A multi-target optimization inspection method based on cooperation of an unmanned aerial vehicle and an inspection platform

Abstract

The application discloses a multi-target optimization inspection method based on cooperation of an unmanned aerial vehicle and an inspection platform, and comprises the following steps: obtaining basic data of a converter station, inputting a pre-constructed inspection scheduling model to obtain an inspection scheduling scheme; performing global path search and endurance verification based on an A-star algorithm according to the inspection scheduling scheme, and generating an initial global inspection track; controlling the unmanned aerial vehicle to fly according to the initial global inspection track, and receiving real-time flight data returned by the unmanned aerial vehicle; obtaining a locally optimized global inspection track based on a visual SLAM algorithm and a rolling window mechanism according to the real-time flight data; and outputting a corrected flight control signal based on a G2O graph optimization framework and an adaptive error compensation algorithm, so that the unmanned aerial vehicle stably flies according to the locally optimized global inspection track. The application can significantly improve the efficiency and operation quality of equipment inspection of an extra-high voltage converter station, and is suitable for intelligent operation and maintenance scenes of large-scale DC transmission converter stations.

Description

Technical Field

[0001] This invention relates to a multi-objective optimized inspection method based on the collaboration between unmanned aerial vehicles (UAVs) and an inspection platform, belonging to the field of UAV inspection technology. Background Technology

[0002] With the rapid development of high-voltage direct current (HVDC) transmission projects, converter stations, as the core hubs for power collection and conversion, possess massive equipment, complex structures, and operate in harsh environments. Taking a ±800kV UHVDC project as an example, a single converter station typically occupies several square kilometers and contains numerous critical electrical devices, including converter valves, smoothing reactors, DC filters, cooling systems, and DC field equipment. The equipment is diverse in type and densely distributed. As the scale of HVDC transmission continues to expand, the operation and maintenance tasks of converter stations are becoming increasingly heavy. How to efficiently and comprehensively complete equipment inspections and ensure the safe and stable operation of the system has become a crucial issue that urgently needs to be addressed.

[0003] Currently, converter station inspections are still primarily conducted manually, requiring maintenance personnel to check each piece of equipment along a predetermined route. This method is not only inefficient and labor-intensive, but also highly dependent on personnel experience, with inspection quality significantly affected by subjective factors. Furthermore, manual inspections pose certain safety risks in high-voltage, strong electromagnetic, and complex environments. In addition, with the expansion of converter station scale and the increase in the number of devices, manual inspections struggle to achieve high-frequency, comprehensive, and detailed testing, easily leading to blind spots or omissions of critical equipment, and are no longer sufficient to meet the current development needs of intelligent operation and maintenance.

[0004] To improve the automation level of inspections, existing technologies have attempted to introduce drones or robots for assisted inspections. By equipping these robots with visible light and infrared sensors, they can detect equipment status and improve local inspection efficiency to some extent. However, current technologies mostly rely on single drones or individual devices for inspection tasks, which has significant shortcomings: First, a single drone's endurance and operating radius limit its ability to achieve full coverage inspections of large-scale converter stations. Second, converter stations are densely packed with equipment and have complex structures, including many tall buildings and narrow passages. Traditional path planning and obstacle avoidance methods are not adaptable to these complex environments, affecting inspection safety and stability. Furthermore, existing methods lack systematic scheduling and collaborative optimization of inspection resources, failing to coordinate the deployment of the inspection platform and the allocation of drone tasks at an overall level, resulting in limited improvements in operational efficiency. Summary of the Invention

[0005] The purpose of this invention is to overcome the shortcomings of the prior art and provide a multi-objective optimized inspection method based on the collaboration of UAVs and inspection platforms. Compared with traditional manual inspection or single UAV inspection, this invention can significantly improve the efficiency and quality of equipment inspection in UHV converter stations and is applicable to intelligent operation and maintenance scenarios of large-scale DC transmission converter stations.

[0006] To achieve the above objectives, the present invention is implemented using the following technical solution:

[0007] This invention discloses a multi-objective optimized inspection method based on the collaboration of unmanned aerial vehicles (UAVs) and inspection platforms, comprising:

[0008] Obtain basic data of the converter station;

[0009] The basic data is input into the pre-built inspection scheduling model, and the solution is obtained with the goal of minimizing the total operation time of the inspection platform.

[0010] According to the inspection scheduling scheme, global path search and endurance verification are performed based on the A-star algorithm to generate an initial global inspection track.

[0011] The initial global inspection trajectory is sent to the UAV, the UAV is controlled to fly according to the initial global inspection trajectory, and the real-time flight data transmitted back by the UAV is received.

[0012] Based on the real-time flight data, the initial global inspection trajectory is locally replanned using the visual SLAM algorithm and the rolling window mechanism to obtain a locally optimized global inspection trajectory.

[0013] Using the locally optimized global inspection trajectory as the target trajectory, and based on the G2O graph optimization framework and adaptive error compensation algorithm, a corrected flight control signal is output to enable the UAV to fly stably along the locally optimized global inspection trajectory.

[0014] Furthermore, the basic data includes the location data of the equipment to be inspected, the location data of alternative deployment points, the flight parameters of the UAV, and the constraint data of the inspection workday.

[0015] Furthermore, the objective function of the inspection scheduling model is expressed as follows:

[0016] ;

[0017] In the formula, Represents the function to be minimized; Indicates the total operation time of the inspection platform; Indicates a weekday gathering; The expression representing the operation time of the inspection platform on the qth working day is as follows:

[0018]

[0019] In the formula, A binary decision variable representing whether the nth alternative deployment point is selected as the actual deployment point; The binary decision variable represents the inspection platform's access to the nth alternative deployment point on the qth working day. This represents the inspection time of the drone at the nth alternative deployment point.

[0020] The binary decision variable represents whether the i-th candidate job node is selected as the actual job node; The binary decision variable representing whether the j-th candidate job node is selected as the actual job node; Let $\mathbf{i}$ represent the binary decision variable when the inspection platform travels from the $i$-th alternative work node to the $j$-th alternative work node on the $q$-th working day. This represents the straight-line geographical distance between the i-th candidate job node and the j-th candidate job node; This represents the equivalent time parameter for the deployment and switching of the inspection platform; This indicates the longest working hours for inspection personnel on a weekday;

[0021] P represents the set of candidate deployment points; G represents the set of candidate job nodes.

[0022] Furthermore, the inspection time of the drone at the nth alternative deployment point The expression is:

[0023]

[0024] In the formula, The task time of the k-th drone at the nth alternative deployment point is expressed as:

[0025]

[0026] In the formula, A binary decision variable representing whether the nth alternative deployment point is selected as the actual deployment point; Let m represent the m-th equipment to be inspected, which is served by the n-th alternative deployment point;

[0027] This represents a binary decision variable when the m-th equipment to be inspected is inspected by the k-th drone;

[0028] This represents the straight-line geographical distance between the m-th device and the n-th alternative deployment point;

[0029] Indicates the drone's inspection flight speed;

[0030] The time taken for the drone to inspect the m-th device is represented by ; T represents the set of devices to be inspected.

[0031] Furthermore, the constraints of the inspection scheduling model include weekday route start and end constraints, single deployment of deployment points and operation continuity constraints, equipment deployment point correspondence constraints, path endpoint validity constraints, access selection consistency constraints, decision variable constraints, single drone flight and single device constraints, and variable attribute constraints.

[0032] The expression for the start and end constraints of the weekday route is:

[0033]

[0034] In the formula, The binary decision variable represents the time when the inspection platform travels from maintenance station a to the j-th alternative work node on the q-th working day. The binary decision variable represents the time when the inspection platform travels from the j-th alternative work node to the maintenance station a on the q-th working day;

[0035] The expression for the single deployment of the deployment point and the operational continuity constraint is:

[0036]

[0037]

[0038] In the formula, A binary decision variable representing whether the nth alternative deployment point is selected as the actual deployment point; Let $\mathbf{i}$ represent the binary decision variable when the inspection platform travels from the $i$-th alternative work node to the $n$-th alternative deployment point on the $q$-th working day. Let $\mathbf{ ...

[0039] The expression for the constraint corresponding to the device deployment point is:

[0040]

[0041] In the formula, Let m represent the m-th equipment to be inspected, which is served by the n-th alternative deployment point; This indicates the collection of equipment to be inspected;

[0042] The expression for the path endpoint validity constraint is:

[0043]

[0044] In the formula, Let $\mathbf{i}$ represent the binary decision variable when the inspection platform travels from the $i$-th alternative work node to the $j$-th alternative work node on the $q$-th working day. The binary decision variable represents the number of times the inspection platform visits the i-th alternative work node on the q-th working day. The binary decision variable represents the number of times the inspection platform visits the j-th alternative work node on the q-th working day.

[0045] The expression for the access selection consistency constraint is:

[0046]

[0047] In the formula, The binary decision variable represents the inspection platform's access to the i-th alternative deployment point on the q-th working day.

[0048] The expression for the decision variable constraint is:

[0049]

[0050]

[0051] The expression for the constraint on a single UAV flight and a single device is:

[0052]

[0053]

[0054] In the formula, This represents a binary decision variable when the m-th equipment to be inspected is inspected by the k-th drone;

[0055] This indicates the collection of equipment to be inspected;

[0056] This represents the mission time of the k-th drone at the n-th alternative deployment point;

[0057] The expression for the variable attribute constraint is:

[0058]

[0059]

[0060] In the formula, This represents the mission time of the k-th drone at the n-th alternative deployment point.

[0061] Furthermore, the inspection scheduling scheme includes:

[0062] The optimal set of deployment points covering all equipment to be inspected;

[0063] The inspection platform deployment scheme is used to determine the access order and dwell time of the inspection platform to each deployment point in the optimal deployment point set on each working day.

[0064] The drone inspection task allocation scheme is used to determine the list of equipment to be inspected by each drone at each deployment point.

[0065] Furthermore, the generation of the initial global inspection track includes:

[0066] According to the inspection scheduling scheme, the A* algorithm is used to perform global path search and generate the original global inspection track.

[0067] Perform a battery life verification on the original global inspection track;

[0068] If the original global inspection track fails the endurance verification, the optimal deployment point set and the inspection platform deployment scheme remain unchanged. The UAV inspection task allocation scheme is divided into multiple UAV inspection sub-task allocation schemes whose range meets the endurance threshold. The original global inspection track is then generated again using the A* algorithm with each UAV inspection sub-task allocation scheme as input.

[0069] If the original global inspection track passes the endurance verification, then the original global inspection track will be used as the initial global inspection track.

[0070] Furthermore, the use of the A* algorithm for global path search includes:

[0071] Based on the optimal deployment point set of the inspection scheduling scheme and the UAV inspection task allocation scheme, determine the take-off deployment point of each UAV and the corresponding list of equipment to be inspected.

[0072] Obtain the spatial distribution data of the equipment to be inspected corresponding to the list of equipment to be inspected in the converter station, and cluster the spatial locations of the equipment to be inspected based on spatial proximity to obtain multiple local inspection areas.

[0073] For each local inspection area, extract the outer edge feature points located at the spatial distribution boundary; construct a local target rectangular region based on the outer edge feature points;

[0074] For each local target rectangular area, based on the preset imaging parameters of the UAV onboard imaging equipment, the corresponding original local inspection track is generated in an S-shaped routing planning manner;

[0075] Based on all the original local inspection tracks, the TSP traveler algorithm is used to traverse the shortest loop access order of all local inspection areas to generate the original global inspection track.

[0076] Furthermore, the real-time flight data includes environmental data within a scrolling window, which includes image data, inertial measurement data, GPS data, barometric pressure data, and radar data.

[0077] Based on the environmental data within the scrolling window, the real-time pose data of the UAV in the current environment is calculated and environmental map data is constructed to obtain a local environment model.

[0078] Based on obstacle detection algorithms and motion prediction algorithms, dynamic obstacles appearing in the local environment model are detected, and the predicted motion trend of the dynamic obstacles is predicted.

[0079] When the predicted trajectory of the dynamic obstacle interferes with the initial global inspection track, a collision risk is determined. The Morphin local path planning algorithm is used to perform local path correction on the interfered segment of the initial global inspection track to generate a safe path that bypasses the dynamic obstacle. The initial global inspection track is then updated with the safe path to obtain the locally optimized global inspection track.

[0080] The scrolling window is updated in real time as the drone moves, and the environmental data in the updated scrolling window serves as the input data for the next local replanning cycle.

[0081] Furthermore, the output corrected flight control signal includes:

[0082] The real-time flight data is input into the G2O graph optimization framework, and nonlinear optimization and extended Kalman filtering are performed to output UAV attitude and position estimation data.

[0083] Based on the UAV attitude and position estimation data, the deviation between the UAV attitude and position estimation data and the locally optimized global inspection trajectory is calculated to obtain the trajectory tracking error.

[0084] Based on Lyapunov stability theory, an adaptive error compensation algorithm is used to compensate for the trajectory tracking error and output a corrected flight control signal so that the UAV can fly stably according to the locally optimized global inspection trajectory under uncertain airflow disturbances.

[0085] Compared with the prior art, the beneficial effects achieved by the present invention are as follows:

[0086] The present invention provides a multi-objective optimized inspection method based on the collaboration of UAV and inspection platform. First, an inspection scheduling scheme is obtained based on the inspection scheduling model. Inspection resources are coordinated from the global level, which effectively reduces the switching time of the inspection platform between deployment points and the waiting time of UAV operation, and significantly reduces the overall inspection operation cost.

[0087] Secondly, by combining global path search with endurance verification using the A-star algorithm, the endurance feasibility of the trajectory is determined and adaptive segmented replanning is performed during the planning stage. This ensures that the generated initial global inspection trajectory meets both the full coverage requirement and the physical constraints of a single UAV endurance, guaranteeing the continuity and completeness of the inspection task in actual execution.

[0088] In addition, by introducing visual SLAM algorithm and rolling window mechanism, environmental map is built in real time, UAV pose is calculated, dynamic obstacles in the rolling window are identified and motion is predicted, and local path replanning is performed on the interfered flight segment, so that UAV has real-time obstacle avoidance capability in complex electrical equipment environment, which significantly enhances the safety and real-time performance of inspection trajectory planning.

[0089] Finally, by fusing multi-source sensor data based on the G2O graph optimization framework to estimate the attitude and position of the UAV, and combining it with an adaptive error compensation algorithm to correct the trajectory tracking error, the UAV can still maintain stable tracking of the target trajectory under airflow disturbance and environmental uncertainty conditions, thus ensuring the flight stability and positioning accuracy of the inspection operation. Attached Figure Description

[0090] Figure 1 This is a flowchart of a multi-objective optimized inspection method based on the collaboration between a drone and an inspection platform, provided by an embodiment of the present invention.

[0091] Figure 2 This is a schematic diagram of a converter station inspection scenario provided in an embodiment of the present invention;

[0092] Figure 3 This is a schematic diagram of the A-star search provided in an embodiment of the present invention;

[0093] Figure 4 This is a schematic diagram of the outer edge feature points and the local target rectangular region provided in the embodiment of the present invention;

[0094] Figure 5 This is a schematic diagram of the lens imaging principle provided in an embodiment of the present invention;

[0095] Figure 6 This is a schematic diagram of the "S"-shaped route planning for local target rectangular areas provided in this embodiment of the invention;

[0096] Figure 7This is a schematic diagram of the front end of the SLAM algorithm provided in an embodiment of the present invention;

[0097] Figure 8 This is a display diagram of the localization results of the SLAM algorithm provided in this embodiment of the invention. Detailed Implementation

[0098] The present invention will be further described below with reference to the accompanying drawings. The following embodiments are only used to more clearly illustrate the technical solution of the present invention, and should not be used to limit the scope of protection of the present invention.

[0099] This embodiment discloses a multi-objective optimized inspection method based on the collaboration between unmanned aerial vehicles (UAVs) and an inspection platform, such as... Figure 1 As shown, it includes:

[0100] Obtain basic data of the converter station;

[0101] The basic data is input into the pre-built inspection scheduling model, and the solution is obtained by minimizing the total operation time of the inspection platform.

[0102] According to the inspection and scheduling plan, global path search and endurance verification are performed based on the A-star algorithm to generate an initial global inspection track.

[0103] The initial global inspection trajectory is sent to the UAV, the UAV is controlled to fly according to the initial global inspection trajectory, and the real-time flight data transmitted back by the UAV is received.

[0104] Based on real-time flight data, the initial global inspection trajectory is locally replanned using the visual SLAM algorithm and the rolling window mechanism to obtain the locally optimized global inspection trajectory.

[0105] Using the locally optimized global inspection trajectory as the target trajectory, and based on the G2O graph optimization framework and adaptive error compensation algorithm, the corrected flight control signal is output to enable the UAV to fly stably according to the locally optimized global inspection trajectory.

[0106] The technical concept of this invention is as follows:

[0107] First, based on the inspection scheduling model, an inspection scheduling scheme is obtained, which coordinates inspection resources from a global perspective, effectively reducing the switching time of the inspection platform between deployment points and the waiting time of drone operations, and significantly reducing the overall inspection operation cost.

[0108] Secondly, by combining global path search with endurance verification using the A-star algorithm, the endurance feasibility of the trajectory is determined and adaptive segmented replanning is performed during the planning stage. This ensures that the generated initial global inspection trajectory meets both the full coverage requirement and the physical constraints of a single UAV endurance, guaranteeing the continuity and completeness of the inspection task in actual execution.

[0109] In addition, by introducing visual SLAM algorithm and rolling window mechanism, environmental map is built in real time, UAV pose is calculated, dynamic obstacles in the rolling window are identified and motion is predicted, and local path replanning is performed on the interfered flight segment, so that UAV has real-time obstacle avoidance capability in complex electrical equipment environment, which significantly enhances the safety and real-time performance of inspection trajectory planning.

[0110] Finally, by fusing multi-source sensor data based on the G2O graph optimization framework to estimate the attitude and position of the UAV, and combining it with an adaptive error compensation algorithm to correct the trajectory tracking error, the UAV can still maintain stable tracking of the target trajectory under airflow disturbance and environmental uncertainty conditions, thus ensuring the flight stability and positioning accuracy of the inspection operation.

[0111] Step 1: Obtain the basic data of the converter station.

[0112] The basic data includes the location data of the equipment to be inspected, the location data of alternative deployment points, the flight parameters of the drone, and the constraint data of the inspection workday.

[0113] Please see Figure 2 , Figure 2 This is a schematic diagram of a converter station inspection scenario involved in this invention. In this scenario, the converter station equipment is densely distributed and has a complex structure. The inspection area includes DC field equipment, cooling systems, and related electrical facilities, providing a basic environment for UAV inspection path planning and task scheduling.

[0114] The equipment to be inspected in the converter station mainly includes smoothing reactors, DC filters, converter transformers, DC field buses, DC circuit breakers, cooling devices, and related auxiliary equipment.

[0115] The drone flight parameters include: drone ensemble, average drone flight speed, drone inspection time at the equipment, drone single-batch endurance, and parameters of the onboard imaging equipment.

[0116] The constraint data for inspection workdays include: the set of workdays, the longest working time of inspection personnel on a workday, and the coordinates of the maintenance station location.

[0117] In this embodiment, the following set is involved:

[0118] Maintenance station set A, in this embodiment, only one maintenance station is set.

[0119] There is a set of candidate deployment points P, and the number of candidate deployment points in set P is N.

[0120] There is a set T of equipment to be inspected, and the number of equipment to be inspected in the set T is M.

[0121] There is a set of candidate job nodes G, and the number of candidate job nodes in the set of candidate job nodes G is I.

[0122] Let U be a set of drones, and L be the number of drones in U.

[0123] Let W be the set of working days, and Q be the number of working days in W.

[0124] Step 2: Input the basic data into the pre-built inspection scheduling model and solve it with the goal of minimizing the total operation time of the inspection platform to obtain the inspection scheduling scheme.

[0125] The objective function of the inspection scheduling model is expressed as follows:

[0126] ;

[0127] In the formula, Represents the function to be minimized; Indicates the total operation time of the inspection platform; Indicates a weekday gathering;

[0128] This represents the operation time of the inspection platform on the qth working day. The operation time of the inspection platform on each working day consists of two parts: the total inspection operation time of the selected deployment points covered on that working day and the platform deployment switching time between deployment points. The expression is as follows, provided that the operation time of the inspection platform on each working day does not exceed the maximum working time for inspection personnel set by the power company:

[0129]

[0130] In the formula, A binary decision variable representing whether the nth alternative deployment point is selected as the actual deployment point; The binary decision variable represents the inspection platform's access to the nth alternative deployment point on the qth working day. This represents the inspection time of the drone at the nth alternative deployment point.

[0131] The binary decision variable represents whether the i-th candidate job node is selected as the actual job node; The binary decision variable representing whether the j-th candidate job node is selected as the actual job node; Let $\mathbf{i}$ represent the binary decision variable when the inspection platform travels from the $i$-th alternative work node to the $j$-th alternative work node on the $q$-th working day. This represents the straight-line geographical distance between the i-th candidate job node and the j-th candidate job node; This represents the equivalent time parameter for the deployment and switching of the inspection platform; This indicates the longest working hours for inspection personnel on a weekday;

[0132] P represents the set of candidate deployment points; G represents the set of candidate job nodes.

[0133] Since the inspection platform needs to release its drones at each deployment point to perform inspection tasks, and multiple drones can operate in parallel, the inspection platform can only proceed to the next deployment phase after all drones have completed their respective inspection tasks and returned to the platform. Therefore, the inspection operation time of the inspection platform at each selected deployment point is equal to the maximum value of the task times of all drones at that deployment point. Thus, the inspection time of the drone at the nth alternative deployment point is... The expression is:

[0134]

[0135] In the formula, This represents the mission time of the k-th drone at the n-th candidate deployment point. At each selected deployment point, the mission time of each drone consists of two parts: the flight time between the deployment point and the equipment, and the inspection time at the equipment. The expression is:

[0136]

[0137] In the formula, A binary decision variable representing whether the nth alternative deployment point is selected as the actual deployment point; Let m represent the m-th equipment to be inspected, which is served by the n-th alternative deployment point;

[0138] This represents a binary decision variable when the m-th equipment to be inspected is inspected by the k-th drone;

[0139] This represents the straight-line geographical distance between the m-th device and the n-th alternative deployment point;

[0140] Indicates the drone's inspection flight speed;

[0141] This represents the inspection time of the drone at the m-th device.

[0142] The constraints of the inspection and scheduling model include weekday route start and end constraints, single deployment and operation continuity constraints, equipment deployment point correspondence constraints, path endpoint validity constraints, access selection consistency constraints, decision variable constraints, single drone deployment and single device constraints, and variable attribute constraints.

[0143] The workday route start and end constraint indicates that the inspection platform uses the maintenance station as the starting node for its work each workday and returns to the maintenance station after completing the day's inspection tasks. The expression is:

[0144]

[0145] In the formula, The binary decision variable represents the time when the inspection platform travels from maintenance station a to the j-th alternative work node on the q-th working day. The binary decision variable represents the time when the inspection platform travels from the j-th alternative work node to the maintenance station a on the q-th working day;

[0146] The single deployment and operation continuity constraint for deployment points indicates that each selected deployment point is deployed by the inspection platform only once on a certain workday, and enters the next deployment phase after the inspection operation of that deployment point is completed. The expression is:

[0147]

[0148]

[0149] In the formula, A binary decision variable representing whether the nth alternative deployment point is selected as the actual deployment point; Let $\mathbf{i}$ represent the binary decision variable when the inspection platform travels from the $i$-th alternative work node to the $n$-th alternative deployment point on the $q$-th working day. Let represent the binary decision variable when the inspection platform travels from the nth alternative deployment point to the jth alternative operation node on the qth working day.

[0150] The device deployment point constraint indicates that all devices to be inspected are provided with inspection services by one and only one selected deployment point. The expression is:

[0151]

[0152] In the formula, Let m represent the m-th equipment to be inspected, which is served by the n-th alternative deployment point; This indicates the collection of equipment to be inspected;

[0153] The path endpoint validity constraint is used to indicate that the two endpoints of any platform deployment switchover process within a given workday are both operation nodes involved in the inspection platform on that workday. The expression is:

[0154]

[0155] In the formula, Let $\mathbf{i}$ represent the binary decision variable when the inspection platform travels from the $i$-th alternative work node to the $j$-th alternative work node on the $q$-th working day. The binary decision variable represents the number of times the inspection platform visits the i-th alternative work node on the q-th working day. The binary decision variable represents the number of times the inspection platform visits the j-th alternative work node on the q-th working day.

[0156] The access selection consistency constraint indicates that if a deployment point is accessed by the inspection platform on a certain workday, then that deployment point must be selected. The expression is:

[0157]

[0158] In the formula, The binary decision variable represents the inspection platform's access to the i-th alternative deployment point on the q-th working day.

[0159] The expression for the decision variable constraint is:

[0160]

[0161]

[0162] In the formula, P represents the set of candidate deployment points;

[0163] G represents the set of candidate work nodes; W represents the set of workdays; T represents the set of equipment to be inspected.

[0164] The single-drone-per-flight, single-device constraint indicates that each device is accessed by one and only one drone, and that at each selected deployment point, the total number of drone takeoffs equals the total number of devices served by that deployment point. Together, these two expressions represent that a single drone inspects only one device per takeoff, as shown in the following expression:

[0165]

[0166]

[0167] In the formula, This represents a binary decision variable when the m-th equipment to be inspected is inspected by the k-th drone;

[0168] This indicates the collection of equipment to be inspected;

[0169] This represents the mission time of the k-th drone at the n-th alternative deployment point;

[0170] The expression for variable attribute constraint is:

[0171]

[0172]

[0173] In the formula, This represents the mission time of the k-th drone at the n-th alternative deployment point.

[0174] The final inspection scheduling plan includes:

[0175] The optimal set of deployment points covering all equipment to be inspected;

[0176] The inspection platform deployment plan is used to determine the order of access to each deployment point in the optimal deployment point set and the dwell time of the inspection platform on each working day.

[0177] The drone inspection task allocation scheme is used to determine the list of equipment to be inspected by each drone at each deployment point.

[0178] Step 3: Based on the inspection and scheduling plan, perform global path search and endurance verification using the A-star algorithm to generate the initial global inspection track.

[0179] Specifically, generating the initial global inspection track includes:

[0180] 3.1. Based on the inspection scheduling plan, the A* algorithm is used to perform global path search and generate the original global inspection track;

[0181] Global path search is performed using the A* algorithm, including:

[0182] Based on the optimal deployment point set of the inspection scheduling plan and the UAV inspection task allocation plan, determine the take-off deployment point of each UAV and the corresponding list of equipment to be inspected.

[0183] Obtain the spatial distribution data of the equipment to be inspected corresponding to the list of equipment to be inspected in the converter station, and cluster the spatial locations of the equipment to be inspected based on spatial proximity to obtain multiple local inspection areas.

[0184] For each local inspection area, extract the outer edge feature points located at the spatial distribution boundary; based on the outer edge feature points, construct a local target rectangular region;

[0185] For each local target rectangular area, based on the preset imaging parameters of the UAV onboard imaging equipment, the corresponding original local inspection track is generated in an S-shaped routing planning manner;

[0186] Based on all the original local inspection tracks, the TSP traveler algorithm is used to traverse the shortest loop access order of all local inspection areas to generate the original global inspection track.

[0187] Specifically, the A* algorithm is a commonly used pathfinding and graph traversal algorithm, such as... Figure 3 As shown. In planet A, starting from the starting point, check its adjacent squares, then expand outwards until the target is found. The specific steps include:

[0188] (1) Starting from the starting point, add the starting point to an open list consisting of squares. Currently, the open list only contains the starting point; more items will be added gradually. The squares in the open list represent those that the path may or may not pass through. The open list is a list of squares to be checked.

[0189] (2) Check the squares adjacent to the starting point, ignore the squares occupied by walls, rivers and other illegal terrain, add the passable squares to the open list, and set the starting point as the parent node of these squares.

[0190] (3) Remove the starting point from the open list and add it to the closed list. Each cell in the closed list is one that no longer needs to be focused on.

[0191] (4) Repeat the above steps, select the node with the lowest replacement value from the open list as the current node for inspection, until the target node is found or the open list is empty.

[0192] Obtain the spatial distribution data of the equipment to be inspected corresponding to the list of equipment to be inspected in the converter station, and cluster the spatial locations of the equipment to be inspected based on spatial proximity to obtain multiple local inspection areas.

[0193] For each local inspection area, extract the outer edge feature points located at the spatial distribution boundary; based on the outer edge feature points, construct a local target rectangular region;

[0194] For each local target rectangular area, based on the preset imaging parameters of the UAV onboard imaging equipment, the corresponding original local inspection track is generated in an S-shaped routing planning manner;

[0195] Based on all the original local inspection tracks, the TSP traveler algorithm is used to traverse the shortest loop access order of all local inspection areas to generate the original global inspection track.

[0196] Within the framework of the aforementioned global path search algorithm, this embodiment further establishes an information model for converter station inspection objects, defining the set of equipment to be inspected in the converter station as T, where T = {t1, t2, ..., t}. m ,…, t M Let t be the elevation of the highest point of the m-th equipment to be inspected. m h;

[0197] Next, based on the spatial distribution of the DC field equipment in the converter station, the area where multiple inspection devices are located within a set spatial threshold is divided into a local inspection area X. f This is used for local trajectory planning of UAVs. Ultimately, multiple local inspection areas X1, X2, ..., X are obtained. f , ..., X F The number of local inspection areas is F, where F ≤ M, and M represents the number of equipment to be inspected in the converter station.

[0198] The minimum value of the spatial threshold is the minimum straight-line distance between any two devices in the set T of the converter station to be inspected, and the maximum value of the spatial threshold is the maximum straight-line distance between any two devices in the set T of the converter station to be inspected. The maximum value of the spatial threshold is the straight-line distance between any two devices in the set T of the converter station to be inspected, and the specific value is determined according to the actual situation.

[0199] Measure the local inspection area X f The latitude and longitude information of the outer edge feature points; the outer edge feature points of the local inspection area must meet the boundary conditions of the equipment distribution area, the outer edge angle is not 180°, and there must be at least one or more straight lines passing through the point that do not intersect with the main body of the equipment within the local target. A schematic diagram of the outer edge feature points is shown below. Figure 4 As shown.

[0200] Next, the UAV inspection route planning within the local target is completed. Among the feature points on the outer edge of the local target, the line connecting the two farthest points is selected as one diagonal of the rectangular inspection area of ​​the local target, denoted as L1. The other diagonal is L2. Diagonals L2 and L1 are of equal length and their center points coincide. The angle between L2 and L1 is adjusted to construct the rectangular area of ​​the local target, so that the rectangular area of ​​the local target includes all the feature points on the outer edge of the local inspection area.

[0201] The angle of view is determined by the size, focal length, and working height of the camera's charge-coupled device (CCD). Based on the imaging principle of a convex lens, the formula is: Working height : Angle of view = Focal length : CCD size. The area of ​​the imaging region is determined by the angle of view, i.e., Imaging region area = Angle of view (length) × Angle of view (width). The lens imaging principle is as follows... Figure 5 As shown.

[0202] Define the starting point of the UAV inspection of the f-th local target rectangular area as S. f Let S be the endpoint of the UAV inspection of the f-th local target rectangular area. f e, define the elevation of the highest point of the equipment structure within the f-th local target rectangular area as S. f h, S fh represents the maximum installation height of the equipment structure of the subordinate converter station to be inspected;

[0203] Taking the local target rectangular area S as an example f The four vertices serve as the inspection starting point S f s, with the local target rectangular area S f Middle and inspection starting point S f The longer side of the connected 's' lines serves as the initial inspection route, and an 'S'-shaped route is used to inspect the local target rectangular area S. f The inspection route is planned to ensure that the imaging area covers the local target. The latitude and longitude information of the UAV inspection route is calculated using the principle of convex lens imaging. The corresponding original local inspection track is generated by planning the route in an S-shaped pattern. Figure 6 As shown.

[0204] The comprehensive optimization of the UAV inspection route completes the output. If the number of local target rectangular areas is 1, then the local target inspection endpoint S1e and the local target inspection starting point S1s are connected in a straight line; if the number of local targets is greater than 1, then the last local target inspection endpoint S1e is connected in a straight line. F e is directly connected to the first local target inspection starting point S1s;

[0205] The inspection route with the shortest total inspection distance for the UAV is selected as the optimal UAV inspection path for the converter station, and the starting point S of any local target along this optimal path is defined. f s is set as the mission start point, and the output is performed according to the format of the UAV flight control script.

[0206] Next, the UAV inspection path planning between multiple local target rectangular areas is completed. Using the converter station UAV inspection route planning system, the straight-line distance between each local target is calculated based on the geographical coordinate information of each local target. Using the existing technology of the Traveling Salesman Problem (TSP), the shortest path scheme that traverses all local target rectangular areas and returns to the starting point is determined.

[0207] Specifically, using the starting and ending coordinates of each local target rectangular area as access nodes, the straight-line flight distance between any two access nodes is calculated. With the goal of minimizing the total cascaded flight distance, the Traveling Salesman Problem algorithm is used to determine the optimal cascaded sequence of the original local inspection tracks that start from the takeoff deployment point, traverse all local target rectangular areas, and return to the takeoff deployment point. According to the optimal cascaded sequence, the original local inspection tracks are cascaded end to end to generate the original global inspection track.

[0208] 3.2 Perform endurance verification on the original global inspection track;

[0209] If the original global inspection track fails the endurance verification, the optimal deployment point set and inspection platform deployment scheme remain unchanged. The UAV inspection task allocation scheme is then divided into multiple UAV inspection sub-task allocation schemes with endurance thresholds based on flight range. The original global inspection track is then regenerated using the A* algorithm with each UAV inspection sub-task allocation scheme as input.

[0210] 3.3 If the original global inspection track passes the endurance verification, the original global inspection track will be used as the initial global inspection track.

[0211] Step 4: Send the initial global inspection track to the UAV, control the UAV to fly according to the initial global inspection track, and receive the real-time flight data transmitted back by the UAV.

[0212] The drone in this embodiment is equipped with a global exposure or rolling shutter exposure grayscale camera, as well as low-cost sensors such as MEMS (Micro-electromechanical Systems) inertial devices, GPS, barometers, and radar.

[0213] Real-time flight data includes environmental data within a scrolling window, which includes image data, inertial measurement data, GPS data, barometric pressure data, and radar data.

[0214] Step 5: Based on real-time flight data, the initial global inspection track is locally replanned using the visual SLAM (Simultaneous Localization and Mapping) algorithm and the rolling window mechanism to obtain the locally optimized global inspection track.

[0215] The specific steps are as follows:

[0216] 5.1 Based on the environmental data in the scrolling window, calculate the real-time pose data of the UAV in the current environment and construct environmental map data to obtain a local environment model.

[0217] Specifically, based on the grayscale information of the image data, inertial measurement data, and radar data in the environmental data within the scrolling window, the real-time pose data of the UAV in the current environment is calculated using the visual SLAM algorithm, and environmental map data is constructed to obtain a local environment model. The front-end of the visual SLAM algorithm is as follows: Figure 7 As shown, the positioning results are as follows: Figure 8 As shown.

[0218] Simultaneously, the Structure from Motion algorithm can be used to reconstruct the 3D shape of the target building. This process can be automated based on the given building's location and the provided human-computer interaction software.

[0219] 5.2 Based on obstacle detection and motion prediction algorithms, detect dynamic obstacles appearing in the local environment model and predict the predicted motion trend of the dynamic obstacles.

[0220] The obstacle detection algorithm used in this embodiment is the YOLOv8 deep learning object detection algorithm and the Euclidean clustering segmentation algorithm. By performing semantic segmentation on the point cloud and image data in the local environment model, dynamic obstacles such as people, vehicles, and other drones and their contour information are identified.

[0221] The motion prediction algorithm is a trajectory extrapolation algorithm based on Kalman filtering and a constant velocity model. It tracks the position of dynamic obstacles in multiple consecutive frames, estimates their instantaneous velocity and heading angle, and then predicts their motion trajectory in a short time window in the future.

[0222] The specific algorithms described above are merely illustrative examples of this embodiment. Those skilled in the art can select other algorithms with similar functions to implement obstacle detection and motion prediction according to actual application scenarios. The scope of protection of this invention is not limited to the above-described exemplary algorithms.

[0223] 5.3 When the predicted trajectory of a dynamic obstacle interferes with the initial global inspection track, a collision risk is determined. The Morphin local path planning algorithm is used to perform local path correction on the interfered segments of the initial global inspection track, generating a safe path to bypass the dynamic obstacle. The initial global inspection track is then updated with the safe path to obtain the locally optimized global inspection track.

[0224] The scrolling window updates in real time as the drone moves, and the updated environmental data within the scrolling window serves as the input data for the next local replanning cycle.

[0225] The specific steps of Morphin's local path planning algorithm are as follows:

[0226] 1: Near the current location of the UAV, based on the UAV's kinematic constraints such as minimum turning radius and maximum yaw rate, a set of candidate local paths are generated, each path being defined by different curvature and orientation angle parameters;

[0227] 2: Based on the static obstacle map and dynamic obstacle prediction trajectory in the local environment model, calculate the collision risk cost, path length cost, and deviation cost from the original global trajectory for each candidate path.

[0228] 3: Eliminate candidate paths that have spatial-temporal conflicts with the predicted motion trajectory, such as candidate paths where the spatial distance between the two is less than the safety threshold at the same time;

[0229] 4: Among the remaining collision-free paths, select the path with the minimum weighted overall cost as the safe path to bypass the dynamic obstacle;

[0230] 5: Smoothly stitch the beginning of the safe path with the current uninterrupted initial global inspection track, and realign the end with the initial global inspection track after passing the dynamic obstacle to obtain a locally optimized global inspection track.

[0231] Step 6: Using the locally optimized global inspection trajectory as the target trajectory, and based on the G2O graph optimization framework and adaptive error compensation algorithm, output the corrected flight control signal to enable the UAV to fly stably according to the locally optimized global inspection trajectory.

[0232] Output corrected flight control signals, including:

[0233] 6.1 Input real-time flight data into the G2O graph optimization framework, perform nonlinear optimization and extended Kalman filtering, and output UAV attitude and position estimation data;

[0234] In this embodiment, the UAV uses a lightweight binocular global exposure or rolling shutter exposure grayscale camera and multiple sensors, combined with a G2O-based backend Bundle Adjustment optimization algorithm, to provide stable and effective visual information under conditions of low power consumption, cost and size, thereby realizing the UAV's own position and environmental perception capabilities on a low-power mobile computing unit.

[0235] The attitude estimation speed of the UAV can reach over 100Hz, far exceeding the speed of other algorithms on mobile computing units. The cumulative position estimation error is reduced to 1m / km, and the attitude estimation error is reduced to 0.5 degrees, enabling high-precision navigation and environmental modeling of the UAV in GPS signal-constrained areas such as the DC field of converter stations.

[0236] 6.2. Based on the UAV attitude and position estimation data, the deviation between the UAV attitude and position estimation data and the locally optimized global inspection trajectory is calculated to obtain the trajectory tracking error.

[0237] 6.3 Based on Lyapunov stability theory, an adaptive error compensation algorithm is used to compensate for trajectory tracking errors and output corrected flight control signals so that the UAV can fly stably on a locally optimized global inspection trajectory under uncertain airflow disturbances.

[0238] The specific steps are as follows:

[0239] First, real-time flight data is input into the G2O graph optimization framework for nonlinear optimization and extended Kalman filtering, outputting UAV attitude and position estimation data. The expression is , and represent the estimated three-dimensional spatial coordinates of the UAV at the current time t.

[0240] Secondly, the locally optimized global inspection trajectory is used as the target trajectory. The expression is: ;

[0241] Based on the target trajectory and real-time position estimation results, the trajectory tracking error is calculated. : .

[0242] Then, based on the trajectory tracking error Construct Lyapunov functions , ;

[0243] For Lyapunov functions Differentiation yields: ;

[0244] The determination is based on Lyapunov stability theory:

[0245] like This indicates that the trajectory tracking error is converging and the system is in a stable or asymptotically stable state. At this time, the original control signal should be kept unchanged.

[0246] like This indicates that the trajectory tracking error has a diverging trend, the system is unstable, and adaptive compensation correction of the control signal is required. An adaptive compensation term is superimposed on the original control signal to obtain the corrected flight control signal. :

[0247]

[0248] In the formula, The original control signal at time t; The adaptive compensation term at time t is expressed as:

[0249]

[0250] In the formula, This represents the estimated value of the airflow disturbance at time t; Represents the adaptive integral gain matrix; The integral term representing the position tracking error; The integral variable represents any moment on the time axis from 0 to t; Indicates time The trajectory tracking error.

[0251] In summary, this invention selects the optimal deployment point set and rationally allocates inspection tasks between the inspection platform and the UAV through a multi-objective optimization method, reducing repetitive inspections and unnecessary waiting time, thereby effectively reducing the overall inspection operation cost. By combining high-precision maps of the converter station, visual SLAM, and dynamic obstacle avoidance mechanisms, it significantly enhances the safety and real-time performance of UAV trajectory planning in complex electrical environments. Employing multi-source sensor fusion positioning technology based on the G2O graph optimization framework, combined with an adaptive error compensation strategy, it enables the UAV to maintain flight trajectory stability and robustness under airflow disturbances and environmental uncertainties. Furthermore, the introduction of endurance determination and inspection task segmentation mechanisms ensures that inspection tasks can still be executed continuously and completely even when single-use endurance is limited, thus significantly improving the intelligence level and operational reliability of the intelligent converter station inspection system that integrates UAVs and the inspection platform.

[0252] Those skilled in the art will understand that embodiments of this application can be provided as methods, systems, or computer program products. Therefore, this application can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, this application can take the form of a computer program product embodied on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0253] This application is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of this application. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart... Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0254] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0255] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0256] The above are merely preferred embodiments of the present invention. It should be noted that those skilled in the art can make various improvements and modifications without departing from the principle of the present invention, and these improvements and modifications should also be considered within the scope of protection of the present invention.

Claims

1. A multi-objective optimized inspection method based on the collaboration between unmanned aerial vehicles (UAVs) and an inspection platform, characterized in that, include: Obtain basic data of the converter station; The basic data is input into the pre-built inspection scheduling model, and the solution is obtained with the goal of minimizing the total operation time of the inspection platform. According to the inspection scheduling scheme, global path search and endurance verification are performed based on the A-star algorithm to generate an initial global inspection track. The initial global inspection trajectory is sent to the UAV, the UAV is controlled to fly according to the initial global inspection trajectory, and the real-time flight data transmitted back by the UAV is received. Based on the real-time flight data, the initial global inspection trajectory is locally replanned using the visual SLAM algorithm and the rolling window mechanism to obtain a locally optimized global inspection trajectory. Using the locally optimized global inspection trajectory as the target trajectory, and based on the G2O graph optimization framework and adaptive error compensation algorithm, a corrected flight control signal is output to enable the UAV to fly stably along the locally optimized global inspection trajectory.

2. The multi-objective optimized inspection method based on the collaboration between UAVs and inspection platforms as described in claim 1, characterized in that, The basic data includes the location data of the equipment to be inspected, the location data of the alternative deployment points, the flight parameters of the UAV, and the constraint data of the inspection workday.

3. The multi-objective optimized inspection method based on the collaboration between UAVs and inspection platforms as described in claim 1, characterized in that, The objective function of the inspection scheduling model is expressed as follows: ； In the formula, Represents the function to be minimized; Indicates the total operation time of the inspection platform; Indicates a weekday gathering; The expression representing the operation time of the inspection platform on the qth working day is as follows: ； In the formula, A binary decision variable representing whether the nth alternative deployment point is selected as the actual deployment point; The binary decision variable represents the inspection platform's access to the nth alternative deployment point on the qth working day. This represents the inspection time of the drone at the nth alternative deployment point. The binary decision variable represents whether the i-th candidate job node is selected as the actual job node; The binary decision variable representing whether the j-th candidate job node is selected as the actual job node; Let $\mathbf{i}$ represent the binary decision variable when the inspection platform travels from the $i$-th alternative work node to the $j$-th alternative work node on the $q$-th working day. This represents the straight-line geographical distance between the i-th candidate job node and the j-th candidate job node; This represents the equivalent time parameter for the deployment and switching of the inspection platform; This indicates the longest working hours for inspection personnel on a weekday; P represents the set of candidate deployment points; G represents the set of candidate job nodes.

4. The multi-objective optimized inspection method based on the collaboration between UAV and inspection platform as described in claim 3, wherein the inspection time of the UAV at the nth alternative deployment point... The expression is: ； In the formula, The task time of the k-th drone at the nth alternative deployment point is expressed as: ； In the formula, A binary decision variable representing whether the nth alternative deployment point is selected as the actual deployment point; Let m represent the m-th equipment to be inspected, which is served by the n-th alternative deployment point; This represents a binary decision variable when the m-th equipment to be inspected is inspected by the k-th drone; This represents the straight-line geographical distance between the m-th device and the n-th alternative deployment point; Indicates the drone's inspection flight speed; The time taken for the drone to inspect the m-th device is represented by ; T represents the set of devices to be inspected.

5. The multi-objective optimized inspection method based on the collaboration between UAVs and inspection platforms according to claim 4, characterized in that, The constraints of the inspection and scheduling model include weekday route start and end constraints, single deployment of deployment points and operation continuity constraints, equipment deployment point correspondence constraints, path endpoint validity constraints, access selection consistency constraints, decision variable constraints, single drone flight and single device constraints, and variable attribute constraints. The expression for the start and end constraints of the weekday route is: ； In the formula, The binary decision variable represents the time when the inspection platform travels from maintenance station a to the j-th alternative work node on the q-th working day. The binary decision variable represents the time when the inspection platform travels from the j-th alternative work node to the maintenance station a on the q-th working day; The expression for the single deployment of the deployment point and the operational continuity constraint is: ； ； In the formula, A binary decision variable representing whether the nth alternative deployment point is selected as the actual deployment point; Let $\mathbf{i}$ represent the binary decision variable when the inspection platform travels from the $i$-th alternative work node to the $n$-th alternative deployment point on the $q$-th working day. Let $\mathbf{ ... The expression for the constraint corresponding to the device deployment point is: ； In the formula, Let m represent the m-th equipment to be inspected, which is served by the n-th alternative deployment point; This indicates the collection of equipment to be inspected; The expression for the path endpoint validity constraint is: ； In the formula, Let $\mathbf{i}$ represent the binary decision variable when the inspection platform travels from the $i$-th alternative work node to the $j$-th alternative work node on the $q$-th working day. The binary decision variable represents the number of times the inspection platform visits the i-th alternative work node on the q-th working day. The binary decision variable represents the number of times the inspection platform visits the j-th alternative work node on the q-th working day. The expression for the access selection consistency constraint is: ； In the formula, The binary decision variable represents the inspection platform's access to the i-th alternative deployment point on the q-th working day. The expression for the decision variable constraint is: ； ； The expression for the constraint on a single UAV flight and a single device is: ； ； In the formula, This represents a binary decision variable when the m-th equipment to be inspected is inspected by the k-th drone; This indicates the collection of equipment to be inspected; This represents the mission time of the k-th drone at the n-th alternative deployment point; The expression for the variable attribute constraint is: ； ； In the formula, This represents the mission time of the k-th drone at the n-th alternative deployment point.

6. The multi-objective optimized inspection method based on the collaboration between UAVs and inspection platforms according to claim 1, characterized in that, The inspection and scheduling scheme includes: The optimal set of deployment points covering all equipment to be inspected; The inspection platform deployment scheme is used to determine the access order and dwell time of the inspection platform to each deployment point in the optimal deployment point set on each working day. The drone inspection task allocation scheme is used to determine the list of equipment to be inspected by each drone at each deployment point.

7. The multi-objective optimized inspection method based on the collaboration of UAV and inspection platform according to claim 6, characterized in that, The generation of the initial global inspection track includes: According to the inspection scheduling scheme, the A* algorithm is used to perform global path search and generate the original global inspection track. Perform a battery life verification on the original global inspection track; If the original global inspection track fails the endurance verification, the optimal deployment point set and the inspection platform deployment scheme remain unchanged. The UAV inspection task allocation scheme is divided into multiple UAV inspection sub-task allocation schemes whose range meets the endurance threshold. The original global inspection track is then generated again using the A* algorithm with each UAV inspection sub-task allocation scheme as input. If the original global inspection track passes the endurance verification, then the original global inspection track will be used as the initial global inspection track.

8. The multi-objective optimized inspection method based on the collaboration between UAVs and inspection platforms according to claim 7, characterized in that, The use of the A* algorithm for global path search includes: Based on the optimal deployment point set of the inspection scheduling scheme and the UAV inspection task allocation scheme, determine the take-off deployment point of each UAV and the corresponding list of equipment to be inspected. Obtain the spatial distribution data of the equipment to be inspected corresponding to the list of equipment to be inspected in the converter station, and cluster the spatial locations of the equipment to be inspected based on spatial proximity to obtain multiple local inspection areas. For each local inspection area, extract the outer edge feature points located at the spatial distribution boundary; construct a local target rectangular region based on the outer edge feature points; For each local target rectangular area, based on the preset imaging parameters of the UAV onboard imaging equipment, the corresponding original local inspection track is generated in an S-shaped routing planning manner; Based on all the original local inspection tracks, the TSP traveler algorithm is used to traverse the shortest loop access order of all local inspection areas to generate the original global inspection track.

9. The multi-objective optimized inspection method based on the collaboration of UAV and inspection platform according to claim 1, characterized in that, The real-time flight data includes environmental data within a scrolling window, which includes image data, inertial measurement data, GPS data, barometric pressure data, and radar data. Based on the environmental data within the scrolling window, the real-time pose data of the UAV in the current environment is calculated and environmental map data is constructed to obtain a local environment model. Based on obstacle detection algorithms and motion prediction algorithms, dynamic obstacles appearing in the local environment model are detected, and the predicted motion trend of the dynamic obstacles is predicted. When the predicted trajectory of the dynamic obstacle interferes with the initial global inspection track, a collision risk is determined. The Morphin local path planning algorithm is used to perform local path correction on the interfered segment of the initial global inspection track to generate a safe path that bypasses the dynamic obstacle. The initial global inspection track is then updated with the safe path to obtain the locally optimized global inspection track. The scrolling window is updated in real time as the drone moves, and the environmental data in the updated scrolling window serves as the input data for the next local replanning cycle.

10. The multi-objective optimized inspection method based on the collaboration of UAV and inspection platform according to claim 1, characterized in that, The output corrected flight control signal includes: The real-time flight data is input into the G2O graph optimization framework, and nonlinear optimization and extended Kalman filtering are performed to output UAV attitude and position estimation data. Based on the UAV attitude and position estimation data, the deviation between the UAV attitude and position estimation data and the locally optimized global inspection trajectory is calculated to obtain the trajectory tracking error. Based on Lyapunov stability theory, an adaptive error compensation algorithm is used to compensate for the trajectory tracking error and output a corrected flight control signal so that the UAV can fly stably according to the locally optimized global inspection trajectory under uncertain airflow disturbances.

Images