weights; use the cross entropy loss method as the loss function; the optimizer selects the stochastic gradient descent method, and monitors the model performance and dynamically adjusts the learning rate to improve the training effect; S2.2.

3. Model evaluation: The hidden danger model is evaluated through evaluation indicators including precision, recall rate, and F1 score, and cross-validation is performed at the same time.

5. The large-scale inspection method for power transmission channels using high-resolution satellite remote sensing data according to claim 1 is characterized in that: The S3 is specifically as follows: creating a fishing net according to the image range, deleting erroneous hidden danger spots, supplementing missing hidden dangers, and correcting spot boundaries for the extracted hidden danger vectors according to the fishing net; quality checking the results after collaborative correction again according to the fishing net until there are no erroneous spots or spots with inaccurate boundaries, and the quality inspection and acceptance is completed.

6. The large-scale inspection method for power transmission channels using high-resolution satellite remote sensing data according to claim 1, characterized in that: Specifically, S4 is as follows: through spatial overlay analysis of geographic data, the finally qualified hidden danger map vectors in S3 are superimposed on the satellite image of the transmission channel, and a data-driven analysis tool is used to automatically generate a hidden danger distribution map of the transmission channel. At the same time, a neighbor analysis of geographic information technology is used to generate a list of hidden dangers in the transmission channel, which can accurately locate the hidden dangers; a detailed hidden danger analysis report is generated in combination with the hidden danger distribution map of the transmission channel and the hidden danger list.

7. The large-scale inspection method for power transmission channels using high-resolution satellite remote sensing data according to claim 6 is characterized in that: The transmission channel hidden danger distribution map includes images, hidden dangers, transmission lines, compass, scale, legend, and image time; the hidden danger list includes map block number, map block type, longitude, latitude, area, nearest tower number, and horizontal distance information from the tower; the hidden danger analysis report includes basic information about transmission lines, basic information about satellite images, the number and type of hidden dangers, the distribution of hidden dangers, and the location relationship between hidden dangers and transmission lines.

Images