Monitoring method and system for full life cycle of transformer substation project

Abstract

The invention relates to a monitoring method and system for the full life cycle of a transformer substation project, and the method comprises the following steps: deploying an unmanned plane automatic airport to collect environment data, and selecting a transformer substation project construction site according to the actual construction demands and the environment data; modeling the power generation and transformation project based on the monitoring data of the power generation and transformation project construction site, and constructing a hidden danger sample library; constructing an intelligent monitoring system which comprises a hidden danger recognition module, an expert rechecking module and a decision generation module; wherein the hidden danger identification module performs hidden danger identification by using an improved YOLOv8n network to obtain a hidden danger target; the improved YOLOv8n network is specifically characterized in that a convolution attention module and an adaptive feature fusion module are embedded in a backbone network; the hidden danger target passing through the expert rechecking module is input into a decision generation module, and a hidden danger decision is generated and issued to an unmanned aerial vehicle automatic airport; the unmanned aerial vehicle automatic airport receives and controls the unmanned aerial vehicle to execute the hidden danger decision.

Description

Technical Field

[0001] The present application relates to the field of power substation project monitoring, and mainly to a monitoring method and system for the entire life cycle of a power substation project. Background Art

[0002] As the scale of power systems continues to expand and their intelligence levels continue to improve, the drawbacks of traditional power engineering construction process control and operation and maintenance inspection models are becoming increasingly prominent. Currently, the planning, site selection, survey and design, construction, and operation and maintenance inspections of power generation and transformation project sites primarily rely on on-site personnel equipped with drones. These personnel use drone aerial surveys to collect spatial information about the site, analyze terrain, and assess construction progress and quality. Inspections and hazard identification are carried out using video and photo capture of the site, combined with manual visual assessment. From initial site selection to commissioning, the collection of site data by drone flights typically requires 5-10 times. During operation and maintenance, due to the increased need for monitoring, the number of collections increases exponentially.

[0003] However, this traditional operation model presents numerous problems. First, labor costs remain high, requiring specialized teams to travel back and forth to the site at each stage. This is particularly true in areas with high altitudes, dense jungles, and complex waters, where personnel mobilization is difficult and the operation cycle is long. Second, operational efficiency is low. A single aerial survey requires multiple steps, including equipment transportation, on-site commissioning, data collection, and post-processing, and is significantly constrained by environmental factors such as weather and terrain. Third, operation and maintenance costs continue to rise, with frequent on-site operations leading to equipment wear and tear, increased labor input, and time costs. Fourth, detection capabilities are inherently flawed. Human visual interpretation is heavily influenced by subjective experience, making it difficult to accurately identify hidden defects. Furthermore, the limited viewing angle and coverage of drone aerial surveys can easily create blind spots, making it even more difficult to ensure the safety and reliability of manual operations in harsh environmental conditions. Furthermore, existing drone inspection technologies are often limited to single scenarios or specific phases, lacking a systematic monitoring solution for the entire lifecycle of power generation and transformation projects. This inability to achieve intelligent monitoring and diagnosis throughout the entire process, from preliminary surveys, construction, and operation and maintenance. Summary of the Invention

[0004] In order to solve the above problems existing in the prior art, the present application provides a monitoring method and system for the entire life cycle of a power substation project.

[0005] The technical solution of this application is as follows:

[0006] In one aspect, the present invention provides a monitoring method for a power substation project throughout its entire life cycle, the method comprising:

[0007] Deploy drone automated airports based on candidate power substation project sites, use drones to collect environmental data of candidate substation project sites, and select power substation construction sites based on actual construction needs and environmental data;

[0008] Obtain monitoring data from the power generation and substation construction site, model the power generation and substation site environment based on the monitoring data, and generate surveying and mapping results; construct an initial hidden danger sample library based on the surveying and mapping results and monitoring data, and use annotation software to mark the hidden danger areas to obtain the hidden danger sample library;

[0009] Construct an intelligent monitoring system, including a hidden danger identification module, an expert review module, and a decision-making module. The hidden danger identification module uses an improved YOLOv8n network to identify hidden dangers in a hidden danger sample library, obtain hidden danger targets, and push them to the expert review module. The improved YOLOv8n network specifically embeds a convolutional attention module and an adaptive feature fusion module in the backbone network.

[0010] The hidden danger targets that have passed the expert review module are input into the decision generation module, and the hidden danger decision is generated and sent to the drone automatic airport; the drone automatic airport receives and controls the drone to execute the hidden danger decision.

[0011] Preferably, drones are used to collect environmental data of candidate substation project sites, specifically:

[0012] The environmental data includes topographic features, distribution of adverse geological conditions, location and height of buildings, and distribution of high-voltage power towers; and marking of no-fly zones, restricted-fly zones, and safe flight paths;

[0013] Build a risk quantification model based on environmental data, calculate the environmental data and its corresponding weights, and obtain the risk value of each candidate power substation project site;

[0014] Select the construction site for the power substation project based on actual construction needs and risk value.

[0015] Preferably, monitoring data of the power substation construction site is obtained, specifically:

[0016] Through the image control points arranged at the four corners and the center of the power substation construction site, all-weather data collection is carried out on the power substation construction site;

[0017] During daytime missions, after verifying the initial inspection coordinates using the global navigation satellite system, the drone flies along a preset optimal inspection route, using its onboard high-resolution optical camera and lidar to collect visible light images and 3D point cloud data; performs geometric and radiometric correction on the visible light images; and performs denoising filtering and ground point classification on the 3D point cloud data.

[0018] During night missions, the drone switches to the infrared thermal imager, extracts temperature information from thermal infrared photos, converts the format of the thermal infrared photos, and obtains thermal infrared images with pixel points as temperature values; it performs temperature field analysis on the thermal infrared images and identifies areas with abnormal temperature rise.

[0019] Preferably, the surveying and mapping results include true orthogonal images, digital elevation models, digital surface models and three-dimensional real scene models;

[0020] The three-dimensional real scene model is specifically based on visible light images and three-dimensional point cloud data, using a dense matching algorithm to generate a high-density three-dimensional point cloud;

[0021] The Poisson surface reconstruction algorithm is used to reconstruct the high-density 3D point cloud to generate a 3D model of the power substation construction site.

[0022] Preferably, the hidden danger identification module is specifically:

[0023] The hidden danger sample library is input into the improved YOLOv8n network, and the backbone network in the improved YOLOv8n network is used to extract feature maps. The improved convolutional attention module is embedded in the key layers of the backbone network to assign weights to the channels and spatial dimensions of the feature maps.

[0024] The improved convolutional attention module includes two submodules working together. The first channel attention submodule is used to analyze the importance of different feature channels. Specifically, it performs global average pooling and maximum pooling on the received feature map to generate two one-dimensional feature vectors. The two one-dimensional feature vectors are weighted according to the channel dimension through a multi-task learning layer and normalized by an activation function to obtain a channel weight feature map.

[0025] The second spatial attention submodule performs average pooling and maximum pooling on the received feature maps to generate two two-dimensional feature maps. The two two-dimensional feature maps are concatenated and input into the convolutional layer for spatial dimension weight distribution to obtain a spatial weight feature map.

[0026] Multiply the channel weight feature map and the spatial weight feature map to obtain a fused feature map;

[0027] An adaptive feature fusion module is embedded in the key layers of the backbone network to dynamically extract feature information from the fused feature map. Specifically, under sufficient lighting conditions, visible light images are used as the primary feature information to extract texture and morphological features; at night or in severe weather, thermal infrared images are used as the primary feature information to extract temperature anomalies and partial discharge features.

[0028] The extracted multi-source feature information is weightedly fused to generate fused features; the improved YOLOv8n network is used to detect hidden dangers on the fused features to obtain hidden danger targets.

[0029] Preferably, the expert review module is a dual-track mechanism that combines preliminary screening of the improved YOLOv8n network with expert review based on expert experience, wherein:

[0030] The monitoring intelligent system pushes the hidden danger target to the expert review interface for multi-expert parallel review, including hidden danger target confirmation, classification correction and severity rating, to obtain the review results.

[0031] Preferably, the decision-making generation module includes extracting key information from the reviewed hidden danger targets and generating a structured detection report.

[0032] On the other hand, the present invention also proposes a monitoring system for the entire life cycle of a substation project, the system comprising a substation site selection module, a sample library construction module, and a hidden danger handling module, wherein:

[0033] The substation site selection module is used to deploy drone automated airports based on candidate substation project sites, use drones to collect environmental data of the candidate substation project sites, and select the substation construction site based on actual construction needs and environmental data;

[0034] The sample library construction module is used to obtain monitoring data of the power generation and transformation station construction site, model the power generation and transformation project site environment based on the monitoring data, and generate surveying and mapping results; construct an initial hidden danger sample library based on the surveying and mapping results and the monitoring data, and use annotation software to mark the hidden danger areas to obtain the hidden danger sample library;

[0035] The hidden danger processing module has a built-in monitoring intelligent system, including a hidden danger identification module, an expert review module, and a decision-making module. The hidden danger identification module uses an improved YOLOv8n network to identify hidden dangers in the hidden danger sample library, obtain hidden danger targets, and push them to the expert review module. The improved YOLOv8n network specifically embeds a convolutional attention module and an adaptive feature fusion module in the backbone network.

[0036] The hidden danger targets that have passed the expert review module are input into the decision generation module, and the hidden danger decision is generated and sent to the drone automatic airport; the drone automatic airport receives and controls the drone to execute the hidden danger decision.

[0037] On the other hand, the present invention also proposes an electronic device, including a memory, a processor, and a computer program stored in the memory and runnable on the processor. When the processor executes the program, it implements a monitoring method for the entire life cycle of a power substation project as described in any embodiment of the present invention.

[0038] On the other hand, the present invention further provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements a monitoring method for the entire life cycle of a power substation project as described in any embodiment of the present invention.

[0039] Compared with the prior art, the present invention has the following beneficial effects:

[0040] 1) This invention provides a monitoring method and system for the entire life cycle of power and substation projects. This system uses drones to collect multi-dimensional environmental data, such as geography, meteorology, and topography, from candidate sites, and conducts data analysis and modeling based on project construction requirements. This data-driven site selection method avoids the subjectivity and limitations of traditional manual site selection, enabling the selection of sites that better meet project requirements. This improves the scientific and rational nature of substation site selection, reducing the risks of later-stage operation and maintenance or increased construction costs caused by inappropriate site selection.

[0041] 2) The present invention provides a monitoring method and system for the entire life cycle of power substation projects. Through an improved YOLOv8n network, a convolutional attention module and an adaptive feature fusion module are embedded in the backbone network. The former focuses on key feature areas through weight distribution, suppresses irrelevant information, and enhances the expressiveness of key features. The latter dynamically fuses features at different levels. By fusing different features, the detection accuracy and recall rate of hidden danger targets can be improved, and the problems of missed detection and false detection can be reduced. The ability to identify early hidden dangers is enhanced, and monitoring efficiency is improved.

[0042] 3) The present invention provides a monitoring method and system for the entire life cycle of power substation projects. The constructed intelligent body forms an intelligent closed loop of the entire process through the coordinated linkage of three modules: "hidden danger identification-expert review-decision generation"; the expert review module introduces an industry experience rule library to verify the results, make up for the limitations of the algorithm, and improve the reliability and explainability of the decision; the decision generation module is linked to the drone automatic airport to execute instructions in real time, realize the automation and immediate response of hidden danger disposal, and shorten the problem response cycle; it improves the intelligence level of substation full life cycle management, hidden danger handling efficiency and system robustness. DETAILED DESCRIPTION

[0043] The specific embodiments of the present invention are described below to facilitate understanding of the present invention by those skilled in the art. However, it should be clear that the present invention is not limited to the scope of the specific embodiments. For those skilled in the art, as long as various changes are within the spirit and scope of the present invention as defined and determined by the appended claims, these changes are obvious, and all inventions and creations utilizing the concepts of the present invention are protected.

[0044] The present invention provides the following technical solution: a monitoring method and system for the entire life cycle of a power substation project.

[0045] Example 1:

[0046] This embodiment provides a monitoring method for the entire life cycle of a power substation project. The specific steps include:

[0047] S1. Deploy drone automated airports based on candidate power substation project sites, use drones to collect environmental data of the candidate substation project sites, and select power substation construction sites based on actual construction needs and environmental data;

[0048] S11. The environmental data includes topographic features, distribution of adverse geological conditions, location and height of buildings, and distribution of high-voltage power towers; and marking of no-fly zones, restricted-fly zones, and safe flight paths;

[0049] In this embodiment, the no-fly zone is a circular area with a radius of 50 meters and a high-voltage line as the center; the restricted-fly zone requires the flight altitude to be higher than 30 meters above the height of the building complex;

[0050] S12. Build a risk quantification model based on environmental data, calculate the environmental data and its corresponding weights, and obtain the risk value of each candidate power substation project site, which can be expressed as:

[0051]

[0052] In the formula, Risk j represents the risk value of the jth candidate power generation and transformation project site; w i,j F represents the weight of the i-th environmental data of the j-th candidate power generation and transformation project site; i,j represents the i-th environmental data of the j-th candidate power generation and transformation project site; n represents the number of candidate power generation and transformation project sites; j represents the index value of the j-th candidate power generation and transformation project site; i represents the index value of the i-th environmental data;

[0053] Select the construction site of the power substation project based on actual construction needs and risk value;

[0054] The method further includes constructing a risk heat map based on the risk value, and setting a first warning area, a second warning area, and a third warning area;

[0055] If the risk value of the current candidate power substation project site is within the first warning area, it means that the current candidate power substation project site is a low-risk area, which is represented by green on the risk heat map;

[0056] If the risk value of the current candidate power substation project site is within the second warning area, it means that the current candidate power substation project site is a medium risk area, which is indicated by yellow on the risk heat map;

[0057] If the risk value of the current candidate power substation project site is within the third warning area, it means that the current candidate power substation project site is a high-risk area, which is indicated by red on the risk heat map;

[0058] The method further includes setting up emergency alternate routes and alternate airports around the candidate power substation project site to form a coordinated rescue network with the drone airport;

[0059] S2. Obtain monitoring data of the power substation construction site;

[0060] Through the image control points arranged at the four corners and the center of the power substation construction site, all-weather data collection is carried out on the power substation construction site;

[0061] During daytime missions, after verifying the initial inspection coordinates using the global navigation satellite system, the drone flies along a preset optimal inspection route, using its onboard high-resolution optical camera and lidar to collect visible light images and 3D point cloud data; performs geometric and radiometric correction on the visible light images; and performs denoising filtering and ground point classification on the 3D point cloud data.

[0062] During night missions, the drone switches to the infrared thermal imager, extracts temperature information from thermal infrared photos, converts the format of the thermal infrared photos, and obtains thermal infrared images with temperature values ​​as pixels. The thermal infrared images are then analyzed for temperature fields to identify areas with abnormal temperature rise.

[0063] S22. Model the power generation and transformation project based on monitoring data and generate surveying and mapping results;

[0064] The surveying and mapping results include true orthogonal images, digital elevation models, digital surface models and three-dimensional real scene models;

[0065] The true orthophoto map uses aerial triangulation technology and ground control points to perform adjustment calculations to generate high-precision image exterior orientation elements; a pixel-level image matching algorithm is used to automatically stitch visible light images to ultimately generate a digital orthophoto map;

[0066] The digital elevation model is based on the classified ground point cloud data and uses an irregular triangulated network interpolation method to generate a grid digital elevation model;

[0067] The digital surface model is based on three-dimensional point cloud data and uses an irregular triangulated network interpolation method to generate a grid digital elevation model. Compared with the digital elevation model, the digital surface model only contains the elevation information of the terrain and does not contain other surface information. The digital surface model covers the elevation of other surface information except the ground.

[0068] The true orthogonal image, digital elevation model and digital surface model are in TIF format and are used in the power design process of the subsequent power substation project to meet the power design requirements;

[0069] The three-dimensional real scene model is in OBJ format, specifically, based on visible light images and three-dimensional point cloud data, a high-density three-dimensional point cloud is generated using a dense matching algorithm;

[0070] Use the Poisson surface reconstruction algorithm to reconstruct the high-density 3D point cloud and generate a 3D model of the substation construction site;

[0071] S23. Build an initial hidden danger sample library based on surveying and mapping results and monitoring data, and use annotation software to mark the hidden danger areas to obtain the hidden danger sample library;

[0072] The hidden danger sample library is in TIF format and covers various hidden danger types, including geological hazards (such as landslides and collapses), high-risk behaviors (failure to tie safety ropes, personnel entering the lifting area, etc.), and environmental risks (open flames, smoke diffusion, etc.). The sample library targets include personnel, vehicles, structures, equipment, supplies and materials, and other targets of concern for power generation and transformation projects.

[0073] S3. Build an intelligent monitoring system, including a hidden danger identification module, an expert review module, and a decision-making module, and push the results to the expert review module. The hidden danger identification module uses an improved YOLOv8n network to identify hidden dangers in the hidden danger sample library and obtain hidden danger targets. The improved YOLOv8n network specifically embeds a convolutional attention module and an adaptive feature fusion module in the backbone network.

[0074] S31, the hidden danger identification module specifically inputs the hidden danger sample library into the improved YOLOv8n network, extracts the feature map through the backbone network in the improved YOLOv8n network; embeds the improved convolutional attention module in the key layer of the backbone network, and assigns weights to the channels and spatial dimensions of the feature map;

[0075] S311, the improved convolutional attention module includes two submodules working together. The first channel attention submodule is used to analyze the importance of different feature channels, specifically performing global average pooling and maximum pooling on the received feature map to generate two one-dimensional feature vectors; the two one-dimensional feature vectors are subjected to a multi-task learning layer to perform channel dimension weight distribution, and are normalized by an activation function to obtain a channel weight feature map;

[0076] The channel weight feature map is expressed as follows:

[0077] M c =σ(MLP(I avg )+MLP(I max ));

[0078] Where M c represents the channel weight feature map; σ represents the Sigmoid function; MLP represents the multi-task learning layer; I avg represents the average pooled one-dimensional feature vector; I max represents the maximum pooled one-dimensional feature vector;

[0079] S312, the second spatial attention submodule performs average pooling and maximum pooling on the received feature map to generate two two-dimensional feature maps; the two two-dimensional feature maps are spliced ​​and input into the convolution layer for weight distribution of the spatial dimension to obtain a spatial weight feature map;

[0080] The spatial weight feature map is expressed as follows:

[0081] M s =σ(K k×k (Concat(I' avg ,I' max )));

[0082] Where M s Represents the spatial weight feature map; K k×k Represents a convolutional layer with k layers; I' avg represents the average pooled two-dimensional feature vector; I' max Represents the maximum pooled two-dimensional feature vector; Concat represents the concatenation function;

[0083] S313, multiplying the channel weight feature map and the spatial weight feature map to obtain a fusion feature map, which is expressed as:

[0084] M r =M c ⊙M s ;

[0085] Where M rRepresents the fused feature map; ⊙ represents the element-by-element multiplication operation;

[0086] S314. Embed an adaptive feature fusion module in the key layers of the backbone network to dynamically extract feature information of the fused feature map. Specifically, under sufficient lighting conditions, the visible light image is used as the main feature information to extract texture and morphological features; at night or in bad weather, the imager image is used as the main feature information to extract temperature anomaly and partial discharge features.

[0087] The extracted multi-source feature information is weighted and fused to generate fused features. The improved YOLOv8n network is used to detect hidden dangers on the fused features and obtain hidden danger targets.

[0088] S32. The expert review module is specifically a dual-track mechanism that combines preliminary screening of the improved YOLOv8n network with expert review based on expert experience, wherein:

[0089] The monitoring intelligent system pushes the hidden danger target to the expert review interface for multi-expert parallel review, including hidden danger target confirmation, classification correction and severity rating, to obtain the review results;

[0090] S33, the decision generation module includes extracting key information from the reviewed hidden danger targets and generating a structured detection report;

[0091] S4. Input the hidden danger targets that have passed the expert review module into the decision generation module, generate the hidden danger decision and send it to the UAV automatic airport; the UAV automatic airport receives and controls the UAV to execute the hidden danger decision.

[0092] Example 2:

[0093] This embodiment provides a monitoring system for the entire life cycle of a substation project. The system includes a substation site selection module, a sample library construction module, and a hidden danger handling module, wherein:

[0094] The substation site selection module is used to deploy drone automated airports based on candidate substation project sites, use drones to collect environmental data of the candidate substation project sites, and select the substation construction site based on actual construction needs and environmental data;

[0095] The sample library construction module is used to obtain monitoring data of the power generation and transformation station construction site, model the power generation and transformation project site environment based on the monitoring data, and generate surveying and mapping results; construct an initial hidden danger sample library based on the surveying and mapping results and the monitoring data, and use annotation software to mark the hidden danger areas to obtain the hidden danger sample library;

[0096] The hidden danger processing module has a built-in monitoring intelligent system, including a hidden danger identification module, an expert review module, and a decision-making module. The hidden danger identification module uses an improved YOLOv8n network to identify hidden dangers in the hidden danger sample library, obtain hidden danger targets, and push them to the expert review module. The improved YOLOv8n network specifically embeds a convolutional attention module and an adaptive feature fusion module in the backbone network.

[0097] The hidden danger targets that have passed the expert review module are input into the decision generation module, and the hidden danger decision is generated and sent to the drone automatic airport; the drone automatic airport receives and controls the drone to execute the hidden danger decision.

[0098] Example 3:

[0099] This embodiment provides an electronic device, including a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the program, a monitoring method for the entire life cycle of a power substation project as described in any embodiment of the present invention is implemented.

[0100] Example 4:

[0101] This embodiment provides a computer-readable storage medium having a computer program stored thereon. When the program is executed by a processor, a monitoring method for the entire life cycle of a power substation project as described in any embodiment of the present invention is implemented.

[0102] It is worth noting that the system, electronic device and computer-readable storage medium described in the present invention are all based on the same principles as the method described in Example 1, and will not be repeated here.

[0103] The above descriptions are merely embodiments of the present invention and are not intended to limit the patent scope of the present invention. Any equivalent structure or equivalent process transformation made using the contents of the present invention specification, or directly or indirectly applied in other related technical fields, are also included in the patent protection scope of the present invention.

Claims

1. A monitoring method for the entire life cycle of a power substation project, characterized in that: The method comprises: Deploy drone automated airports based on candidate power substation project sites, use drones to collect environmental data of candidate substation project sites, and select power substation construction sites based on actual construction needs and environmental data; Obtain monitoring data from the power generation and substation construction site, model the power generation and substation site environment based on the monitoring data, and generate surveying and mapping results; construct an initial hidden danger sample library based on the surveying and mapping results and monitoring data, and use annotation software to mark the hidden danger areas to obtain the hidden danger sample library; Construct an intelligent monitoring system, including a hidden danger identification module, an expert review module, and a decision-making module. The hidden danger identification module uses an improved YOLOv8n network to identify hidden dangers in a hidden danger sample library, obtain hidden danger targets, and push them to the expert review module. The improved YOLOv8n network specifically embeds a convolutional attention module and an adaptive feature fusion module in the backbone network. The hidden danger targets that have passed the expert review module are input into the decision generation module, and the hidden danger decision is generated and sent to the drone automatic airport; the drone automatic airport receives and controls the drone to execute the hidden danger decision.

2. A monitoring method for the entire life cycle of a power substation project according to claim 1, characterized in that: Use drones to collect environmental data of candidate substation project sites, specifically: The environmental data includes topographic features, distribution of adverse geological conditions, location and height of buildings, and distribution of high-voltage power towers; and marking of no-fly zones, restricted-fly zones, and safe flight paths; Build a risk quantification model based on environmental data, calculate the environmental data and its corresponding weights, and obtain the risk value of each candidate power substation project site; Select the construction site for the power substation project based on actual construction needs and risk value.

3. A monitoring method for the entire life cycle of a power substation project according to claim 1, characterized in that: Obtain monitoring data for the power substation construction site, specifically: Through the image control points arranged at the four corners and the center of the power substation construction site, all-weather data collection is carried out on the power substation construction site; During daytime missions, after verifying the initial inspection coordinates using the global navigation satellite system, the drone flies along a preset optimal inspection route, using its onboard high-resolution optical camera and lidar to collect visible light images and 3D point cloud data; performs geometric and radiometric correction on the visible light images; and performs denoising filtering and ground point classification on the 3D point cloud data. During night missions, the drone switches to the infrared thermal imager, extracts temperature information from thermal infrared photos, converts the format of the thermal infrared photos, and obtains thermal infrared images with pixel points as temperature values; it performs temperature field analysis on the thermal infrared images and identifies areas with abnormal temperature rise.

4. A monitoring method for the entire life cycle of a power substation project according to claim 1, characterized in that: The surveying and mapping results include true orthogonal images, digital elevation models, digital surface models and three-dimensional real scene models; The three-dimensional real scene model is specifically based on visible light images and three-dimensional point cloud data, using a dense matching algorithm to generate a high-density three-dimensional point cloud; The Poisson surface reconstruction algorithm is used to reconstruct the high-density 3D point cloud to generate a 3D model of the power substation construction site.

5. The monitoring method for the entire life cycle of a power substation project according to claim 1, characterized in that: The hidden danger identification module is specifically: The hidden danger sample library is input into the improved YOLOv8n network, and the backbone network in the improved YOLOv8n network is used to extract feature maps. The improved convolutional attention module is embedded in the key layers of the backbone network to assign weights to the channels and spatial dimensions of the feature maps. The improved convolutional attention module includes two submodules working together. The first channel attention submodule is used to analyze the importance of different feature channels. Specifically, it performs global average pooling and maximum pooling on the received feature map to generate two one-dimensional feature vectors. The two one-dimensional feature vectors are passed through the multi-task learning layer to perform weight distribution on the channel dimension and normalized by the activation function to obtain the channel weight feature map; The second spatial attention submodule performs average pooling and maximum pooling on the received feature maps to generate two two-dimensional feature maps. The two two-dimensional feature maps are concatenated and input into the convolutional layer for spatial dimension weight distribution to obtain a spatial weight feature map. Multiply the channel weight feature map and the spatial weight feature map to obtain a fused feature map; An adaptive feature fusion module is embedded in the key layers of the backbone network to dynamically extract feature information from the fused feature map. Specifically, under sufficient lighting conditions, visible light images are used as the primary feature information to extract texture and morphological features; at night or in severe weather, thermal infrared images are used as the primary feature information to extract temperature anomalies and partial discharge features. Perform weighted fusion on the extracted multi-source feature information to generate fusion features; The improved YOLOv8n network is used to detect hidden dangers based on the fusion features and obtain hidden danger targets.

6. A monitoring method for the entire life cycle of a power substation project according to claim 1, characterized in that: The expert review module is specifically a dual-track mechanism that combines preliminary screening with an improved YOLOv8n network and expert review based on expert experience, where: The monitoring intelligent system pushes the hidden danger target to the expert review interface for multi-expert parallel review, including hidden danger target confirmation, classification correction and severity rating, to obtain the review results.

7. A monitoring method for the entire life cycle of a power substation project according to claim 1, characterized in that: The decision-making generation module includes extracting key information from the reviewed hidden danger targets and generating a structured detection report.

8. A monitoring system for the entire life cycle of a substation project, characterized in that: The system includes a substation site selection module, a sample library construction module, and a hidden danger processing module, wherein: The substation site selection module is used to deploy drone automated airports based on candidate substation project sites, use drones to collect environmental data of the candidate substation project sites, and select the substation construction site based on actual construction needs and environmental data; The sample library construction module is used to obtain monitoring data of the power generation and transformation station construction site, model the power generation and transformation project based on the monitoring data, and generate surveying and mapping results; and construct a hidden danger sample library based on the surveying and mapping results; The hidden danger processing module has a built-in monitoring intelligent system, including a hidden danger identification module, an expert review module, and a decision-making module. The hidden danger identification module uses an improved YOLOv8n network to identify hidden dangers in the hidden danger sample library, obtain hidden danger targets, and push them to the expert review module. The improved YOLOv8n network specifically embeds a convolutional attention module and an adaptive feature fusion module in the backbone network. The hidden danger targets that have passed the expert review module are input into the decision generation module, and the hidden danger decision is generated and sent to the drone automatic airport; the drone automatic airport receives and controls the drone to execute the hidden danger decision.

9. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein: When the processor executes the program, it implements a monitoring method for the entire life cycle of a power substation project as described in any one of claims 1 to 7.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the program is executed by a processor, a monitoring method for the entire life cycle of a power substation project as described in any one of claims 1 to 7 is implemented.

Images