Intelligent thermal power generation facility inspection method and system based on unmanned aerial vehicle

Abstract

The invention relates to the field of unmanned aerial vehicle inspection, in particular to an intelligent thermal power generation facility inspection method and system based on an unmanned aerial vehicle. Comprising an unmanned aerial vehicle, a binocular vision camera, a sensor, communication equipment, a flight management module, an autonomous charging station, an energy management module, a ground control station and an augmented reality (AR) interface. According to the method, the binocular vision technology is combined with the SLAM algorithm to construct the three-dimensional image, the structural problem of the thermal power generation facility is identified through the target detection algorithm, and the unmanned aerial vehicle can rapidly and accurately perform autonomous navigation and inspection in the complex thermal power plant depending on the efficient algorithm and the strong computing power, so that the reliability of the thermal power generation facility is improved. Not only are the inspection efficiency and safety improved, but also the labor cost is reduced, and the risk caused by human errors is reduced.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of unmanned aerial vehicle inspection, in particular to an intelligent fire power generation facility inspection method and system based on unmanned aerial vehicles. BACKGROUND

[0002] In a thermal power plant, the boiler furnace, desulfurization tower, chimney, cooling tower and rooftop photovoltaic system and other equipment are difficult for personnel to enter or climb to check due to their special structure. When defects occur in these equipment or need to be checked, a scaffold must be erected or personnel must climb the stairs platform to check. The scaffold erection is usually large in engineering quantity, high in cost and long in time, which not only increases the maintenance cost, but also increases the maintenance time. Therefore, for the above-mentioned similar closed space such as the boiler furnace, desulfurization tower and rooftop photovoltaic system, if an unmanned aerial vehicle intelligent inspection system can be developed, the work efficiency will be undoubtedly improved, the personnel operation risk will be reduced, and the production management level of enterprises will be improved, which is of great significance to the intelligentization and technologicalization of thermal power plants.

[0003] Therefore, it is urgent to develop an intelligent fire power generation facility inspection method and system based on unmanned aerial vehicles to overcome the shortcomings in the prior art. SUMMARY

[0004] The present application aims to provide an intelligent fire power generation facility inspection method and system based on unmanned aerial vehicles to solve the problems raised in the background.

[0005] To achieve the above-mentioned purpose, the present application provides the following technical solution: an intelligent fire power generation facility inspection method based on unmanned aerial vehicles, the specific implementation manner being: S101: The unmanned aerial vehicle cluster carries out an inspection task on the fire power generation facility according to a flight task; S102: The unmanned aerial vehicle identifies structural defects existing in the fire power generation facility, and monitors the operation state and temperature distribution of the fire power generation facility through a sensor; S103: The unmanned aerial vehicle automatically adjusts the subsequent flight plan and inspection route according to the comprehensive analysis result; S104: Meanwhile, the unmanned aerial vehicle clusters realize information sharing and task coordination through data communication, share the monitoring data and flight state information, and realize the adjustment of the overall subsequent flight plan and inspection route of the unmanned aerial vehicle cluster; S105: The unmanned aerial vehicle cluster transmits the monitoring data to a ground control station, and the ground control station combines the monitoring data with a three-dimensional model of the fire power generation facility to clearly display the problems existing in the fire power generation facility.

[0006] Further, when the unmanned aerial vehicle identifies the structural defects, the double-vision technology is carried on the unmanned aerial vehicle and combined with the SLAM algorithm, so that the unmanned aerial vehicle can capture and process complex visual information in real time during flight. The unmanned aerial vehicle can construct a high-precision three-dimensional model of the thermal power plant in real time according to the depth information and three-dimensional structure of the environment obtained through the double-vision technology and the SLAM algorithm. Specifically, the depth information calculated by the double-vision technology is combined with the pose of the unmanned aerial vehicle to generate a dense three-dimensional point cloud representing the three-dimensional structure of the surrounding environment. Then, the three-dimensional point cloud is filtered, denoised and optimized to eliminate redundant data and measurement errors. Finally, the point cloud data is converted into a three-dimensional model of the thermal power plant by using a three-dimensional reconstruction technology. Based on the constructed three-dimensional model of the thermal power plant, the unmanned aerial vehicle can accurately perceive the surrounding environment and avoid obstacles in real time through the path optimization algorithm deployed in the unmanned aerial vehicle, so as to ensure that the unmanned aerial vehicle flies along the predetermined inspection route or conducts more detailed inspection on the thermal power plant according to the three-dimensional model.

[0007] Further, the unmanned aerial vehicle uses a target detection algorithm to judge the size and shape of the defects existing in the stereo image provided by the double-vision technology, optimizes and corrects the depth information of the thermal power plant obtained by the double-vision technology by using the advantage of the deep learning algorithm, and identifies the cracks, corrosion and wear defects existing on the surface or shielding part of the facility under different lighting conditions. At the same time, various sensors are used to monitor the thermal power plant, and the monitoring content includes the running state and temperature of the thermal power plant.

[0008] Further, when the unmanned aerial vehicle identifies the structural defects in the thermal power plant, a defect identification model pre-deployed on the unmanned aerial vehicle is used. The defect identification model is constructed by a target detection algorithm Faster RCNN, and the structural defects include cracks, corrosion and wear. The specific steps of training the defect identification model to identify common structural defects of the thermal power plant are as follows: S111: First, collect an image data set for training and verification, which includes structural defects of the thermal power plant. At the same time, label the images in the image data set, that is, draw a bounding box for each defect existing on the image and mark the category of the defect. Then, classify the images with category labels into training set, verification set and test set; S112: Perform a pre-processing operation on the image data set, which includes adjusting the image size and normalizing the image. At the same time, rotate, flip, scale and crop the images in the image data set with category labels to increase the diversity of the data; S113: Pre-training the defect identification model using the training set to determine whether each feature block in the picture is a structural defect and to determine the precise location of the defect; S114: Evaluating the pre-trained defect identification model using the validation set, fine-tuning the defect identification model according to the evaluation results, the evaluation including accuracy, recall rate and F1 score; wherein the calculation method of accuracy is: Precision = Correctly identified image number / Total number of images identified x 100%, The calculation method of recall rate is: Recall = Correctly identified image / Total number of images x 100%, The calculation method of F1 score is: 2 x Precision x Recall / (Precision + Recall) x 100%; S115: Finally, the final performance of the defect identification model is evaluated using the test set, and the result is determined whether it meets the expected standard, no, execute step S101; yes, deploy the defect identification model to the unmanned aerial vehicle control platform, and perform the actual power plant inspection task to identify the structural defects of the power plant.

[0009] Further, the unmanned aerial vehicle automatically adjusts the subsequent flight plan according to the comprehensive analysis result, and when a problem is detected in one of the areas of the power plant, the unmanned aerial vehicle compares and analyzes the structural defect identification result of the monitored power plant with the preset normal range threshold to determine whether the operation state of the power plant in the area is abnormal; when it is determined that the operation state of the facility in the area is outside the normal range threshold, the unmanned aerial vehicle sends a cooperation request to the unmanned aerial vehicle cluster to increase the number of unmanned aerial vehicles for inspecting the area, and finds the optimal flight path of the unmanned aerial vehicle cluster in the area through a preset algorithm; when it is determined that the operation state of the facility in the area is within the normal range threshold, the unmanned aerial vehicle continues to inspect according to the predetermined flight task.

[0010] Further, the unmanned aerial vehicle finds the optimal flight path through the particle swarm optimization algorithm to plan the flight path of the unmanned aerial vehicle, and when the unmanned aerial vehicle detects a problem in one of the areas of the power plant, it not only increases the inspection frequency of the area, but also realizes information sharing and task coordination through real-time data communication of the unmanned aerial vehicle cluster, and initiates a request for assistance to other unmanned aerial vehicles in the unmanned aerial vehicle cluster to cooperatively perform more complex inspection tasks and collect more detailed data from more angles; wherein when the unmanned aerial vehicle plans the flight path through the particle swarm optimization algorithm, the calculation formula of the speed of the next time node of the unmanned aerial vehicle is:

[0011] The calculation formula of the position of the next time node of the UAV is:

[0012] wherein, is a particle, that is, a UAV in the UAV cluster; is the position of the UAV at the current time node; is an inertia weight, which determines the tendency of the UAV to maintain the current speed; is the position of the UAV at the current time node; and is a learning factor, which determines the degree to which the UAV approaches the optimal position and the global optimal position; and is a random number, the range of the random number is [0, 1], and the random number introduces randomness in the search process of the UAV; is the optimal position of the UAV , that is, the optimal solution found by the UAV in the historical data; is the global optimal position, that is, the optimal solution found by all particles in history; is the position of the UAV at the current time node; is the position of the UAV at the next time node; the position of the UAV is taken as a point of the UAV in the search space, the speed is taken as the moving direction and size of the UAV in the search space, the optimal path of the UAV is obtained by iteratively calculating the position and the speed.

[0013] Further, when the ground control station clearly displays the problems existing in the thermal power generation facility, the ground operator views the images returned by the UAV in real time through intelligent glasses or a mobile device, and overlays the three-dimensional model to clearly view and further confirm the problems existing in the thermal power generation facility; the three-dimensional model realizes operations including zooming in, zooming out, rotating, and switching to a view of a specified position through a touch screen or a voice command to display the defects in the thermal power generation facility; the three-dimensional model can superimpose the running state and the temperature of the thermal power generation facility monitored by the sensor to the corresponding positions of the three-dimensional model.

[0014] Further, a plurality of autonomous charging stations are also arranged in the thermal power plant, the autonomous charging stations are evenly distributed in the thermal power plant, that is, the distance from one of the autonomous charging stations to a nearby autonomous charging station is the same, and the autonomous charging stations are internally provided with unmanned aerial vehicle autonomous charging devices for automatic charging of the unmanned aerial vehicles.

[0015] Furthermore, the drone is set with a minimum charging power level. That is, when the drone's own power level is close to the minimum charging power level, the drone automatically plans a flight path from the current position to the nearest autonomous charging station; at the same time, the drone also has a self-diagnosis function, and automatically returns to the maintenance center for repair when it detects that it needs maintenance or the equipment is damaged.

[0016] Furthermore, an intelligent inspection system for thermal power generation facilities based on drones includes a flight management module, an autonomous charging station, an energy management module, a ground control station, and an augmented reality (AR) interface; The flight management module is used to plan and adjust the UAV flight plan, automatically adjusting the UAV's flight route and mission priority based on real-time inspection data; The autonomous charging station is an autonomous charging facility for the drone. When the drone is low on power, it will automatically return to the autonomous charging station for autonomous charging to ensure the continuity of the inspection mission. The energy management module is used to determine the preferred autonomous charging station for charging the drone; The ground control station is used to monitor the status of the drone, plan flight missions, view inspection data, and build and view three-dimensional models; The augmented reality (AR) interface is used to combine the real-time images transmitted by the drone with the three-dimensional model of the thermal power generation facility, and to visually display the problems in the thermal power generation facility through smart glasses or display devices. Beneficial effects

[0017] 1. This invention uses binocular vision technology combined with SLAM algorithms to construct three-dimensional images of thermal power generation facilities and identifies structural problems on thermal power generation facilities through target detection algorithms. Relying on efficient algorithms and powerful computing capabilities, the invention enables drones to quickly and accurately navigate and inspect complex thermal power plants autonomously. This not only improves inspection efficiency and safety, but also reduces labor costs and mitigates the risks caused by human error. 2. By adding adaptive learning algorithms to drones, they can automatically adjust their inspection frequency and flight paths. Furthermore, drone swarms enable information sharing and task coordination, enabling drones to more efficiently and intelligently complete inspections of thermal power generation facilities. This not only improves inspection quality and efficiency, but also reduces potential safety risks. 3. By using an energy management system to manage the charging issues between drones and autonomous charging stations, drones can be made more efficient, flexible and intelligent. In this way, drones can maximize their work efficiency while ensuring timely completion of tasks and reasonable distribution of power. BRIEF DESCRIPTION OF THE DRAWINGS

[0018] Figure 1 It is a schematic diagram of the process of the present invention. DETAILED DESCRIPTION

[0019] The application will be further described below in conjunction with the accompanying drawings and examples. Example One

[0020] An unmanned aerial vehicle-based intelligent thermal power plant inspection method, as shown in the figure, in order to overcome the problem of needing to set up a scaffold or having a worker climb a stair platform to inspect the thermal power plant in the prior art, a unmanned aerial vehicle-based intelligent thermal power plant inspection method is designed. In this embodiment, the unmanned aerial vehicle cluster carries out an inspection task on the thermal power plant according to a flight task. Figure 1

[0021] The unmanned aerial vehicle cluster orderly carries out a comprehensive inspection of the entire thermal power plant according to a pre-planned inspection route and task allocation.

[0022] In the unmanned aerial vehicle-based intelligent thermal power plant inspection method, a binocular vision technology is used in combination with a SLAM algorithm to establish a three-dimensional model, wherein the binocular vision technology simulates human binocular vision by using two cameras, calculates the distance and position information of objects in the scene by comparing the images captured by the two cameras, and thus perceives structural defects on the thermal power plant; at the same time, the SLAM algorithm can enable the unmanned aerial vehicle to position and construct a three-dimensional model of the thermal power plant environment while moving; by combining the binocular vision technology and the SLAM algorithm, the intelligent thermal power plant inspection system constructs a three-dimensional model of the thermal power plant in real time, the unmanned aerial vehicle avoids obstacles according to the three-dimensional model of the thermal power plant constructed in real time, and flies along the predetermined inspection route after avoiding the obstacles, or adjusts the flight path according to real-time data after avoiding the obstacles to better inspect the facility; at the same time, the three-dimensional model of the thermal power plant can also be used to identify defects, wear and tear or other problems requiring maintenance on the structure of the thermal power plant; finally, the worker checks the monitoring data of the thermal power plant transmitted by the unmanned aerial vehicle in real time through the ground control station, in addition, the worker can also check the three-dimensional model of the thermal power plant by wearing an AR device, more accurately find possible problems on the facility, and determine whether the problem on the facility is a mistake of the unmanned aerial vehicle or the problem on the facility does not affect the safe operation of the facility.

[0023] Using the above method can greatly reduce the risk of high-altitude work of the worker, improve the safety during the inspection operation, at the same time, the unmanned aerial vehicle can work continuously and is not limited by the physical strength and working time of the human worker, and using the binocular vision in combination with the SLAM algorithm can construct an accurate three-dimensional model of the thermal power plant, which is helpful to more accurately identify structural defects, wear and tear or other problems.

[0024] ​In the embodiment, the unmanned aerial vehicle identifies structural defects existing in the thermal power generation facility; and monitors the operation state and temperature distribution of the thermal power generation facility through the sensor.

[0025] The unmanned aerial vehicle identifies common structural defects through deep learning technology, mainly through a defect identification model constructed based on a target detection algorithm Faster RCNN to help the unmanned aerial vehicle identify various common structural defects, including cracks, corrosion and wear and tear and the like; wherein the specific steps of training the defect identification model to identify common structural defects are as follows: S111: First, collect image data sets for training and verification, which contain structural defect images of the thermal power generation facility; while collecting the image data sets, label each image in the image data sets, that is, draw a bounding box on each defect contained in the image and mark the corresponding defect category; then divide the images marked with the defect category into a training set, a verification set and a test set for subsequent analysis and processing.

[0026] S112: Perform preprocessing operations on the image data set, including adjusting the image size, normalizing the image and the like to adapt to the data input requirements of the defect identification model; at the same time, perform rotation, flipping, scaling and cropping operations on the image to increase the diversity of the data.

[0027] S113: Then use the training set to train the defect identification model; the training steps are as follows: first, use the CNN model in the defect identification model to extract the features in the picture; then generate candidate regions through the RPN in the defect identification model, wherein these candidate regions may contain structural defects such as cracks, corrosion and wear and tear; then perform RoI pooling on each candidate region to obtain a fixed-size feature block; perform classification and bounding box regression operations on each feature block, that is, determine whether it is a structural defect and determine the exact position of the defect.

[0028] S114: Then use the verification set to evaluate the defect identification model, and according to the evaluation result, fine-tune the defect identification model, wherein the evaluation result is tested using accuracy, recall rate and F1 score; wherein the calculation method of the accuracy is as follows: Precision = Correctly identified image number / Total number of images identified × 100%, The calculation method of the recall rate is as follows: Recall = Correctly identified image / Total number of images × 100%, The calculation method of the F1 score is as follows: 2 × Precision × Recall / (Precision + Recall) × 100%.

[0029] S115: Finally, evaluate the final performance of the defect recognition model on the test set and deploy the defect recognition model to the unmanned aerial vehicle control platform to perform actual fire power generation facility inspection tasks. The unmanned aerial vehicle trained using the Faster R-CNN algorithm can automatically identify structural defects of the fire power generation facility, reduce manual participation, and improve inspection efficiency and accuracy; compared with traditional manual detection or simple automation methods, the use of the Faster R-CNN algorithm realizes higher levels of automation and intelligence.

[0030] In this embodiment, the stereo image provided by the binocular vision technology can help the defect recognition model to more accurately judge the size and shape of the defect, accurately locate the position of the problem in the three-dimensional space, and better understand the geometric structure of the fire power generation facility with the help of the three-dimensional model, while the target detection algorithm can optimize and correct the depth information of the fire power generation facility obtained by the binocular vision technology, reduce the influence of environmental factors such as light changes or occlusions on the monitoring accuracy, and thus realize more accurate defect depth estimation and improve the detection accuracy. Through the combination of binocular vision technology and target detection algorithm, the three-dimensional detection capability and accuracy of the unmanned aerial vehicle in inspecting the fire power generation facility are improved.

[0031] When the unmanned aerial vehicle identifies structural defects, the binocular vision technology is mounted on the unmanned aerial vehicle and combined with the SLAM algorithm, so that the unmanned aerial vehicle can capture and process complex visual information in real time during flight. The unmanned aerial vehicle constructs a high-precision three-dimensional model of the fire power generation facility in real time according to the depth information and three-dimensional structure of the environment obtained by the binocular vision technology and the SLAM algorithm, and the depth information can be directly calculated from the binocular image by triangulation.

[0032] Specifically, the depth information calculated by the binocular vision technology is combined with the pose of the unmanned aerial vehicle to generate a dense three-dimensional point cloud, which represents the three-dimensional structure of the surrounding environment. Then, the three-dimensional point cloud is filtered, denoised and optimized to eliminate redundant data and measurement errors and improve the accuracy of the three-dimensional model. Then, the point cloud data is converted into a more detailed and smoother three-dimensional model of the fire power generation facility using three-dimensional reconstruction techniques such as mesh generation. Based on the constructed three-dimensional model of the fire power generation facility, the unmanned aerial vehicle can accurately perceive the surrounding environment, identify and avoid obstacles in real time through the path optimization algorithm deployed in the unmanned aerial vehicle, ensure that the unmanned aerial vehicle flies along the predetermined inspection route, or conduct more detailed inspection of the fire power generation facility according to the three-dimensional model. Then, as the unmanned aerial vehicle continues to fly and collect more data, the SLAM algorithm continuously updates and optimizes the three-dimensional model to ensure the real-time and accuracy of the model. The unmanned aerial vehicle maintains dynamic perception and adjustment of the environment during the entire inspection process to ensure that the constructed three-dimensional model accurately reflects the actual state of the fire power generation facility In the embodiment, the unmanned aerial vehicle automatically adjusts the subsequent flight plan and inspection route according to the comprehensive analysis result.

[0033] The unmanned aerial vehicle obtains the detection result of the thermal power plant through the three-dimensional model and the target detection algorithm. The unmanned aerial vehicle continuously optimizes the inspection strategy according to the detection result and the autonomous decision-making ability of the unmanned aerial vehicle, identifies which area is more likely to have problems, and the event regularity of the problem, and actively adjusts the inspection frequency and flight path of the unmanned aerial vehicle in a certain area and preferentially checks the high-risk area in the thermal power plant.

[0034] In the embodiment, the unmanned aerial vehicle cluster realizes information sharing and task coordination through data, shares the monitoring data and flight state information, and realizes the overall subsequent flight plan and inspection route of the unmanned aerial vehicle cluster.

[0035] Specifically, when a problem is found in a certain area of the thermal power plant by an unmanned aerial vehicle, the unmanned aerial vehicle not only automatically optimizes the inspection strategy and increases the inspection frequency in the area, but also realizes information sharing and task coordination with the unmanned aerial vehicle cluster in the thermal power plant in real time, and cooperatively performs more detailed inspection tasks. The unmanned aerial vehicle first compares and analyzes the structure defect identification result of the thermal power plant in the area with the preset defect normal range threshold, judges whether the operation state of the thermal power plant in the area is abnormal, and when the operation state of the thermal power plant in the area is outside the normal range, i.e. the thermal power plant may have defect problems or abnormal conditions affecting production safety, the unmanned aerial vehicle sends a cooperation request to the unmanned aerial vehicle cluster, increases the number of inspection unmanned aerial vehicles, and finds the optimal flight path of the unmanned aerial vehicle cluster in the area through a preset algorithm, i.e. a particle swarm optimization algorithm, to further check and judge the area in detail; when the operation state of the thermal power plant in the area is within the normal range threshold, the unmanned aerial vehicle performs the next inspection according to the predetermined flight task.

[0036] Specifically, the unmanned aerial vehicle cluster in the thermal power plant is connected through real-time communication between the unmanned aerial vehicles, forming an efficient information sharing network. When any unmanned aerial vehicle in the unmanned aerial vehicle cluster discovers a potential problem or abnormal condition in the inspection process, including but not limited to temperature reading abnormality, structural damage, leakage or other safety hazards, other members in the unmanned aerial vehicle cluster can quickly obtain the potential problem or abnormal condition discovered in the area, and according to the severity, the unmanned aerial vehicles in the unmanned aerial vehicle cluster automatically adjust the inspection strategy to more concentratedly and thoroughly check the area where the problem is found, and through the use of different types of sensors including high-resolution photography, spectral scanning or thermal imaging, etc., more detailed inspection or more comprehensive and multi-angle data collection is realized.

[0037] Specifically, the drone automatically adjusts its inspection strategy through its own particle swarm optimization algorithm. At the same time, the particle swarm optimization algorithm can also be used to dynamically adjust the drone's flight path to cope with sudden obstacles, emergencies or environmental changes. First, the drone collects environmental data in real time during flight, including obstacles, wind speed, temperature and other data; then, through the fitness function defined for the drone, the quality of the current flight path is evaluated based on real-time data; the drone uses the particle swarm optimization algorithm to update the speed and position of each particle in real time according to the feedback of the fitness function, that is, to adjust the flight direction and distance. The idea of ​​the particle swarm optimization algorithm originated from the study of bird foraging behavior. Each particle, that is, the drone, has its corresponding position and speed. The position of the particle is the coordinate of the drone at a certain moment, and the speed is the movement trend from one position to another. Through continuous iteration, the particle swarm gradually converges to the optimal flight path solution; when the drone When a drone detects an obstacle or a problem in a certain area, it immediately feeds its location information back to the particle swarm optimization algorithm, which then replans a new flight path and uses it as a potential flight route. When a swarm of drones conducts simultaneous inspections, they share real-time data via communication equipment, and the particle swarm optimization algorithm coordinates the flight paths between the drones to avoid interference and improve inspection efficiency. Finally, the new flight paths adjusted in real time by the particle swarm optimization algorithm are adopted by the drones, which continue to collect data during the inspection and feed this data back into the particle swarm optimization algorithm, forming a closed feedback control loop that ensures the drones can continuously optimize their paths. The particle swarm optimization algorithm can provide an effective real-time path adjustment mechanism for drones, enhancing their autonomous flight capabilities within complex paths like those in thermal power plants and improving the safety and efficiency of inspection missions.

[0038] When the UAV plans the flight path through the particle swarm optimization algorithm, the calculation formula for the speed of the UAV at the next time node is:

[0039] The calculation formula for the position of the drone at the next time node is: The first formula is the speed update formula, and the second formula is the position update formula, which is calculated based on the new speed; It is a particle in the particle swarm optimization algorithm, that is, a drone in the drone swarm; It's a drone Speed ​​at the next time point; is the inertia weight, which determines the tendency of the drone to maintain its current speed; It's a drone Speed ​​at the current time point; and is the learning factor, which determines the degree to which the UAV moves towards its optimal position and the global optimal position; and is a random number in the range of [0,1], which is the randomness introduced during the drone search process; It's a drone The best position for drones The optimal solution found in historical data; is the global optimal position, i.e., the optimal solution found historically by all drones; It's a drone The position at the current time node; It's a drone The position of the drone at the next time node; the position of the drone is a point in the search space, and the speed represents the direction and size of the drone's movement in the search space. By iteratively calculating the position and speed of the drone, the drone can find the optimal path in the search space. The calculation formula for fitness is: , in is the fitness function, which is used to evaluate the inspection path of the UAV; is the drone inspection coverage, that is, the regional coverage of the drone inspection; is the path length, i.e. the total length of the drone inspection path; Safety, that is, the ability of the drone to avoid hitting obstacles during inspections; is the energy consumption, i.e. the amount of electricity consumed by the drone to complete the inspection task; 、 、 、 is the weight coefficient used to balance different optimization objectives.

[0040] In this embodiment, the drone cluster transmits monitoring data to a ground control station, which combines the monitoring data with a three-dimensional model of the thermal power generation facility to clearly display problems existing in the thermal power generation facility.

[0041] In the process of the unmanned aerial vehicle inspecting the thermal power plant, the unmanned aerial vehicle can feed back data and analysis results to the ground control center in real time. The ground control center can integrate augmented reality (AR) technology into smart glasses or a mobile device, and the smart glasses or the mobile device can receive data and analysis results of the ground control center through the augmented reality interface. In this way, an operator can see superimposed real-time images and a three-dimensional model of a scene returned by the unmanned aerial vehicle in a field of view of the real world. The superimposed real-time images and the three-dimensional model of the scene can include effects achieved by superimposed display of damage positions, abnormal hot spots, or other detected defects. In this way, the operator can visually and more accurately locate problems of the thermal power plant. Meanwhile, the three-dimensional model can fuse data of other sensors such as a thermal imager and a laser scanner, and provide more comprehensive information for the operator, so as to help the operator better understand real-time conditions of the thermal power plant and how to effectively maintain and repair the thermal power plant. The operator can also interact with the three-dimensional model through a touch screen or a voice command. For example, the operator can zoom in, rotate, or switch to a detailed view of a specific part of the three-dimensional model. The data is displayed in the form of the three-dimensional model by using the augmented reality technology, and the operator can interact with the three-dimensional model in real time. This is easier for the operator to understand and analyze, and improves a higher level of user experience and accuracy. The data from different sensors is fused to provide more comprehensive decision support for the operator, and improve the quality of maintenance and repair of the thermal power plant.

[0042] In the embodiment, in order to solve the problem of charging of the unmanned aerial vehicle in the thermal power plant, a plurality of unmanned aerial vehicle autonomous charging stations are further arranged in the thermal power plant. The autonomous charging stations are arranged on an inspection route of the unmanned aerial vehicle, and are evenly distributed in the thermal power plant. The distance from each autonomous charging station to a nearby autonomous charging station is the same. The autonomous charging stations adopt wireless charging technology, so that the unmanned aerial vehicle can be inductively charged in flight or above a parking apron, and energy consumption caused by landing or taking off is reduced.

[0043] The unmanned aerial vehicle is provided with a minimum charging amount, which can be set to 5%. When the amount of electricity of a certain unmanned aerial vehicle is lower than the minimum charging amount, an energy management system calculates a position of the nearest autonomous charging station to the unmanned aerial vehicle. Whether the autonomous charging station is worth waiting for other unmanned aerial vehicles to be fully charged is determined by monitoring whether the number of unmanned aerial vehicles being charged in the autonomous charging station exceeds a threshold and a task state of the unmanned aerial vehicle. If the determination result is yes, the autonomous charging station is selected as a preferred charging station. If the determination result is no, the next autonomous charging station is selected for next determination. By using the unmanned aerial vehicle autonomous charging and the energy management system, the inspection efficiency of the unmanned aerial vehicle is improved, the safety in the inspection process is ensured, manual intervention is reduced, and the continuous operation capability of the unmanned aerial vehicle is enhanced. Embodiment Two

[0044] The unmanned aerial vehicle-based intelligent firepower power generation facility inspection system comprises an unmanned aerial vehicle, a binocular vision camera, a sensor, a communication device, a flight management module, an autonomous charging station, an energy management module, a ground control station and an augmented reality (AR) interface.

[0045] The unmanned aerial vehicle is responsible for inspecting the firepower power generation facility, and multiple unmanned aerial vehicles form an unmanned aerial vehicle cluster to realize accurate and multi-directional detection of the firepower power generation facility. The binocular vision camera is used in combination with a SLAM algorithm to process image data acquired by the unmanned aerial vehicle in real time to construct an accurate three-dimensional model of the firepower power generation facility. The other sensors include a thermal imager and a laser scanner and the like to monitor the temperature distribution of the firepower power generation facility and the operation state of the firepower power generation facility. The communication device is responsible for real-time data transmission between the unmanned aerial vehicle cluster and the ground control station. The flight management module is used to plan and adjust the flight plan of the unmanned aerial vehicle, can automatically adjust the flight route and task priority of the unmanned aerial vehicle according to real-time inspection data, and realizes autonomous planning of the flight route of the unmanned aerial vehicle by carrying a particle swarm optimization algorithm. The autonomous charging station is an autonomous charging facility of the unmanned aerial vehicle, and the unmanned aerial vehicle can automatically return to the charging station for autonomous charging when the power is low to ensure the continuity of the inspection task, and has wireless charging capability to reduce the time required for the unmanned aerial vehicle to dock and charge. The energy management module judges the preferred autonomous charging station for charging of the unmanned aerial vehicle and adjusts the charging strategy according to the environmental conditions of the unmanned aerial vehicle. The ground control station is used for ground operators to monitor the state of the unmanned aerial vehicle, plan flight tasks, view inspection data, and construct and view three-dimensional models. The augmented reality (AR) interface is used to combine the real-time image returned by the unmanned aerial vehicle with the three-dimensional model of the firepower power generation facility, realize instant visualization of the state of the firepower power generation facility through smart glasses or other display devices, and enable the operator to more intuitively acquire problems existing in the firepower power generation facility and the priority of the problems.

[0046] Compared with the prior art, the unmanned aerial vehicle-based intelligent firepower power generation facility inspection system provides higher automation level, better data collection and analysis capability and higher operation efficiency.

[0047] The above-described embodiments only express the preferred embodiments of the present application, and the description is more specific and detailed, but it cannot be understood as a limitation on the scope of the patent of the present application. It should be noted that for ordinary skilled persons in the art, without departing from the concept of the present application, a number of modifications, improvements and substitutions can be made, which all belong to the protection scope of the present application. Therefore, the protection scope of the patent of the present application should be subject to the appended claims.

Claims

1. An intelligent inspection method for thermal power generation facilities based on drones, characterized in that: The specific implementation is as follows: S101: The drone swarm conducts inspections of thermal power generation facilities based on flight missions; S102: UAVs identify structural defects in thermal power generation facilities and use sensors to monitor the operating status and temperature distribution of thermal power generation facilities. S103: The drone automatically adjusts subsequent flight plans and inspection routes based on the comprehensive analysis results; S104: At the same time, the drone clusters share information and coordinate tasks through data communication, share monitoring data and flight status information, and adjust the overall subsequent flight plan and inspection route of the drone cluster; S105: The drone cluster transmits the monitoring data to the ground control station, which combines the monitoring data with the three-dimensional model of the thermal power generation facility to clearly display the problems existing in the thermal power generation facility.

2. The intelligent inspection method for thermal power generation facilities based on drones according to claim 1 is characterized in that: When the drone identifies structural defects, by equipping the drone with binocular vision technology and combining it with the SLAM algorithm, the drone can capture and process complex visual information in real time during flight. The drone constructs a high-precision three-dimensional model of the thermal power generation facility in real time based on the depth information and three-dimensional structure of the environment obtained through the binocular vision technology and the SLAM algorithm. Specifically, the depth information calculated by the binocular vision technology is combined with the posture of the drone to generate a dense three-dimensional point cloud, which represents the three-dimensional structure of the surrounding environment. The three-dimensional point cloud is then filtered, denoised, and optimized to eliminate redundant data and measurement errors. The three-dimensional reconstruction technology is then used to convert the point cloud data into a three-dimensional model of the thermal power generation facility. Based on the constructed three-dimensional model of the thermal power generation facility, the drone can accurately perceive the surrounding environment and identify and avoid obstacles in real time through the path optimization algorithm deployed in the drone, ensuring that the drone flies along the predetermined inspection route, or performs a more detailed inspection of the thermal power generation facility based on the three-dimensional model.

3. The intelligent inspection method for thermal power generation facilities based on drones according to claim 2 is characterized in that: The drone uses a target detection algorithm to determine the size and shape of defects in the stereo image provided by binocular vision technology, and takes advantage of the deep learning algorithm to optimize and correct the depth information of the thermal power generation facility obtained by binocular vision technology, and identify cracks, corrosion and wear defects on the surface of the facility or the obscured parts under different lighting conditions. At the same time, various sensors are used to monitor the thermal power generation facility, including the operating status and temperature of the thermal power generation facility.

4. The intelligent inspection method for thermal power generation facilities based on drones according to claim 1 is characterized in that: When the drone identifies structural defects in thermal power generation facilities, it uses a defect recognition model pre-deployed on the drone. The defect recognition model is constructed using the target detection algorithm Faster RCNN. The structural defects include cracks, corrosion, and wear. The specific steps for training the defect recognition model to identify common structural defects in thermal power generation facilities are as follows: S111: First, an image dataset for training and validation is collected, wherein the image dataset includes structural defects of thermal power generation facilities; images in the image dataset are annotated, i.e., a bounding box is drawn for each defect in the image and the category of the defect is marked; then, the images with the category marked are classified into a training set, a validation set, and a test set; S112: performing a preprocessing operation on the image dataset, wherein the preprocessing operation includes adjusting the image size and normalizing the image; and performing rotation, flipping, scaling, and cropping operations on the category-labeled images in the image dataset to increase data diversity; S113: Pre-training the defect recognition model using a training set to determine whether each feature block in the image is a structural defect and to determine the precise location of the defect; S114: Evaluate the pre-trained defect recognition model using the validation set, and fine-tune the defect recognition model based on the evaluation results. The evaluation includes accuracy, recall, and F1 score. The accuracy is calculated as follows: Precision = number of correctly identified images / total number of identified images × 100%, The recall rate is calculated as: Recall = correctly identified images / total number of images × 100%, The F1 score is calculated as: 2× Precision× Recall / ( Precision+ Recall)×100%; S115: Finally, the final performance of the defect recognition model is evaluated using the test set to determine whether the result meets the expected standard. If not, execute step S101; if yes, deploy the defect recognition model to the UAV control platform to perform actual thermal power generation facility inspection tasks to identify structural defects of thermal power generation facilities.

5. The intelligent inspection method for thermal power generation facilities based on drones according to claim 1 is characterized in that: The drone automatically adjusts subsequent flight plans based on the comprehensive analysis results. When a problem is detected in one area of ​​the thermal power plant, the drone compares the structural defect identification results of the monitored thermal power facilities with the preset normal range threshold to determine whether there is any abnormality in the operating status of the thermal power facilities in the area. When it is determined that the operating status of the facilities in the area is outside the normal range threshold, the drone sends a cooperation request to the drone cluster, increases the number of drones patrolling the area, and uses a preset algorithm to find the optimal flight path for the drone cluster in the area; when it is determined that the operating status of the facilities in the area is within the normal range threshold, the drone continues to patrol according to the scheduled flight mission.

6. The intelligent inspection method for thermal power generation facilities based on drones according to claim 5 is characterized in that: The drone uses the particle swarm optimization algorithm to find the optimal flight path and plan the drone's flight path. When the drone detects a problem in one area of ​​the thermal power plant, it not only increases the inspection frequency of the area, but also conducts data communication in real time through the drone cluster to achieve information sharing and task coordination, and initiates assistance requests to other drones in the drone cluster to jointly perform more complex inspection tasks, conduct more detailed inspections, or collect data from more angles. When the drone plans the flight path using the particle swarm optimization algorithm, the formula for calculating the drone's speed at the next time node is: ； The calculation formula for the position of the drone at the next time node is: ； in, is a particle, i.e., a drone in the drone cluster; It's a drone Speed ​​at the next time point; is the inertia weight, which determines the tendency of the drone to maintain its current speed; It's a drone Speed ​​at the current time point; and is the learning factor, which determines the degree to which the UAV moves towards its optimal position and the global optimal position; and is a random number in the range of [0,1], which is the randomness introduced during the drone search process; It's a drone The best position for drones The optimal solution found in historical data; is the global best position, i.e. the best solution found by all particles in history; It's a drone The position at the current time node; It's a drone The position of the drone at the next time node; the position of the drone is regarded as a point in the search space, and the speed is regarded as the moving direction and size of the drone in the search space. The optimal path of the drone is obtained by iteratively calculating the position and speed.

7. The intelligent inspection method for thermal power generation facilities based on drones according to claim 1 is characterized in that: When the ground control station clearly displays the problems existing in the thermal power generation facilities, the ground operators view the images sent back by the drone in real time through smart glasses or mobile devices, and superimpose them with the three-dimensional model to clearly view the problems existing in the thermal power generation facilities and further confirm the problems; the three-dimensional model uses a touch screen or voice commands to realize operations including zooming in, zooming out, rotating, and switching to a view at a specified position to display the defects in the thermal power generation facilities; the three-dimensional model can superimpose the operating status and temperature of the thermal power generation facilities monitored by sensors to the corresponding positions of the three-dimensional model.

8. The intelligent inspection method for thermal power generation facilities based on drones according to claim 1 is characterized in that: There are also multiple autonomous charging stations in the thermal power plant. The autonomous charging stations are evenly distributed in the thermal power plant, that is, the distance from one of the autonomous charging stations to the adjacent autonomous charging station is the same. At the same time, drone autonomous charging equipment is installed inside the autonomous charging stations for automatic charging of drones.

9. The intelligent inspection method for thermal power generation facilities based on drones according to claim 8, characterized in that: The drone is set with a minimum charging power level. That is, when the drone's own power level approaches the minimum charging power level, the drone automatically plans a flight path from its current location to the nearest autonomous charging station. At the same time, the drone also has a self-diagnosis function and automatically returns to the maintenance center for repair when it detects that it needs maintenance or the equipment is damaged.

10. An intelligent inspection system for thermal power generation facilities based on drones, characterized in that: It includes a flight management module, an autonomous charging station, an energy management module, a ground control station, and an augmented reality (AR) interface. The flight management module is used to plan and adjust the UAV flight plan, automatically adjusting the UAV's flight route and mission priority based on real-time inspection data; The autonomous charging station is an autonomous charging facility for the drone. When the drone is low on power, it will automatically return to the autonomous charging station for autonomous charging to ensure the continuity of the inspection mission. The energy management module is used to determine the preferred autonomous charging station for charging the drone; The ground control station is used to monitor the status of the drone, plan flight missions, view inspection data, and build and view three-dimensional models; The augmented reality (AR) interface is used to combine real-time images transmitted by drones with three-dimensional models of thermal power generation facilities, visually displaying problems within the thermal power generation facilities through smart glasses or display devices.

Images