Power distribution network operation inspection method and system based on cooperation of inspection unmanned aerial vehicle and robot

Abstract

The invention relates to the technical field of power distribution network operation inspection, and discloses a power distribution network operation inspection method and system based on cooperation of an inspection unmanned aerial vehicle and a robot, and the method comprises the steps: obtaining an inspection network diagram of an area where a power distribution network needing operation inspection is located; constructing a routing inspection route model considering routing inspection time and routing inspection cost based on the routing inspection network diagram; solving the inspection route model through a parameter optimization algorithm so as to determine an optimal unmanned aerial vehicle placement point, an optimal unmanned aerial vehicle inspection route, an optimal robot placement point and an optimal robot inspection route from the inspection network diagram; performing inspection according to the optimal unmanned aerial vehicle inspection route and the optimal robot inspection route so as to acquire images of the power distribution network; performing data fusion on the collected initial high-altitude image and initial low-altitude image and the corresponding time information and geographic position information to obtain a fusion result; judging whether the running state of the power distribution network is abnormal or not; and performing the next inspection according to the set inspection frequency. According to the invention, the comprehensiveness of routing inspection is ensured.

Description

Technical Field

[0001] The present invention relates to the technical field of distribution network operation and inspection, and in particular to a distribution network operation and inspection method and system based on the collaboration of a patrol drone and a robot. Background Art

[0002] The distribution network is composed of overhead lines, cables, poles, distribution transformers, disconnectors, reactive power compensators and some ancillary facilities. It plays an important role in distributing electric energy in the power grid.

[0003] Distribution network operation and maintenance (O&M) involves inspecting, maintaining, and repairing distribution equipment within a power system to ensure power quality and stability. With the development of intelligent technologies, intelligent applications are becoming increasingly important in distribution network O&M. However, some regions still rely primarily on manual inspections and lack the application of intelligent technologies, resulting in low O&M efficiency and quality. This approach also significantly increases labor costs and safety risks. Summary of the Invention

[0004] In order to solve the problem mentioned in the above background technology that the lack of application of intelligent technology leads to low efficiency and quality of operation and maintenance work, and also significantly increases the labor cost and safety risks of operation and maintenance work, the present invention proposes a distribution network operation and inspection method and system based on the collaboration of inspection drones and robots to solve this problem.

[0005] The present invention is achieved through the following technical solutions: A distribution network operation and inspection method based on collaboration between inspection drones and robots, comprising: Obtain the inspection network diagram of the area where the distribution network that needs to be inspected is located; Based on the inspection network diagram, an inspection route model is constructed that takes into account inspection time, inspection cost, load balancing, synchronization time difference, and path conflict risk; Solving the inspection route model through a parameter optimization algorithm, thereby determining an optimal inspection plan from the inspection network diagram; Performing an inspection based on the optimal inspection plan to collect an image of the distribution network; Performing data fusion on the initial high-altitude images and initial low-altitude images collected by the drone and the robot, as well as the corresponding time information and geographic location information, to obtain a fusion result, and transmitting the fusion result to the terminal; Determine whether the operating status of the distribution network is abnormal based on the fusion result. If so, generate an inspection warning result and continue the inspection. Otherwise, continue the inspection until one round of inspection is completed. Perform the next inspection according to the set inspection frequency.

[0006] As an optimization, the specific steps to obtain the inspection network diagram of the area where the distribution network that needs to be inspected is located are: Obtaining basic power grid data of the area where the distribution network is located, the first basic power grid data including a first tower coordinate set, a first conductor direction set, and a first substation location set; Acquire aerial survey information of the area where the distribution network is located, the aerial survey information including a second tower coordinate set, a second conductor direction set, and a second substation location set; Determine the altitude of the drone’s flight path; Marking obstacles in the distribution network area based on the aerial survey information; Converting the data in the first tower coordinate set, the first conductor orientation set, the first substation location set, the second tower coordinate set, the second conductor orientation set, and the second substation location set into the same coordinate system, and using GCP to correct the deviation between the aerial survey information and the basic power grid data, thereby obtaining a final tower coordinate set, a final conductor orientation set, and a final substation location set; An inspection network diagram is formed based on the distribution network area, obstacles, final tower coordinate set, final conductor direction set and final substation location set.

[0007] As an optimization, the inspection route model includes an objective function and constraints, wherein the objective function is: ; ; ; ; ; in, 、 、 、 is the weight of the corresponding coefficient, and , 、 、 、 There are 5 sub-goals: It is the normalized reference value of the total inspection time, used to Map to The interval, eliminating the dimension effect, can be a longer time limit value calculated based on experience or theory. It is the normalized reference value of the total inspection distance. A possible maximum distance value is set according to the scenario. It is the normalized reference value of the load variance. A larger variance limit value can be set according to the actual situation. It is the normalized reference value of the synchronization time difference and can be set as a maximum allowed time difference; , represents the total time of the drone moving between the drone placement points, m is the number of drones, is the number of stages of the moving path of UAV u, represents the moving time of the u-th UAV in the i-th stage of the moving path; , represents the total movement time of the robot between the robot placement points, n is the number of robots, is the number of stages of the movement path of robot r, represents the moving time of the rth robot in the jth stage of the moving path; , represents the total inspection time of the areas corresponding to all placement points, is the inspection time of the area corresponding to the placement point p; represents the set of drone placement points, Represents the set of robot placement points , represents the total distance the drone moves, represents the moving distance of the u-th UAV in the i-th stage of the moving path; , represents the total distance the robot moves, represents the moving distance of the rth UAV in the jth stage of the moving path; , represents the UAV load variance, , Represents the load of UAV u. The load here refers to the sum of the relevant tasks such as the movement time and operation time of the UAV in the inspection task it is responsible for. It is used to measure the overall scale of the tasks undertaken by the UAV. Indicates the drone placement point that drone u reaches in the i-th stage of the movement path The inspection operation time of the corresponding area, that is, the time it takes for the drone to perform the inspection task at the placement point, Indicates the average load of the drone equipment; , represents the robot load variance, , Represents the load of robot r. The load here refers to the sum of the robot's movement time, operation time and other related tasks in the inspection task it is responsible for. It is used to measure the overall scale of the robot's tasks. Indicates the robot placement point reached by robot r in the jth stage of the moving path The inspection operation time of the corresponding area, that is, the time the robot spends performing the inspection task at the placement point, Indicates the average load of the robot device; S represents the set of nodes that are visited by both drones and robots. represents the time when the UAV arrives at node s, represents the time when the robot arrives at node s; The constraints include safety constraints, regulatory constraints, environmental constraints, technical constraints, node regression constraints, and inspection time interval constraints for drones and robots at the same geographical location.

[0008] As an optimization, the inspection route model is solved by a parameter optimization algorithm, specifically using a multi-objective genetic algorithm based on NSGA-II. The specific process is: Randomly generate the number m of drones, the number n of robots, and the corresponding paths to form an initialization population, wherein the initialization population includes a number of individuals, each of which includes the number of drones, the number of robots, and the corresponding path; Calculate the function values ​​of the four sub-goals of all individuals to obtain the specific values ​​of each individual in these four dimensions; Perform non-dominated sorting + crowding calculation: Non-dominated sorting: For individuals X and Y in the population, if: (1) the target values ​​of X are not worse than those of Y; (2) at least one target value of X is better than that of Y, then X can dominate Y; first find all individuals in the population that are not dominated by other individuals, classify them into the first layer (non-dominated layer), and assign them the same non-dominated order; then, remove these individuals from the population, repeat the above operation for the remaining individuals, find the non-dominated individuals in the second layer, and so on, until all individuals in the population are stratified; in this way, each individual has its own non-dominated order, reflecting its superiority and inferiority level in the population; crowding calculation: in the same non-dominated layer, construct a rectangle with the individual's nearest neighbor as the vertex. The crowding of an individual is related to the area of ​​the rectangle. The larger the area, the less crowded the individual. By calculating the crowding, the superiority and inferiority of individuals in the same non-dominated layer can be further distinguished, avoiding the concentration of individuals in certain areas of the Pareto front, and maintaining the diversity of the solution set; The parent population is selected through tournament selection: the number of individuals N to be compared is determined, and N individuals are randomly selected from the population, with each individual having equal probability of being selected. The fitness of these N individuals is then compared (this combines non-dominated order and crowding, with individuals with smaller non-dominated order being prioritized; if the non-dominated order is the same, individuals with smaller crowding are prioritized). The individual with the best fitness is selected to enter the parent population, and this process is repeated multiple times until the parent population reaches a predetermined size. Merge the parent and offspring generations based on the elite strategy, retaining the top M excellent individuals: merge the parent population and offspring population together, then sort them according to the non-dominated order and crowding degree of the individuals, give priority to individuals with smaller non-dominated order, and for individuals in the same non-dominated layer, select individuals with smaller crowding degree. Select the top N excellent individuals from the merged population to form a new parent population and enter the next round of iteration; Repeat the above steps until the maximum number of iterations is reached or convergence occurs, and the placement point found is the optimal placement point.

[0009] As an optimization, if the drone is a charging drone, the drone placed at the optimal drone placement point ; If the drone is a battery replacement drone, the number of drones placed at the optimal drone placement point is 1 or 2; If the robot is a charging robot, the robot placed at the optimal robot placement point ; If the robot is a battery replacement robot, the number of robots placed at the optimal robot placement point is 1 or 2.

[0010] As an optimization, the initial high-altitude images and initial low-altitude images collected by the UAV and the robot, as well as the corresponding time information and geographic location information, are fused to obtain a fusion result, and the specific process of transmitting the fusion result to the terminal is as follows: Performing radiation correction and geometric correction on the initial high-altitude image and the initial low-altitude image, respectively, and then performing contrast enhancement, brightness adjustment, and denoising processing on the initial high-altitude image and the initial low-altitude image, respectively, to obtain a first intermediate high-altitude image and a first intermediate low-altitude image; Performing feature extraction on the first intermediate high-altitude image and the first intermediate low-altitude image corresponding to the same geographic location information of the same inspection batch to obtain corresponding feature points, and matching the feature points of the first intermediate high-altitude image with the feature points of the first intermediate low-altitude image to achieve alignment of the first intermediate high-altitude image and the first intermediate low-altitude image; Inputting the aligned first intermediate high-altitude image and the first intermediate low-altitude image into a neural network model for image fusion to obtain an initial fused image; Post-processing and optimizing the initial fused image to obtain a fusion result; The fusion result is transmitted to the terminal.

[0011] As an optimization, the specific process of post-processing and optimizing the initial fused image to obtain the fusion result is as follows: The initial fused image is deblurred, and the fusion parameters are adjusted through cross-validation, visual evaluation or field verification to optimize the stitching effect, so as to achieve accuracy evaluation and adjustment and finally obtain the fusion result.

[0012] As an optimization, the specific process of judging whether the operating state of the distribution network is abnormal according to the fusion result is as follows: Building a monitoring database, wherein the monitoring database is provided with a fault image related to a distribution network fault; The fusion result is matched with several fault images. If there is a matching fault image, the fault type of the fault image is used as the fault type of the fusion result, and an alarm message is recorded and issued; if there is no matching fault image, the fusion result is deleted.

[0013] The present invention discloses a distribution network operation and inspection system based on the collaboration of inspection drones and robots, which is used to implement the aforementioned distribution network operation and inspection method based on the collaboration of inspection drones and robots, including: Inspection system is used to realize the operation and inspection of distribution network, including: The inspection system includes an unmanned aerial vehicle system, a robot system, a wireless communication system, a collaborative system, and a positioning system. The unmanned aerial vehicle system is used to inspect high-altitude distribution networks, the robot system is used to inspect low-altitude distribution networks, the wireless communication system is used to connect the unmanned aerial vehicle system, the robot system, and the remote control system, the collaborative system is used to synchronize the operation of the unmanned aerial vehicle system and the robot system, and the positioning system is used to locate the positions of the unmanned aerial vehicle system and the robot system. A remote control system for remotely controlling the inspection system; The data storage system is used to store the operation and inspection data of the inspection system.

[0014] As an optimization, the drone system and the robot system are also equipped with an obstacle avoidance unit, which calculates the shortest distance between the obstacle avoidance unit and the obstacle through a parameter optimization algorithm based on the travel speed, reaction speed, relative angle between the travel direction of the drone / robot and the wind direction, and wind speed of the drone / robot.

[0015] Compared with the prior art, the present invention has the following advantages and beneficial effects: 1. The present invention issues instructions to the inspection system through a remote control system, and the UAV system and the robot system receive the instructions through a wireless communication system. Under the action of the collaborative system, the UAV system and the robot system work synchronously, that is, it can achieve the inspection purpose of high-altitude and low-altitude distribution networks, and then achieve the comprehensive inspection purpose of the distribution network, ensure the comprehensiveness and accuracy of the inspection, and meet the inspection purpose of harsh geographical environments. It saves time and labor, has high work efficiency, strong practicality, and a wide range of applications.

[0016] 2. The robot in the present invention has an obstacle avoidance function, which can reduce the error rate and ensure the inspection effect. It can also grayscale and denoise the collected high-altitude images and low-altitude images to obtain enhanced images, effectively ensuring the quality of the inspection.

[0017] 3. Since the inspection process of the present invention requires collaborative inspection by robots and drones, the present invention sets node regression constraints and inspection time interval constraints for drones and robots at the same geographical location. Under this condition, an inspection route model that takes into account inspection time, inspection cost, load balancing, synchronization time difference, and path conflict risk is constructed and solved. This can ensure that the number of drones and robots solved meets the requirements of each inspection. At the same time, the placement points solved can also meet the requirements of inspection time, inspection cost, load balancing, synchronization time difference, and path conflict risk under the premise of meeting the set inspection frequency. This can balance inspection time and cost under the premise of greater safety and reliability, that is, find the optimal collaborative number and collaborative placement points of drones and robots and the optimal inspection route. BRIEF DESCRIPTION OF THE DRAWINGS

[0018] The drawings described herein are used to provide a further understanding of the embodiments of the present invention, constitute a part of this application, and do not constitute a limitation of the embodiments of the present invention. In the drawings: Figure 1 This is a module diagram of a distribution network operation and inspection system based on collaboration between inspection drones and robots according to the present invention; Figure 2 Schematic diagram of the relationship between obstacles and obstacle avoidance units. DETAILED DESCRIPTION

[0019] In order to make the objectives, technical solutions and advantages of the present invention more clearly understood, the present invention is further described in detail below in conjunction with examples and drawings. The exemplary embodiments of the present invention and their descriptions are only used to explain the present invention and are not intended to limit the present invention.

[0020] It should be noted that since this invention utilizes drones and robotic systems to inspect distribution networks, it does not consider distribution networks within no-fly and no-travel zones. In other words, the distribution networks referred to in this invention are those where both drone and robotic inspections are permitted. Distribution networks within no-fly and no-travel zones are inspected by separate robots and drones.

[0021] This embodiment 1 provides a distribution network operation and inspection method based on the collaboration of an inspection drone and a robot, including: S1. Obtain an inspection network diagram of the area where the distribution network requiring inspection is located.

[0022] In some embodiments, the specific steps of obtaining the inspection network diagram of the area where the distribution network requiring inspection is located are: S1.1. Obtain basic power grid data of the distribution network area, where the first basic power grid data includes a first tower coordinate set, a first conductor direction set, and a first substation location set.

[0023] Here, the first tower coordinate set, the first conductor direction set, and the first substation location set include all tower coordinates, conductor directions, and substation locations in the area where the distribution network requiring inspection is located.

[0024] Basic power grid data can be obtained from the power grid GIS system, and distribution network vector data can be exported from power companies.

[0025] S1.2. Acquire aerial survey information of the area where the distribution network is located, where the aerial survey information includes a second tower coordinate set, a second conductor direction set, and a second substation location set.

[0026] Since the data of the power grid GIS system may not be updated in a timely manner or may contain construction errors, when using the data of the power grid GIS system, it is also necessary to integrate the data through aerial survey data.

[0027] S1.3. Determine the altitude of the drone’s flight path.

[0028] The altitude of the drone route can be determined through the DEM digital elevation model (such as NASASRTM30m data).

[0029] S1.4. Mark obstacles in the distribution network area based on the aerial survey information.

[0030] Here, obstacle marking refers to marking the edge of the obstacle, that is, the marked obstacle is not a coordinate point, but the entire edge coordinate set (vector boundary) of the marked obstacle. The marking tools of professional software (such as ArcGIS Pro, QGIS or drone inspection platform) can be used to manually verify the automated detection results.

[0031] S1.5. Convert the data in the first tower coordinate set, the first conductor direction set, the first substation location set, the second tower coordinate set, the second conductor direction set, and the second substation location set into the same coordinate system, and use GCP to correct the deviation between the aerial survey information and the basic power grid data, so as to obtain the final tower coordinate set, the final conductor direction set, and the final substation location set; in this way, the offset between the basic power grid system and the aerial survey information can be measured, and finally the basic power grid data can be updated according to the offset. This is the existing technology and will not be repeated here.

[0032] S1.5. An inspection network diagram is formed based on the distribution network area, obstacles, a final pole tower coordinate set, a final conductor direction set, and a final substation location set.

[0033] Specifically, aerial survey data is used as a benchmark, and vector boundaries (such as obstacle contours) are extracted after GCP correction, and then reversely updated to the GIS database.

[0034] Since the inspection network diagram is a comprehensive data of aerial survey data and GIS database, three-dimensional coordinates can be constructed in the inspection network diagram. The three-dimensional coordinates can be used to accurately locate the monitoring location, making it easier for inspection and maintenance personnel to perform immediate inspection and maintenance.

[0035] S2. Constructing an inspection route model based on the inspection network diagram that takes into account inspection time, inspection cost, load balancing, synchronization time difference, and path conflict risk.

[0036] In some embodiments, the inspection route model includes an objective function and constraints, wherein the objective function is: ; ; ; ; ; ; in, 、 、 、 、 is the weight of the corresponding coefficient, and , 、 、 、 、 There are 5 sub-goals: It is the normalized reference value of the total inspection time, used to Map to The interval, eliminating the dimension effect, can be a longer time limit value calculated based on experience or theory. It is the normalized reference value of the total inspection distance. A possible maximum distance value is set according to the scenario. It is the normalized reference value of the load variance. A larger variance limit value can be set according to the actual situation. It is the normalized reference value of the synchronization time difference, which can be set as a maximum allowed time difference. It is the normalized baseline value of the path conflict score, which can be set to the maximum conflict score value that may occur theoretically; , represents the total time of the drone moving between the drone placement points, m is the number of drones, is the number of stages of the moving path of UAV u, represents the moving time of the u-th UAV in the i-th stage of the moving path; , represents the total movement time of the robot between the robot placement points, n is the number of robots, is the number of stages of the movement path of robot r, represents the moving time of the rth robot in the jth stage of the moving path; , represents the total inspection time of the areas corresponding to all placement points, is the inspection time of the area corresponding to the placement point p; represents the set of drone placement points, Represents the set of robot placement points , represents the total distance the drone moves, represents the moving distance of the u-th UAV in the i-th stage of the moving path; , represents the total distance the robot moves, represents the moving distance of the rth UAV in the jth stage of the moving path; , represents the UAV load variance, , Represents the load of UAV u. The load here refers to the sum of the relevant tasks such as the movement time and operation time of the UAV in the inspection task it is responsible for. It is used to measure the overall scale of the tasks undertaken by the UAV. Indicates the drone placement point that drone u reaches in the i-th stage of the movement path The inspection operation time of the corresponding area, that is, the time it takes for the drone to perform the inspection task at the placement point, Indicates the average load of the drone equipment; , represents the robot load variance, , Represents the load of robot r. The load here refers to the sum of the robot's movement time, operation time and other related tasks in the inspection task it is responsible for. It is used to measure the overall scale of the robot's tasks. Indicates the robot placement point reached by robot r in the jth stage of the moving path The inspection operation time of the corresponding area, that is, the time the robot spends performing the inspection task at the placement point, Indicates the average load of the robot device; S represents the set of nodes that are visited by both drones and robots. represents the time when the UAV arrives at node s, represents the time when the robot arrives at node s; K represents the number of time steps, Indicates that when there is a path conflict between the drone u and the robot r in the kth step, ; When there is no conflict, , The conflict severity score is the score of the collision between drone u and robot r during their kth movement. This score can be calculated based on factors such as the distance and duration of the collision, with scores increasing for shorter distances and longer durations. The overall goal of this formula is to minimize the sum of the path conflict severity scores across all drones, robots, and movement steps.

[0037] When a drone and a robot need to detect the same target (such as a tower) within a preset time difference, and are responsible for high-altitude and low-altitude shooting tasks respectively, there may be a risk of path conflict. The specific analysis is as follows: 1. Spatial trajectory intersection under time difference constraint: Mandatory order of target access: If the drone is required to complete high-altitude shooting first (such as time t1), and the robot is required to complete low-altitude shooting (such as time t1+Δt, Δt=1-2 minutes), both must arrive at the same target point in sequence. At this time: The route of the drone flying from high altitude to the target point may form a horizontal projection intersection with the ground path of the robot rushing to the target point from other areas.

[0038] If the drone completes its mission and evacuates ahead of schedule, or the robot is delayed due to terrain obstacles, the time windows of the two near the target point may overlap (for example, the robot has arrived before the drone leaves), leading to close vertical contact (for example, when the drone descends to a height of 10 meters, the robot is directly below).

[0039] The pursuit effect caused by speed difference: The flying speed of the drone (20-50 km / h) is much faster than the moving speed of the robot (3-5 km / h). If the two start from different starting points and head towards the same destination, the following may occur: The drone "rapidly catches up" with the robot in front, creating a risk of "rear-end collision" when the horizontal distance is less than 5 meters (especially when the robot is in a low-speed state such as climbing a slope).

[0040] The robot had not yet completely left the target area, and the drone had returned to the area for a recheck as planned, resulting in a secondary conflict.

[0041] 2. Positioning error and environmental interference: GNSS signal obstruction: In urban high-rise buildings or mountainous scenes, the drone may trigger visual navigation due to loss of satellite signals, resulting in target point positioning deviation (error ±2 meters); when the robot moves on complex ground (such as grass and water), the wheeled / tracked odometer may accumulate errors (error ±1-3 meters per 100 meters). The actual position of the target point reached by both may deviate from the preset coordinates, causing unexpected spatial proximity.

[0042] Weather affects mission duration: Light rain may reduce the robot's movement speed by 30%, causing Δt to actually be extended to 15 minutes. The drone returns to the target point at t1+10 minutes as originally planned to take additional photos. At this time, the robot is still working on site, resulting in spatial overlap.

[0043] 3. Path coupling caused by task dependencies: If the robot needs to adjust its low-altitude shooting angle based on the drone's high-altitude image results (such as identifying the angle of a wire hanging), the following may occur: The drone failed to send data in time due to a delay in image processing, and the robot entered the target area as planned, resulting in both arriving at the same time.

[0044] The robot temporarily changed its path to adapt to the new instructions, intersecting with the drone's re-inspection route.

[0045] If there is a conflict in the paths of drones and robots during distribution network inspections, it may cause the following dangers: 1. Equipment collision risk: 1. Direct physical collision: If the paths of a drone (flying at high altitude) and a ground robot (moving at low altitude or on the ground) intersect or are too close, a collision may occur due to delayed response of the obstacle avoidance system, navigation errors, or communication interruption, resulting in damage to the equipment or even a crash.

[0046] Case: A drone's flight altitude was lower than expected (e.g., due to strong winds), and it scraped against the robotic arm or antenna above a ground robot, causing the drone's propeller to be damaged and lose control.

[0047] 2. Indirect collision caused by environmental interference: The airflow generated by drones during flight may disturb the stability of ground robots (such as small wheeled robots), especially in narrow areas (such as around distribution rooms), which may cause the robots to deviate from their path and collide with obstacles (such as poles and fences).

[0048] 2. Risk of damage to power grid equipment: 1. Accidentally touching live equipment: Path conflicts may force equipment to make emergency avoidance maneuvers, causing drones to accidentally touch high-voltage lines (such as a sudden drop in flight altitude), or ground robots to mistakenly enter dangerous areas (such as uninsulated cable trenches), causing short circuits, tripping, or even fires.

[0049] Principle: Distribution network equipment (such as transformers and insulators) is sensitive to distance. If the metal parts of a drone or the metal chassis of a robot approach within a safe distance, arc discharge may be induced.

[0050] 2. Indirect risks caused by data collection bias: Path conflicts may prevent equipment from collecting data along the preset route (e.g., a drone may miss capturing a section of the line due to obstacle avoidance), leading to missed defects (e.g., undetected insulator cracks), which may cause equipment failures in the long term.

[0051] 3. Personnel safety threats: 1. Falling objects injuring people: A drone falling after a collision or a ground robot colliding out of control may cause physical harm to on-site workers, surrounding residents or vehicles, and the risk is especially high in densely populated areas.

[0052] Regulatory requirements: According to the Interim Regulations on the Management of Unmanned Aerial Vehicle Flights, drones must maintain a safe distance from people (usually ≥50 meters). Path conflicts may exceed this limit.

[0053] 2. Secondary risks caused by electromagnetic interference: Equipment collisions may cause communication module failures, leading to abnormal operations (such as sudden acceleration of the drone or accidental touch of a control button on the robot). If workers are nearby repairing equipment, abnormal movement of the equipment may cause electric shock or mechanical injury.

[0054] 4. Operation and maintenance efficiency and cost loss: 1. Task interruption and repeated work: Path conflicts can cause inspection tasks to be suspended or restarted, requiring manual intervention to troubleshoot the problem and extending work time. For example, a collision between a drone and a robot could require return to the factory for repair, delaying the inspection schedule and increasing labor and time costs.

[0055] 2. Data synchronization failure: If the same target needs to be inspected within a preset time difference (such as high-altitude and low-altitude images of the same electric pole), path conflicts may cause data collection time misalignment, making it impossible to form a complete three-dimensional inspection report, affecting the accuracy of defect analysis.

[0056] 5. Key measures to address path conflicts: To reduce risks, the following technical and management measures can be adopted: 1. Collaborative path planning: Use a three-dimensional geographic information system (GIS) to assign independent spatial layers to drones and robots (e.g., drone flight altitude ≥ 20 meters, robot activity altitude ≤ 5 meters) to avoid vertical path overlap.

[0057] Share location data through real-time communication (such as centimeter-level positioning based on UWB) and dynamically adjust routes to avoid conflicts.

[0058] 2. Intelligent obstacle avoidance and priority management: Priority is set for devices (e.g., drones give priority to avoiding ground robots), and a multi-sensor fusion obstacle avoidance system (vision + lidar + millimeter-wave radar) is installed to improve response speed in complex environments.

[0059] 3. Dynamic monitoring and emergency response: Establish a central monitoring platform to display the equipment position, speed and status in real time. If a conflict risk is detected (such as the distance is less than the safety threshold), emergency braking or route adjustment will be automatically triggered.

[0060] 4. Task timing optimization: High-altitude and low-altitude inspections of the same target are time-sliced. For example, the drone completes high-altitude photography first, and then the robot enters the area for low-altitude inspection to avoid simultaneous operations.

[0061] In some embodiments, the inspection route model includes an objective function and constraints, wherein the objective function is: ; ; ; ; ; in, 、 、 、 is the weight of the corresponding coefficient, and , 、 、 、 There are 5 sub-goals: It is the normalized reference value of the total inspection time, used to Map to The interval, eliminating the dimension effect, can be a longer time limit value calculated based on experience or theory. It is the normalized reference value of the total inspection distance. A possible maximum distance value is set according to the scenario. It is the normalized reference value of the load variance. A larger variance limit value can be set according to the actual situation. It is the normalized reference value of the synchronization time difference and can be set as a maximum allowed time difference; , represents the total time of the drone moving between the drone placement points, m is the number of drones, is the number of stages of the moving path of UAV u, represents the moving time of the u-th UAV in the i-th stage of the moving path; , represents the total movement time of the robot between the robot placement points, n is the number of robots, is the number of stages of the movement path of robot r, represents the moving time of the rth robot in the jth stage of the moving path; , represents the total inspection time of the areas corresponding to all placement points, is the inspection time of the area corresponding to the placement point p; represents the set of drone placement points, Represents the set of robot placement points , represents the total distance the drone moves, represents the moving distance of the u-th UAV in the i-th stage of the moving path; , represents the total distance the robot moves, represents the moving distance of the rth UAV in the jth stage of the moving path; , represents the UAV load variance, , Represents the load of UAV u. The load here refers to the sum of the relevant tasks such as the movement time and operation time of the UAV in the inspection task it is responsible for. It is used to measure the overall scale of the tasks undertaken by the UAV. Indicates the drone placement point that drone u reaches in the i-th stage of the movement path The inspection operation time of the corresponding area, that is, the time it takes for the drone to perform the inspection task at the placement point, Indicates the average load of the drone equipment; , represents the robot load variance, , Represents the load of robot r. The load here refers to the sum of the robot's movement time, operation time and other related tasks in the inspection task it is responsible for. It is used to measure the overall scale of the robot's tasks. Indicates the robot placement point reached by robot r in the jth stage of the moving path The inspection operation time of the corresponding area, that is, the time the robot spends performing the inspection task at the placement point, Indicates the average load of the robot device; S represents the set of nodes that are visited by both drones and robots. represents the time when the UAV arrives at node s, Indicates the time when the robot arrives at node s.

[0062] The constraints include safety constraints, regulatory constraints, environmental constraints, technical constraints, node regression constraints, and inspection time interval constraints for drones and robots at the same geographical location.

[0063] 1. Security constraints include: 1. Electrical safety: 1.1. Minimum Safety Distance: According to the "Electric Power Safety Work Regulations" in China, drones must maintain a distance of ≥3 meters (10kV), ≥5 meters (110kV), and ≥10 meters (500kV) from energized equipment. Internationally, the IEEE 1584 standard requires an arc safety distance to be calculated based on the voltage level (e.g., ≥6.7 meters for 500kV).

[0064] 1.2. Electromagnetic Interference: Strong electromagnetic fields around high-voltage lines may interfere with the drone's flight control signals. Therefore, you need to use an anti-interference model (such as the DJI M300RTK + anti-magnetic kit).

[0065] 2. Anti-collision: 2.1. Wire obstacle avoidance: The drone must be equipped with a lidar or visual obstacle avoidance system (such as LivoxMID-40) to prevent collisions with wires or poles.

[0066] 2.2. Multi-machine collaboration: When operating a swarm, a horizontal distance of ≥20 meters must be maintained to avoid mid-air collisions.

[0067] 2. Regulatory constraints include: 1. Airspace restrictions: Flight altitude: The Civil Aviation Administration of China stipulates that the maximum altitude of civil drones is 120 meters (an over-limit flight permit must be applied for).

[0068] 2. Approval process: 2.1. Operation Permit: A flight plan (including time, coordinates, and altitude) must be submitted to the local air traffic control department.

[0069] 2.2. Radio compliance: The image transmission frequency of the drone must comply with the regulations of the National Radio Regulatory Commission (e.g., 5.8GHz frequency band power ≤ 200mW).

[0070] 3. Environmental constraints: 1. Weather conditions: 1.1. Wind speed: If the wind speed exceeds level 8 (17m / s), the aircraft must be grounded (for example, the DJI M300 has a wind resistance of 15m / s).

[0071] 1.2. Precipitation: Rainy or snowy weather may cause the sensor to fail (a waterproof model such as the DJI Matrice 300+ protective cover is required).

[0072] 1.3. Temperature: -20℃~50℃ is the common operating range of drones. Preheating or heat dissipation measures are required in extremely cold or hot environments.

[0073] 2. Topography and vegetation: 2.1. Complex terrain: Mountainous areas and forests require lidar modeling to assist navigation (such as DJIL1).

[0074] 2.2. Vegetation obstruction: When tree canopies obstruct the wires, it is necessary to lower the flight altitude or use an obstacle-penetrating drone (such as a bionic flapping-wing drone).

[0075] 4. Technical constraints: 1. Battery life: 1.1. Battery Limitations: Mainstream industry drones have a flight time of 25-40 minutes (e.g., the DJI M300 has a 55-minute flight time with dual batteries).

[0076] 1.2. Charging solution: fixed charging station or battery swapping system (i.e. battery replacement).

[0077] 2. Sensor performance: 2.1. Detection Accuracy: Infrared thermal imaging: Temperature resolution must be ≤ 0.5°C (e.g., FLIRT1020). Visible light camera: Pixels ≥ 20 million (e.g., Zenmuse P1).

[0078] 2.2. Data backhaul: 5G / 4G network latency must be ≤ 200ms. Otherwise, edge computing (local AI recognition defects) will be enabled.

[0079] 3. Navigation and positioning: 3.1. GPS-denied environment: The GPS signal may be weak inside the substation, requiring the integration of visual SLAM + UWB ultra-wideband positioning.

[0080] 3.2. High precision requirements: The tower coordinate positioning error must be ≤10cm (RTK / PPK differential positioning is required).

[0081] 5. Node regression constraints: 1. When a drone is set up at a drone placement point and the drone is conducting an inspection, the drone placement point will be used as the starting point, and the starting point will be used as the destination of the drone after the inspection is completed. For a drone placement point with a drone, only one drone will be sent out in the same inspection batch.

[0082] That is, assuming there are 10 drone placement points, the best drone placement points A, B, C, and D are found through optimization. Then, drones are placed at these four placement points. At the initial moment, A (drone a is placed), B (drone b is placed), C (drone b is placed), and D (drone b is placed). During the inspection, drone a at point A starts the inspection. After the inspection is completed, only one drone will return to point A (the returning drone may be a, or it may be b, c, or d). According to the optimal drone inspection route, the placement point corresponding to the drone returning to point A at the previous moment can be obtained.

[0083] For example, if the initial state is: A (a), B (b), C (c), D (d), and the placement order is set to the drone at placement point A will be placed at placement point B next time, the drone at placement point B will be placed at placement point C next time, the drone at placement point C will be placed at placement point D next time, and the drone at placement point D will be placed at placement point A next time. Then, after a round of inspection, the state is: A (d), B (a), C (b), D (c), and so on.

[0084] If there are multiple drones at a certain location, only one drone will be dispatched from that location for an inspection batch. For example, if there are three drones at location A, only one will be dispatched for an inspection batch, and the other two will serve as backup drones.

[0085] 2. When a robot is set up at a robot placement point and the robot is conducting an inspection, the robot placement point will be used as the starting point, and the starting point will be used as the end point of the robot after the inspection is completed. For a robot placement point with a robot placed, only one robot will be sent out in the same inspection batch.

[0086] 6. Time interval constraints for inspections of drones and robots at the same geographic location: When in the same inspection batch, the time interval between drones and robots inspecting the same geographical location does not exceed the preset time threshold.

[0087] For example, for the insulator at the geographic coordinates (x1, y1), the time when the drone patrols to the location and takes the initial high-altitude image of the insulator is 12:53 on a certain day of a certain month of a certain year. Assuming that the preset time threshold is 30 seconds, then during the time period from 12:52:30 to 12:53:30 on the same day of the same month of the same year, the robot must patrol to the insulator and take the initial low-altitude image of the insulator.

[0088] In the inspection network diagram, multiple drone placement nodes and robot placement points are set. At the drone placement node, the drone can be charged and data can be transmitted, that is, the collected image can be transmitted to the terminal. At the robot placement node, the robot can be charged and data can be transmitted, that is, the collected image can be transmitted to the terminal, and the robot placement point can be the same placement point as the drone placement point.

[0089] It should also be noted that if the optimal drone placement node and the optimal robot placement node are found (there may be several optimal drone placement nodes and optimal robot placement nodes), then in each inspection batch, all the optimal human-machine placement nodes and the optimal robot placement nodes will respectively send out a drone / robot for inspection.

[0090] That is, for each inspection, only one drone / robot will be deployed from a designated drone / robot placement point. Therefore, during a single inspection, the number of optimal robot and drone placement points equals the number of robots and drones required for the inspection, respectively. Therefore, determining the optimal drone placement points, optimal drone inspection routes, optimal robot placement points, and optimal robot inspection routes is a key issue addressed by this invention.

[0091] Generally speaking, in the existing inspection schemes for distribution networks using drones and robots, only the optimal path of drones and robots is taken into consideration, while the placement and number of drones and robots are ignored. Although the optimal path of drones and robots is found, which can save inspection time, for distribution networks with a large inspection intensity (inspection frequency), even if one drone and robot inspects the distribution network through the optimal path, it will take a lot of time. If more drones and robots are set up, and given the nature of drones and robots requiring charging, if the parking positions (charging positions) set are not good, the optimal path will not be optimal, resulting in the time spent not being optimal, and the cost will also increase. Therefore, how to balance the relationship between inspection cost and inspection time is a problem that the present invention needs to solve.

[0092] Based on this, step S3 of the present invention solves the inspection route model through a parameter optimization algorithm, thereby determining the optimal drone placement point, the optimal drone inspection route, the optimal robot placement point and the optimal robot inspection route from the inspection network diagram.

[0093] In some embodiments, solving the inspection route model by using a parameter optimization algorithm is specifically solving it using a multi-objective genetic algorithm based on NSGA-II, and the specific process is: Randomly generate the number m of drones, the number n of robots, and the corresponding paths to form an initialization population, wherein the initialization population includes a number of individuals, each of which includes the number of drones, the number of robots, and the corresponding path; Calculate the function values ​​of the 4 / 5 sub-goals of all individuals to obtain the specific values ​​of each individual in these 4 / 5 dimensions; Perform non-dominated sorting + crowding calculation: Non-dominated sorting: For individuals X and Y in the population, if: (1) the target values ​​of X are not worse than those of Y; (2) at least one target value of X is better than that of Y, then X can dominate Y; first find all individuals in the population that are not dominated by other individuals, classify them into the first layer (non-dominated layer), and assign them the same non-dominated order; then, remove these individuals from the population, repeat the above operation for the remaining individuals, find the non-dominated individuals in the second layer, and so on, until all individuals in the population are stratified; in this way, each individual has its own non-dominated order, reflecting its superiority and inferiority level in the population; crowding calculation: in the same non-dominated layer, construct a rectangle with the individual's nearest neighbor as the vertex. The crowding of an individual is related to the area of ​​the rectangle. The larger the area, the less crowded the individual. By calculating the crowding, the superiority and inferiority of individuals in the same non-dominated layer can be further distinguished, avoiding the concentration of individuals in certain areas of the Pareto front, and maintaining the diversity of the solution set; The parent population is selected through tournament selection: the number of individuals N to be compared is determined, and N individuals are randomly selected from the population, with each individual having equal probability of being selected. The fitness of these N individuals is then compared (this combines non-dominated order and crowding, with individuals with smaller non-dominated order being prioritized; if the non-dominated order is the same, individuals with smaller crowding are prioritized). The individual with the best fitness is selected to enter the parent population, and this process is repeated multiple times until the parent population reaches a predetermined size. Merge the parent and offspring generations based on the elite strategy, retaining the top M excellent individuals: merge the parent population and offspring population together, then sort them according to the non-dominated order and crowding degree of the individuals, give priority to individuals with smaller non-dominated order, and for individuals in the same non-dominated layer, select individuals with smaller crowding degree. Select the top N excellent individuals from the merged population to form a new parent population and enter the next round of iteration; Repeat the above steps until the maximum number of iterations is reached or convergence occurs, and the placement point found is the optimal placement point.

[0094] Through the above steps, the coordinates and number of drone placement points for each inspection are found, and the flight time is determined by the optimal inspection route and the flight speed of the drone. Similarly, the coordinates and number of robot placement points for each inspection are found, and the travel time is determined by the optimal inspection route and the travel speed of the robot.

[0095] At the same time, it should be noted that once the optimal drone placement point and the optimal robot placement point are determined, they will not change unless there are major changes in the distribution network equipment.

[0096] After the optimal inspection route is designed, it can be numbered to facilitate the UAV system and the robot system to perform automatic inspections according to the route number.

[0097] S4. Perform inspections based on the optimal inspection plan to collect images of the distribution network, that is, place at least one drone and at least one robot at the optimal drone placement point and the optimal robot placement point, respectively, and perform inspections according to the optimal drone inspection route and the optimal robot inspection route to collect images of the distribution network.

[0098] In some embodiments, if the inspection intensity is not strong (i.e., the inspection frequency is not high), then when the drone completes the inspection and flies to the next set placement point, it needs to be charged. If the battery is full and it is not yet time for the next inspection, then only one drone needs to be set at the optimal drone placement point. If the next inspection time comes before the battery is fully charged, then multiple drones need to be set at this time. Specifically, in some embodiments, if the drone is a charging drone, then the drones placed at the optimal drone placement point The inspection interval is the time interval between two consecutive inspections. Here, the inspection interval is in hours. That is, if an inspection is required every hour, then the inspection interval is 1.

[0099] If the drone is a battery replacement drone, the number of drones placed at the optimal drone placement point is 1 or 2.

[0100] Assuming that the drone is a battery-replacement drone, if the inspection interval is greater than the inspection time + battery replacement time, then only one drone is needed, and the number of batteries is set according to actual needs. If the inspection interval is less than or equal to the inspection time + battery replacement time, then two drones are required. The reason why two drones are set when the inspection interval is equal to the inspection time + battery replacement time is to avoid special circumstances that cause the inspection time + battery replacement time to be slightly longer than the inspection time, such as suddenly finding that the replaced battery is broken and a new battery needs to be replaced.

[0101] If the robot is a charging robot, the robot placed at the optimal robot placement point ; If the robot is a battery replacement robot, the number of robots placed at the optimal robot placement point is 1 or 2 The situation of robots can be referred to as drones, so I will not go into details here.

[0102] S5. Perform data fusion on the initial high-altitude image and the initial low-altitude image respectively collected by the UAV and the robot, as well as the corresponding time information and geographic location information to obtain a fusion result, and transmit the fusion result to the terminal.

[0103] The geographical location information here can be obtained through a positioning system.

[0104] In some embodiments, the specific process of fusing the initial high-altitude image and the initial low-altitude image respectively collected by the drone and the robot and the corresponding time information and geographic location information to obtain a fusion result, and transmitting the fusion result to the terminal is as follows: S5.1. Perform radiation correction and geometric correction on the initial high-altitude image and the initial low-altitude image, respectively, to ensure that the coordinate systems of the two images are consistent. Then, perform contrast enhancement, brightness adjustment, and denoising on the images, respectively, to obtain a first intermediate high-altitude image and a first intermediate low-altitude image, thereby improving image clarity and laying the foundation for subsequent feature extraction.

[0105] Both radiation correction and geometric correction are existing technologies and will not be described in detail here.

[0106] Image enhancement includes edge sharpening and grayscale transformation. Among them, edge sharpening uses the Unsharp Mask (USM) tool to achieve edge sharpening. USM sharpening emphasizes edges based on differential operations. Its formula is expressed as: G(i,j)=f(i,j)-f(i-1,j) and G(i,j)=f(i,j)-f(i,j-1), and G(i,j)=\sqrt{(f(i,j)-f(i-1,j))^2+(f(i,j)-f(i,j-1))^2}, Where f(i, j) represents the pixel value at position (i, j) in the image, and G(i, j) represents the pixel value of the edge map after differential operation.

[0107] Grayscale transformation is based on linear transformation, and its formula is: y(x,y)=k*x(x,y)+b, Where y(x,y) represents the target pixel value; x(x,y) represents the source pixel value; k represents the slope; b represents the intercept.

[0108] Edge sharpening is used to enhance the contrast of image edges, making the edges clearer and ensuring the quality and effect of image collection; grayscale transformation improves the display effect of the image by changing the grayscale value of each pixel in the image. Its purpose is to enhance or weaken the grayscale of the image, improve image quality, highlight features or suppress unnecessary features, change the histogram distribution, make the pixel distribution more uniform, and make the image more suitable for processing through image enhancement.

[0109] S5.2. Feature extraction is performed on the first intermediate high-altitude image and the first intermediate low-altitude image corresponding to the same geographic location information of the same inspection batch to obtain corresponding feature points, and the feature points of the first intermediate high-altitude image are matched with the feature points of the first intermediate low-altitude image, thereby achieving feature extraction, matching, and alignment of the first intermediate high-altitude image and the first intermediate low-altitude image.

[0110] In the feature extraction stage, SIFT, SURF and other algorithms are used to extract local features such as edges and corners of the image to enhance the matching ability of overlapping areas. For global features, principal component analysis or deep learning features are combined; In the matching stage, the extracted features are matched through brute force matching, FLANN matching or graph cut optimization-based methods.

[0111] Feature extraction uses one of the Fourier transform, window Fourier transform, wavelet transform and least squares method; among them, the Fourier transform has the advantages of fast scanning speed, high resolution, high sensitivity, high wave number accuracy, wide application range, strong characteristic, rich information and no restriction on sample physical state; the window Fourier transform has the advantages of balancing time resolution and frequency resolution, being able to capture local information of the signal, being suitable for non-stationary signals, having good compatibility with classical algorithms, high computational efficiency and a wide range of applications; the least squares method has the advantages of simplicity and intuition, wide application, efficient calculation, small sample fitting, estimation accuracy, reduced error influence, support vector machine and good flexibility; the wavelet transform provides finer time-frequency resolution, and the wavelet basis function can be selected as needed.

[0112] S5.3. Input the aligned first intermediate high-altitude image and the first intermediate low-altitude image into a neural network model for image fusion to obtain an initial fused image; Specifically, a generative adversarial network or a convolutional neural network is used to automatically learn multi-scale feature representation to achieve high-precision fusion. This is an existing technology and will not be described in detail here.

[0113] Through deep learning models, multi-level and multi-scale feature information is automatically extracted and fused from images of different resolutions (such as low-altitude images and high-altitude images) or different modalities to generate more accurate and complete alignment or fusion results.

[0114] Combining features at different scales can not only preserve details (such as the texture of low-altitude images) but also utilize global context (such as the layout of high-altitude images) to improve the robustness of matching or fusion.

[0115] S5.4. Post-process and optimize the initial fused image to obtain a fusion result.

[0116] In some embodiments, the specific process of this step is: The initial fused image is deblurred (e.g., sharpening based on wavelet transform), and the fusion parameters (e.g., weight coefficient, fusion algorithm) are adjusted through cross-validation, visual evaluation (e.g., PSNR, SSIM) or field verification to optimize the stitching effect, so as to achieve accuracy evaluation and adjustment and ultimately obtain the fusion result.

[0117] S5.5. Transmit the fusion result to the terminal.

[0118] It should be noted that the initial high-altitude images may be images taken by a high-definition camera and an infrared thermometer (surface temperature measurement). Therefore, a set of initial high-altitude images includes camera images and temperature images.

[0119] Similarly, the initial low-altitude images can be images taken by a high-definition camera and an infrared thermometer (surface temperature measurement). Therefore, a set of initial high-altitude images includes camera images and temperature images.

[0120] When S5 is executed, the camera image shot at a high altitude and the camera image shot at a low altitude are fused, and the temperature image shot at a high altitude and the temperature image shot at a low altitude are fused.

[0121] Therefore, the fusion result includes the camera fusion result obtained by fusion processing the captured camera images and the temperature image obtained by fusion processing the captured temperature images.

[0122] S6. Determine whether the operating status of the distribution network is abnormal based on the fusion result. If so, generate an inspection warning result and continue the inspection. Otherwise, continue the inspection until one round of inspection is completed. In some embodiments, the specific process of determining whether the operating state of the distribution network is abnormal according to the fusion result is as follows: S6.1. Build a monitoring database containing images of distribution network faults, providing a comparison standard for the collected image data. For example, pre-stored images of faults such as bird damage, tree obstructions, floating objects, partial discharges, and insulation failures can be matched against images captured by cameras. Temperature threshold data for each device in the distribution network can also be pre-stored. This temperature threshold data can be used to match the temperature image, obtain temperature data from the temperature image, and then determine whether the temperature threshold data has been exceeded.

[0123] Of course, the infrared thermometer can also be a point temperature measuring thermometer. In this way, image fusion only needs to fuse the images taken by the camera, and the temperature data measured by the point temperature measuring infrared thermometer on the drone and the point temperature measuring infrared thermometer on the robot at the same geographical location are averaged.

[0124] S6.2. Match the fusion result with several of the fault images. If there is a matching fault image, the fault type of the fault image is used as the fault type of the fusion result, and an alarm message is recorded to help staff handle the fault in a timely and accurate manner; if there is no matching fault image, delete the fusion result.

[0125] Deleting fusion results without abnormalities can save storage space. At the same time, it eliminates the need for staff to repeatedly process image data, shortens processing time, and improves work efficiency.

[0126] S7. Perform the next inspection according to the set inspection frequency.

[0127] Inspection frequency is the inspection interval time. Example 2 discloses a distribution network operation and inspection system based on the collaboration of inspection drones and robots, such as Figure 1 As shown, the method for executing a distribution network operation and inspection method based on collaboration between an inspection drone and a robot as described in Example 1 includes: The inspection system is used to implement the operation and inspection of the distribution network.

[0128] The remote control system is used to remotely control the inspection system. Specifically, the terminal uploads the completed inspection route to the remote control system, which stores and plans it. The remote control system issues inspection commands to the drone system based on the set inspection route, causing it to conduct inspections. The remote control system also issues inspection commands to the robot system based on the set inspection route, causing it to conduct inspections.

[0129] The data storage system is used to store the operation and inspection data of the inspection system.

[0130] The working principle and usage process of this embodiment are as follows: instructions are issued to the inspection system through the remote control system, and the UAV system and the robot system receive the instructions through the wireless communication system. Under the action of the collaborative system, the UAV system and the robot system work synchronously, that is, the inspection purpose of the distribution network at high altitude and low altitude can be achieved, and then the comprehensive inspection purpose of the distribution network can be achieved, ensuring the comprehensiveness and accuracy of the inspection, and can meet the inspection purpose of harsh geographical environments. It saves time and effort, has high work efficiency, strong practicality, and a wide range of applications.

[0131] Specifically, the inspection system includes an unmanned aerial vehicle system, a robot system, a wireless communication system, a collaborative system and a positioning system, wherein the unmanned aerial vehicle system is used to inspect high-altitude distribution networks, the robot system is used to inspect low-altitude distribution networks, the wireless communication system is used to connect the unmanned aerial vehicle system, the robot system and the remote control system, the collaborative system is used to realize the synchronous operation of the unmanned aerial vehicle system and the robot system, and the positioning system is used to realize the position positioning of the unmanned aerial vehicle system and the robot system.

[0132] In some embodiments, the UAV system and the robotic system are both equipped with a data acquisition unit, and the data acquisition unit is selected from a high-definition camera and an infrared thermometer. The UAV system and the robotic system are also equipped with a positioning unit, and the positioning unit is selected from one of the GPS, radar, and Beidou satellite positioning systems.

[0133] The drone system and the robot system are also equipped with an obstacle avoidance unit, which uses an infrared rangefinder. More specifically, the drone system and the robot system are also equipped with an obstacle avoidance unit that calculates the shortest distance between the obstacle avoidance unit and the obstacle based on the drone / robot's travel speed, reaction speed, wind speed, and the relative angle between the drone / robot's travel direction and the wind direction using a parameter optimization algorithm.

[0134] Taking the genetic algorithm as an example to find the shortest distance between the drone and the obstacle, the specific steps are: B1, the genes of the chromosome include the UAV’s travel speed, reaction speed, wind speed, the relative angle between the UAV / robot’s travel direction and the wind direction, and the distance between the UAV and the nearest edge of the obstacle, e.g. Figure 2 As shown, Z is the obstacle, a is the drone, and the distance is l. Consider a chromosome as an individual.

[0135] B2. Set the number of iterations of the genetic algorithm and the number of individuals to be solved in each generation of the population; B3. Set the objective function. Specifically, the objective function is: ; in, The actual distance the drone travels around obstacles. This indicates the preset distance between the drone and the obstacle when it passes over it under safety constraints. Passing the obstacle here means that the drone's projection will coincide with the obstacle in a direction perpendicular to the drone's direction of travel.

[0136] B4. Encode the drone's travel speed, reaction speed, wind speed, the relative angle between the drone / robot's travel direction and the wind direction, and the distance from the drone to the nearest obstacle to obtain the encoded value of the individual to be solved, and randomly generate an initial population composed of several individuals to be solved. Let this initial population be the parent population, and let the individual to be solved in the parent population be the parent individual; B5. Performing a drone inspection obstacle avoidance simulation based on the coding values ​​of each parent individual of the parent population to obtain an obstacle avoidance simulation result; B6. Calculate the fitness function according to the obstacle avoidance simulation results, and calculate the fitness value of each parent individual in the parent population according to the fitness function and sort them; the fitness function is .

[0137] B7. Save the first M parent individuals with the largest fitness values ​​in the parent population, select parent individuals from all parent individuals except the first M parent individuals with the largest fitness values ​​through roulette wheel, perform crossover and mutation operations to obtain offspring individuals, calculate the fitness values ​​of the offspring individuals after crossover and mutation and sort them, reinsert the offspring individuals into the parent population according to their fitness values, select a set number of individuals to be solved to form a new parent population, and then return to B5; B8. Repeat B5-B7 until the number of iterations is reached or the objective function value is within the specified threshold range.

[0138] Calculating the shortest distance between the robot's obstacle avoidance unit and the obstacle through the parameter optimization algorithm is similar to that of the drone, so it will not be repeated here.

[0139] In this way, the shortest distance for obstacle avoidance under different conditions can be obtained, which can prevent the robot / drone from being unable to avoid obstacles in extreme situations.

[0140] The specific implementation methods described above further illustrate the objectives, technical solutions and beneficial effects of the present invention in detail. It should be understood that the above description is only a specific implementation method of the present invention and is not intended to limit the scope of protection of the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention should be included in the scope of protection of the present invention.

Claims

1. A distribution network operation and inspection method based on the collaboration of inspection drones and robots, characterized in that: include: Obtain the inspection network diagram of the area where the distribution network that needs to be inspected is located; Based on the inspection network diagram, an inspection route model is constructed that takes into account inspection time, inspection cost, load balancing, and synchronization time difference; Solving the inspection route model through a parameter optimization algorithm to obtain an optimal inspection plan; Performing an inspection based on the optimal inspection plan to collect an image of the distribution network; Performing data fusion on the initial high-altitude images and initial low-altitude images collected by the drone and the robot, as well as the corresponding time information and geographic location information, to obtain a fusion result, and transmitting the fusion result to the terminal; Determine whether the operating status of the distribution network is abnormal based on the fusion result. If so, generate an inspection warning result and continue the inspection. Otherwise, continue the inspection until one round of inspection is completed. Perform the next inspection according to the set inspection frequency.

2. A distribution network operation and inspection method based on collaboration between inspection drones and robots according to claim 1, characterized in that: The specific steps to obtain the inspection network diagram of the area where the distribution network that needs to be inspected is located are: Obtaining basic power grid data of the area where the distribution network is located, the first basic power grid data including a first tower coordinate set, a first conductor direction set, and a first substation location set; Acquire aerial survey information of the area where the distribution network is located, the aerial survey information including a second tower coordinate set, a second conductor direction set, and a second substation location set; Determine the altitude of the drone’s flight path; Marking obstacles in the distribution network area based on the aerial survey information; Converting the data in the first tower coordinate set, the first conductor orientation set, the first substation location set, the second tower coordinate set, the second conductor orientation set, and the second substation location set into the same coordinate system, and using GCP to correct the deviation between the aerial survey information and the basic power grid data, thereby obtaining a final tower coordinate set, a final conductor orientation set, and a final substation location set; An inspection network diagram is formed based on the distribution network area, obstacles, final tower coordinate set, final conductor direction set and final substation location set.

3. The distribution network operation and inspection method based on the collaboration of inspection drones and robots according to claim 1 is characterized in that: The inspection route model includes an objective function and constraints, wherein the objective function is: ； ； ； ； ； in, 、 、 、 is the weight of the corresponding coefficient, and , 、 、 、 There are 4 sub-goals: It is the normalized reference value of the total inspection time, used to Map to The interval, eliminating the dimension effect, can be a longer time limit value calculated based on experience or theory. It is the normalized reference value of the total inspection distance. A possible maximum distance value is set according to the scenario. It is the normalized reference value of the load variance. A larger variance limit value can be set according to the actual situation. It is the normalized reference value of the synchronization time difference and can be set as a maximum allowed time difference; , represents the total time of the drone moving between the drone placement points, m is the number of drones, is the number of stages of the moving path of UAV u, represents the moving time of the u-th UAV in the i-th stage of the moving path; , represents the total movement time of the robot between the robot placement points, n is the number of robots, is the number of stages of the movement path of robot r, represents the moving time of the rth robot in the jth stage of the moving path; , represents the total inspection time of the areas corresponding to all placement points, is the inspection time of the area corresponding to the placement point p; represents the set of drone placement points, Represents the set of robot placement points , represents the total distance the drone moves, represents the moving distance of the u-th UAV in the i-th stage of the moving path; , represents the total distance the robot moves, represents the moving distance of the rth UAV in the jth stage of the moving path; , represents the UAV load variance, , Represents the load of UAV u. The load here refers to the sum of the relevant tasks such as the movement time and operation time of the UAV in the inspection task it is responsible for. It is used to measure the overall scale of the tasks undertaken by the UAV. Indicates the drone placement point that drone u reaches in the i-th stage of the movement path The inspection operation time of the corresponding area, that is, the time it takes for the drone to perform the inspection task at the placement point, Indicates the average load of the drone equipment; , represents the robot load variance, , Represents the load of robot r. The load here refers to the sum of the robot's movement time, operation time and other related tasks in the inspection task it is responsible for. It is used to measure the overall scale of the robot's tasks. Indicates the robot placement point reached by robot r in the jth stage of the moving path The inspection operation time of the corresponding area, that is, the time the robot spends performing the inspection task at the placement point, Indicates the average load of the robot device; S represents the set of nodes that are visited by both drones and robots. represents the time when the UAV arrives at node s, represents the time when the robot arrives at node s; The constraints include safety constraints, regulatory constraints, environmental constraints, technical constraints, node regression constraints, and inspection time interval constraints for drones and robots at the same geographical location.

4. A distribution network operation and inspection method based on collaboration between inspection drones and robots according to claim 3, characterized in that: The inspection route model is solved by a parameter optimization algorithm, specifically using a multi-objective genetic algorithm based on NSGA-II. The specific process is as follows: Randomly generate the number m of drones, the number n of robots, and the corresponding paths to form an initialization population, wherein the initialization population includes a number of individuals, each of which includes the number of drones, the number of robots, and the corresponding path; Calculate the function values ​​of the four sub-goals of all individuals to obtain the specific values ​​of each individual in these four dimensions; Perform non-dominated sorting and crowding calculation: Non-dominated sorting: For individuals X and Y in a population, if: X's objective values ​​are not worse than Y's or X has at least one objective value better than Y, then X can dominate Y. First, find all individuals in the population that are not dominated by other individuals, assign them to the first layer (the non-dominated layer), and assign them the same non-dominated order. Then, remove these individuals from the population and repeat the above process for the remaining individuals to find the non-dominated individuals in the second layer, and so on, until all individuals in the population are stratified. In this way, each individual has its own non-dominated order, reflecting its level of superiority and inferiority in the population. Crowding calculation: In the same non-dominated layer, rectangles are constructed with the individual's nearest neighbors as vertices. The crowding degree of an individual is related to the area of ​​the rectangle. The larger the area, the less crowded the individual. By calculating the crowding degree, it is possible to further distinguish the superiority and inferiority of individuals in the same non-dominated layer, avoid individuals concentrating in certain areas of the Pareto front, and maintain the diversity of the solution set. The parent generation is selected through tournament selection: the number of individuals N to be compared each time is determined, N individuals are randomly selected from the population, and each individual has the same probability of being selected. Then the fitness of these N individuals is compared, and the individual with the best fitness is selected to enter the parent population. This operation is repeated multiple times until the size of the parent population reaches the preset size. Merge the parent and offspring generations based on the elite strategy, retaining the top M excellent individuals: merge the parent population and offspring population together, then sort them according to the non-dominated order and crowding degree of the individuals, give priority to individuals with smaller non-dominated order, and for individuals in the same non-dominated layer, select individuals with smaller crowding degree. Select the top N excellent individuals from the merged population to form a new parent population and enter the next round of iteration; Repeat the above steps until the maximum number of iterations is reached or convergence occurs, and the placement point found is the optimal placement point.

5. A distribution network operation and inspection method based on collaboration between inspection drones and robots according to claim 4, characterized in that: If the drone is a charging drone, the optimal drone placement point is ; If the drone is a battery replacement drone, the number of drones placed at the optimal drone placement point is 1 or 2; If the robot is a charging robot, the robot placement point is ; If the robot is a battery replacement robot, the number of robots placed at the optimal robot placement point is 1 or 2.

6. The distribution network operation and inspection method based on the collaboration of inspection drones and robots according to claim 1 is characterized in that: The specific process of fusing the initial high-altitude image and the initial low-altitude image collected by the UAV and the robot, as well as the corresponding time information and geographic location information, to obtain a fusion result, and transmitting the fusion result to the terminal is as follows: Performing radiation correction and geometric correction on the initial high-altitude image and the initial low-altitude image, respectively, and then performing contrast enhancement, brightness adjustment, and denoising processing on the initial high-altitude image and the initial low-altitude image, respectively, to obtain a first intermediate high-altitude image and a first intermediate low-altitude image; Performing feature extraction on the first intermediate high-altitude image and the first intermediate low-altitude image corresponding to the same geographic location information of the same inspection batch to obtain corresponding feature points, and matching the feature points of the first intermediate high-altitude image with the feature points of the first intermediate low-altitude image to achieve alignment of the first intermediate high-altitude image and the first intermediate low-altitude image; Inputting the aligned first intermediate high-altitude image and the first intermediate low-altitude image into a neural network model for image fusion to obtain an initial fused image; Post-processing and optimizing the initial fused image to obtain a fusion result; The fusion result is transmitted to the terminal.

7. A distribution network operation and inspection method based on collaboration between inspection drones and robots according to claim 6, characterized in that: The specific process of post-processing and optimizing the initial fused image to obtain the fusion result is as follows: The initial fused image is deblurred, and the fusion parameters are adjusted through cross-validation, visual evaluation or field verification to optimize the stitching effect, so as to achieve accuracy evaluation and adjustment and finally obtain the fusion result.

8. The distribution network operation and inspection method based on the collaboration of inspection drones and robots according to claim 1 is characterized in that: The specific process of judging whether the operating state of the distribution network is abnormal according to the fusion result is as follows: Building a monitoring database, wherein the monitoring database is provided with a fault image related to a distribution network fault; The fusion result is matched with several fault images. If there is a matching fault image, the fault type of the fault image is used as the fault type of the fusion result, and an alarm message is recorded and issued; if there is no matching fault image, the fusion result is deleted.

9. A distribution network operation and inspection system based on the collaboration of an inspection drone and a robot, used to implement the distribution network operation and inspection method based on the collaboration of an inspection drone and a robot according to any one of claims 1 to 8, characterized in that: include: Inspection system is used to realize the operation and inspection of distribution network, including: The inspection system includes an unmanned aerial vehicle system, a robot system, a wireless communication system, a collaborative system, and a positioning system. The unmanned aerial vehicle system is used to inspect high-altitude distribution networks, the robot system is used to inspect low-altitude distribution networks, the wireless communication system is used to connect the unmanned aerial vehicle system, the robot system, and the remote control system, the collaborative system is used to synchronize the operation of the unmanned aerial vehicle system and the robot system, and the positioning system is used to locate the positions of the unmanned aerial vehicle system and the robot system. A remote control system for remotely controlling the inspection system; The data storage system is used to store the operation and inspection data of the inspection system.

10. A distribution network operation and inspection system based on collaboration between inspection drones and robots according to claim 9, characterized in that: The drone system and the robot system are also equipped with an obstacle avoidance unit, which calculates the shortest distance between the obstacle avoidance unit and the obstacle through a parameter optimization algorithm based on the travel speed, reaction speed, relative angle between the travel direction of the drone / robot and the wind direction, and wind speed of the drone / robot.

Images