Vehicle-machine cooperative power inspection path optimization method and system considering endurance constraint

Abstract

A method and system for optimizing power inspection path of vehicle-machine cooperation considering endurance constraint. A vehicle-machine cooperation space-time network model is constructed, and the inspection task is discretized into a space-time coordinate system to form a graph structure containing vehicle nodes, unmanned aerial vehicle nodes and edge attributes. Multi-dimensional optimization objectives such as maximum inspection coverage, minimum total energy consumption and task time, and maximum multi-unmanned aerial vehicle load balancing are set. Double-layer coding structure is used for iterative optimization. Space partition, collision avoidance constraint and energy consumption prediction are performed on the unmanned aerial vehicle, and resource feasibility is checked to obtain a set of Pareto optimal solutions. When selecting the optimal solution for execution, the flight state is monitored in real time and the actual energy consumption is calculated. If the deviation exceeds the threshold, re-planning is triggered, the model and optimization objectives are updated, and the local path is regenerated to achieve dynamic, cooperative and efficient inspection.

Description

Technical Field

[0001] This invention belongs to the field of vehicle-machine cooperative control technology, specifically involving a vehicle-machine cooperative power inspection path optimization method and system that takes into account range constraints. Background Technology

[0002] With the continuous expansion of power systems, power line inspection work faces increasingly severe challenges. Traditional manual inspection methods are inefficient and risky, while drone inspection, although effective in improving efficiency, is limited by its limited endurance. Vehicle-to-drone (V2D) collaborative inspection, using vehicles as mobile resupply platforms for drones, can effectively extend the drone's operating range and improve overall inspection efficiency. Therefore, developing path optimization methods suitable for this model is of significant practical importance.

[0003] Existing vehicle-to-machine (V2M) collaborative power line inspection path planning methods suffer from the following shortcomings: First, most methods employ a hierarchical optimization strategy, planning vehicle paths first and then allocating drone tasks. This segmented optimization approach struggles to achieve optimal overall system performance and can easily lead to a mismatch between vehicle stop selection and drone endurance. Second, existing methods handle drone endurance constraints in a simplistic manner, typically using simplified models with fixed flight radius or fixed flight time. This fails to fully consider the impact of dynamic factors such as wind speed and load on energy consumption during actual flight, potentially rendering the planned path infeasible in practice. Third, existing optimization methods often focus on a single objective, such as minimizing total time or total distance, making it difficult to balance multiple conflicting objectives such as inspection quality, energy efficiency, and time cost. Finally, most methods lack effective dynamic adjustment mechanisms. When actual execution deviates from the plan, timely online replanning is impossible, impacting system robustness.

[0004] Therefore, to address the above issues, a vehicle-machine collaborative power distribution inspection path optimization method is proposed. This method can effectively coordinate vehicle and drone operations, fully consider the actual range constraints in power distribution scenarios (including the increased flight distance caused by flying around buildings and the additional energy consumption caused by local airflow changes in urban areas), balance multiple objective requirements, and has online adjustment capabilities. Summary of the Invention

[0005] To address the shortcomings of existing technologies, this invention provides a vehicle-machine collaborative power inspection path optimization method that takes into account range constraints. The inspection task is discretized into a spatiotemporal coordinate system, and multi-dimensional optimization objectives are set, including maximizing inspection coverage, minimizing total energy consumption and task time, and maximizing load balancing among multiple drones. A two-layer coding structure is used for iterative optimization, considering spatial partitioning, collision avoidance constraints, and energy consumption prediction for drones, while also checking resource feasibility to find the optimal execution solution.

[0006] The present invention adopts the following technical solution: Construct a vehicle-machine collaborative spatiotemporal network model and formulate spatiotemporal constraints; discretize the inspection task in the spatial and temporal dimensions, and construct a spatiotemporal coordinate system with the adaptive grid and time slice obtained after discretization; define vehicle nodes, UAV nodes and edge attributes, and construct the vehicle-machine inspection process as a graph structure. Based on the vehicle-machine collaborative spatiotemporal network model, the target dimensions of inspection path planning are clarified, and multi-dimensional optimization objectives are formulated, including maximizing the integrity of inspection coverage, minimizing the total system energy consumption, minimizing the total task completion time, and maximizing the load balancing of multiple drones. Based on the spatiotemporal network model and multidimensional optimization objective, a two-layer coding structure is constructed; for each UAV, spatial partitioning and collision avoidance distance constraint checks are performed, flight energy consumption is calculated and predicted, and resource feasibility checks are performed; the two-layer coding structure is iteratively solved and a Pareto optimal solution set is output. The final path planning scheme is selected from the Pareto optimal solution set and executed; during the execution process, the flight status parameters of the UAV are monitored and the actual energy consumption of the flight mission is calculated; the actual energy consumption of the flight mission is compared with the predicted flight energy consumption to calculate the relative deviation; when the relative deviation is greater than the corresponding threshold, the replanning is triggered to update the vehicle-machine cooperative spatiotemporal network model and optimization objective, and the local path planning scheme is solved again.

[0007] The beneficial effects of this invention are that, compared with the prior art, 1. This invention integrates vehicle path planning and UAV task allocation into a single optimization framework by constructing a unified spatiotemporal network model. It employs a network structure where nodes represent spatiotemporal locations and edges represent movement relationships, achieving integrated modeling of collaborative operations between vehicles and UAVs. This modeling approach accurately describes the complex spatiotemporal relationships in power distribution inspection scenarios, such as vehicles traveling on urban roads and multiple UAVs navigating between buildings. Through collaborative edges, it effectively establishes operational connections between vehicles and UAVs at docking points, thereby avoiding suboptimal problems caused by hierarchical optimization, improving overall system performance, and enhancing the feasibility and coordination of the solution.

[0008] 2. This invention transforms the drone's endurance constraint into a connectivity constraint within a spatiotemporal network, establishing an energy consumption model based on actual flight conditions. It comprehensively considers the impact of multiple factors on energy consumption, including flight distance, ambient wind speed, equipment load, and the actual increase in flight distance due to detours caused by building obstructions in power distribution scenarios. This method can accurately assess the energy consumption of drone flight, ensuring that any flight path is within the endurance range. Furthermore, by linking energy consumption with spatiotemporal location, it guarantees the drone's safe return to the vehicle for resupply, significantly improving the feasibility and reliability of path planning and reducing the risk of mission interruption due to insufficient power.

[0009] 3. This invention employs a resource-aware multi-objective optimization algorithm, introducing a resource feasibility check mechanism during the optimization process. It designs a specialized two-layer encoding scheme to represent vehicle paths and UAV task allocation, and develops problem-specific genetic operators. This algorithm design maintains solution feasibility throughout the evolutionary process, effectively handling the complex spatiotemporal constraints of power distribution inspection, where equipment is distributed along streets, is numerous and densely packed, and has a wide range of available stopping points. Simultaneously, it provides multiple trade-off solutions through the Pareto optimal solution set, enabling decision-makers to select the most suitable implementation plan based on actual needs, achieving multi-objective collaborative optimization while ensuring endurance safety.

[0010] 4. This invention establishes a dynamic adjustment mechanism during the execution phase, monitors the UAV's energy consumption in real time, and initiates an online replanning process when significant deviations are detected. Based on the current system state, a local spatiotemporal network model is constructed, and an adjustment plan is quickly generated. This mechanism effectively addresses uncertainties in actual power distribution inspections, including sudden increases in energy consumption due to local airflow turbulence around buildings and actual flight distances exceeding expectations due to the need to detour around temporary obstacles. It ensures the system maintains a feasible solution even when deviations occur, guarantees task completion rates by timely adjusting path plans, enhances the system's adaptability and robustness in complex urban environments, and improves overall operational efficiency. Attached Figure Description

[0011] Figure 1 This is a diagram of the overall architecture of the vehicle-machine collaborative power inspection system of the present invention. Figure 2 Detailed diagram of the spatiotemporal network model construction of this invention; Figure 3 This is a diagram illustrating the multi-objective optimization and dynamic adjustment mechanism of the present invention. Detailed Implementation

[0012] To make the objectives, technical solutions, and advantages of this invention clearer, the technical solutions of this invention will be clearly and completely described below with reference to the accompanying drawings of the embodiments of this invention. The embodiments described in this application are merely some embodiments of this invention, and not all embodiments. Based on the spirit of this invention, all other embodiments obtained by those skilled in the art without creative effort are within the protection scope of this invention.

[0013] With the continuous expansion of power distribution networks, the inspection of power distribution equipment faces increasingly severe challenges. Power distribution equipment is numerous and widely distributed, situated in complex environments such as urban blocks, industrial parks, and residential communities. Traditional manual inspection methods require inspectors to visit each equipment point individually, resulting in low efficiency, long inspection cycles, and poor accessibility for equipment located in narrow alleys or enclosed areas. Furthermore, manual inspections often fail to detect defects in concealed parts of the equipment. While drone inspection can effectively improve efficiency, its limited flight time makes it difficult to cover all equipment in a single flight, especially given the dense distribution of power distribution equipment within an area. The vehicle-to-drone (V2V) collaborative inspection model uses vehicles as mobile resupply platforms for drones, enabling them to operate from multiple docking points. This effectively extends the drone's operational range and improves overall inspection efficiency. Therefore, developing path optimization methods suitable for this model is of significant practical importance.

[0014] In power distribution scenarios, power distribution equipment is densely distributed along streets and buildings in a planar pattern. While there is a wide range of docking points to choose from, it is necessary to consider the obstruction of the drone's line of sight by buildings and the obstruction of its flight path. The problem of the disconnect between the selection of docking points and the actual reachable area of ​​the drone is particularly prominent due to the hierarchical planning. Existing methods handle the drone's endurance constraints in a relatively simple way, usually using a simplified model with a fixed flight radius or fixed flight time. This fails to fully consider the impact of multiple factors on energy consumption, such as the increase in actual flight distance due to building obstruction causing the flight path to become a detour, the additional power consumption caused by local airflow changes between buildings in urban areas, and the impact of load. This may lead to the infeasibility of the planned path in actual execution. Most methods lack an effective dynamic adjustment mechanism. When the actual execution deviates from the plan, such as a sudden strong wind in the urban area or temporary obstacles causing a sharp increase in energy consumption, online replanning cannot be performed in time, affecting the robustness of the system.

[0015] Example 1 As one embodiment of this application, as shown in the appendix Figure 1 The vehicle-machine collaborative power inspection path optimization method shown is based on UAV endurance constraints. By establishing a unified spatiotemporal network model, the collaborative operation process of vehicles and multiple UAVs is integrated into the same optimization framework. This model can accurately describe the relationship between various elements of the system in the spatiotemporal dimension.

[0016] This embodiment provides a vehicle-to-machine (V2M) collaborative power grid inspection route optimization method that takes into account range constraints. This method addresses the following problem: An inspection vehicle carries multiple drones to inspect various power distribution equipment (such as distribution transformers, ring main units, switch stations, and power line towers) distributed along streets and buildings within a power distribution network area. Under the constraint of drone range limitations, it is necessary to plan the vehicle's driving route and parking scheme, as well as the flight inspection path of each drone, so that the inspection task can achieve multi-objective optimization under safe and feasible conditions, even in urban environments with building obstructions and local airflow changes.

[0017] This method integrates the collaborative operation of inspection vehicles and multiple drones into a single optimization framework by establishing a unified spatiotemporal network model. The entire system consists of inspection vehicles, multiple drones, power distribution equipment, and a path planning and monitoring center. The inspection vehicles serve as mobile supply platforms, traveling on urban roads and parking at pre-designated stops; drones take off from the inspection vehicles to perform power distribution equipment inspection tasks, flying between buildings, and returning to the inspection vehicles to replace batteries or recharge after completing their tasks.

[0018] S1: In one implementation, refer to Figure 2 A vehicle-machine collaborative spatiotemporal network model is constructed, the inspection task is discretized in the spatial and temporal dimensions, the node types and edge types in the network are defined, and spatiotemporal constraints are established.

[0019] This step transforms the continuous inspection task in the real physical world into a discretized graph structure model that can be processed by a computer. In the initial modeling phase, the continuous inspection task needs to be discretized in both spatial and temporal dimensions. Spatial discretization uses a quadtree mesh generation method. The mesh size needs to comprehensively consider the terrain features and equipment distribution density of the local area; finer meshes are used in areas with dense power distribution equipment or complex terrain to ensure inspection accuracy, while coarser meshes are used in areas with sparse power distribution equipment and flat terrain to reduce computational burden.

[0020] 1.1: In a further implementation, the initial basic grid size is calculated based on the density of power distribution equipment in the spatial dimension: ; Indicates the initial grid size. and These are the upper and lower limits of the grid size, respectively; This represents the distribution density of power distribution equipment in the local area where the grid to be calculated is located. This represents the maximum equipment density for the entire inspection area. The extent of a local area is determined by the current level of the quadtree partitioning. During each recursive partitioning, the current area is divided into four sub-regions, and the number of devices and density in each sub-region are calculated. The termination condition for quadtree recursive partitioning is: when the initial grid size of the sub-region is calculated according to the above formula... Less than or equal to When the number of power distribution devices in a sub-area is less than or equal to 1, the division shall be stopped.

[0021] When local equipment density Approaching the global maximum value When, the ratio Approaching 1, the grid size approaches 1. To achieve fine division; when When the size is small, the grid size tends to be close to A coarser partitioning is used. This adaptive mechanism ensures sufficient spatial accuracy in critical areas (where equipment is dense or the terrain is complex) while maintaining a reasonable computational load in non-critical areas.

[0022] In a further implementation, the initial grid size described above only considers the density of power distribution equipment and does not yet take into account the limitations imposed on UAV flight by urban environmental factors. In complex urban environments with dense buildings and narrow streets, the UAV flight path is constrained by building obstructions and street orientation, requiring a finer spatial resolution to ensure flight safety and accurate path planning. Therefore, this embodiment introduces a terrain complexity factor to correct the initial grid size.

[0023] Calculate the terrain complexity factor using a digital elevation model. : ; in, This represents the number of points within the current grid area. and Points within the grid Elevation and grid average elevation.

[0024] when When the terrain complexity threshold is preset, even with sparse equipment, local densification is still required (e.g., size reduction of 30%~50%); in this case, the final grid resolution is determined jointly by the equipment and the terrain. ; in, For the final grid resolution, The terrain correction factor is less than 1.

[0025] In further implementation details, the grid size needs to satisfy the field-of-view coverage constraint of the UAV sensor: assuming the sensor's field-of-view radius is... Then the grid size needs to meet the following requirements. This constraint requires that the diagonal length of the grid does not exceed the sensor's field of view diameter. This ensures that when the drone performs inspections at the center of two adjacent grids, the sensor coverage areas of the two grids can overlap, leaving no blind spots.

[0026] 1.2: In a further implementation, the continuous time axis of the entire inspection task is divided into several time slices in the time dimension.

[0027] The duration of each time slice is marked as follows: The unit is minutes. The value of needs to balance the requirements of time resolution and computational complexity: if the value is too large, the time precision is insufficient and it cannot accurately reflect the rapidly changing system state; if the value is too small, the number of time slices increases sharply, resulting in an excessively large spatiotemporal network, making it difficult to solve subsequent optimization algorithms.

[0028] In this embodiment, a routine inspection scenario Set to 5 minutes. The system will automatically update when the real-time monitoring data meets any of the following conditions. Adjusted to 3 minutes: Condition 1: Within a 30-minute time window, starting from the current moment, the difference between the maximum and minimum ambient wind speeds exceeds 5 meters per second. Condition 2: The task priority attribute value of the current inspection task is 1. The task priority attribute is set by the operator according to the urgency of the task during task initialization, and the value is 0 (normal task) or 1 (high priority task).

[0029] When neither of the above two conditions is met Keep the default value for 5 minutes.

[0030] The time span of the entire inspection task (from the start of the task) Until the latest allowed end time of the mission By time slice length Divide the time slice into equal intervals and number them sequentially as time slice 1, time slice 2, ..., time slice Total number of time slices .

[0031] Each time slice corresponds to a discrete moment, meaning the start time of each time slice is its timestamp. For example, the timestamp of time slice 1 is... The timestamp for time slice 2 is And so on. When constructing the spatiotemporal network subsequently, the nodes'... The attribute records the timestamp of the time slice to which the node belongs.

[0032] 1.3: In a further implementation, a spatiotemporal network model for vehicle-machine collaboration is constructed; a spatiotemporal coordinate system is constructed using an adaptive grid and time slices, and the vehicle-machine inspection process is abstracted into a graph structure using graph theory methods.

[0033] (1) In the graph structure, node types include vehicle nodes and UAV nodes; where: Vehicle nodes are defined as triples This indicates that the vehicle is in the time slice. Time located in grid coordinates A new vehicle node is generated at each time slice as the vehicle moves. The drone node is a quintuple ; indicates that the drone is in the time slice Located in grid coordinates Energy is The load is As the drone flies from one grid to another, a new drone node is generated at the new location after each time slice. It should be noted that the payload in the drone node... The weight of mission equipment, excluding the drone's own weight (aircraft structure, power battery, and flight control system), such as the total weight of sensing devices used to collect status information of power distribution equipment, including visible light cameras, infrared thermal imagers, and lidar.

[0034] In a vehicle-machine collaborative inspection system, a vehicle is equipped with multiple drones (usually 2 to 4). Therefore, the total number of drone nodes is approximately equal to the product of the number of vehicle nodes and the number of drones.

[0035] (2) In the graph structure, edge types include vehicle movement edges, UAV flight edges, and collaborative edges; (2.1) Vehicle movement edge: connects two vehicle nodes of the same inspection vehicle in adjacent time slices; the set of all vehicle movement edges is labeled as follows. .

[0036] For example, the inspection vehicle in the time slice Located at grid position coordinates Corresponding vehicle node In time slice Move to grid position coordinates Corresponding vehicle node Then there exists a path from... point to The vehicle movement side indicates that the inspection vehicle has completed the process from [location] during this period. arrive Spatial movement. It should be noted that here... and The grids must be adjacent or connected by roads, and vehicles can only travel on permitted road networks. In the power distribution inspection scenario, the road network is an urban road network with dense intersections, providing a wide range of vehicle parking spots. However, some narrow streets may restrict the passage of inspection vehicles.

[0037] The properties of the vehicle's moving edge include: Spatial distance between two vehicle nodes The unit is kilometers. The distance traveled is calculated based on the actual road distance between the center points of two grid points, not the straight-line distance between the two points. Vehicle movement time consumption The unit is minutes. Travel time is calculated by dividing the spatial distance by the average vehicle speed, and must be an integer multiple of the time slice length. Mobile energy consumption The unit is kilowatt-hour. In this embodiment, the inspection vehicle is calculated as a pure electric vehicle, and the unit energy consumption coefficient is determined according to the specific vehicle model.

[0038] (2.2) UAV flight edge: connects two UAV nodes of the same UAV in adjacent time slices; the set of all UAV flight edges is denoted as .

[0039] For example, a drone in a time slice Located at grid position coordinates Corresponding drone node In time slice Fly to grid position coordinates Corresponding drone node Then there exists a path from... point to The drone's flight path indicates that the drone completed its journey from [location] during this period. arrive Space flight. Drones can fly between any two grids, unrestricted by road networks, but limited by their endurance. In power distribution inspection scenarios, the actual flight path of a drone is affected by the distribution of buildings, and it may need to fly around tall buildings, with the actual flight distance potentially exceeding the straight-line distance between the two points.

[0040] The attributes of the drone's flight path include: Spatial distance between two nodes The unit is kilometers. This distance is calculated based on the straight-line distance between the center points of two grids.

[0041] Flight time consumption The unit is minutes. Flight time is obtained by dividing the spatial distance by the drone's cruising speed.

[0042] Predicted flight energy consumption The unit is ampere-hours (Ah). This value is not a direct measurement, but a theoretical estimate calculated based on the energy consumption model during the path planning phase. It is used to determine whether the flight path theoretically meets the range constraints. The specific calculation formula for the energy consumption model is given in subsequent step S3.

[0043] Inspection Equipment Identifier List: This list records the globally unique index numbers of each power distribution device covered by the UAV sensors within the spatial range traversed by this flight edge. One flight edge can cover multiple power distribution devices, and a power distribution device can also be covered by multiple flight edges. This list is a key data structure connecting "UAV flight actions" and "inspection task coverage determination": when calculating the inspection coverage integrity target in step S2, by traversing the inspection equipment identifier list of all selected flight edges, it is possible to count which power distribution devices have been inspected.

[0044] (2.3) Collaborative Edge: A collaborative edge connects a vehicle node and a UAV node located at the same time slice and spatial grid position. If multiple UAVs are docked at the same spatiotemporal location, there is an independent collaborative edge between each UAV and the vehicle node. The collaborative edge has no directionality, indicating that the spatiotemporal location is a vehicle-UAV collaborative operation point, i.e., a docking point. At the docking point, the UAV can take off from the inspection vehicle to start performing the inspection task, or land back in the inspection vehicle for refueling after completing the inspection task.

[0045] The properties of collaborative edges include: Service Hours The unit is minutes. This refers to the total time from when the drone lands and touches the inspection vehicle until it completes takeoff preparations (including battery removal and replacement or charging, equipment checks, and pre-takeoff self-checks). If multiple drones need to be resupplyed sequentially at the same docking point and within the same time period, the service time must also include the queuing time for each drone. The service time directly affects the length of time the inspection vehicle stays at that docking point.

[0046] Energy replenishment The unit is ampere-hours (Ahs). Energy replenishment refers to the additional usable energy the drone gains at the docking point by replacing its battery with a fully charged one or by using a fast-charging device. In this embodiment, it is assumed that the drone is fully charged after replenishment at the docking point, meaning the drone's remaining battery power equals its full battery capacity. .

[0047] Based on the above definitions of nodes and edges, the physical process of vehicle-machine collaborative inspection is completely mapped into a spatiotemporal network graph. The path selection problem on the [website / platform]. The node set [is involved]. edge set The subsequent optimization task is to select a set of edges from the graph to form a vehicle-machine collaborative operation scheme.

[0048] 1.4: In a further implementation, to ensure the safety and feasibility of the inspection task in actual execution, spatiotemporal constraints are established for the vehicle-machine collaborative spatiotemporal network model.

[0049] (1) Connectivity Constraint: After each flight inspection mission, the UAV must be able to rendezvous with the inspection vehicle at the same time and location, and land on the inspection vehicle for refueling. If the UAV cannot find the inspection vehicle at the end of its flight, it will crash due to power depletion. Therefore, this constraint is the core constraint to ensure the safe operation of the system. The mathematical form of this constraint is: For any drone flight edge in the spatiotemporal network that is selected as the execution plan, , and For the starting and ending drone nodes of this flight edge, there must be a vehicle node. Simultaneously satisfying: Condition 1: The destination drone node and the vehicle node belong to the same time slice, i.e. The symbol .t indicates that the timestamp attribute value of the node is retrieved.

[0050] Condition 2: The endpoint drone node and the vehicle node are located in the same spatial grid, i.e. and The symbols .x and .y represent the x-coordinate and y-coordinate attribute values ​​of the node, respectively.

[0051] When both conditions one and two are met, the endpoint drone node With vehicle nodes There must be a collaborative edge between them, through which the drone can land on the inspection vehicle.

[0052] The connectivity constraint ensures that each selected UAV flight edge in the spatiotemporal network can establish a collaborative connection with the inspection vehicle at its endpoint. In the optimization algorithm of step S3, this constraint is enforced as a hard condition during the decoding and feasibility check stages.

[0053] (2) Flight time constraint: The duration of any single continuous flight of the UAV shall not exceed the maximum flight time supported by its battery.

[0054] Set up a drone flight path The flight time consumption attribute is The maximum single-charge endurance time specified in the drone technical manual is In this embodiment Set to 45 minutes.

[0055] The flight time constraint is expressed by the following inequality:

[0056] This constraint is verified one by one during the flight edge check mechanism in step S3.

[0057] (3) Cumulative energy consumption constraint: For each UAV at any point during the entire inspection mission, the cumulative flight energy consumed must not exceed the remaining power of the UAV when it departs from the inspection vehicle, and the energy is reset after refueling at the docking point. This constraint ensures that the UAV's remaining power is always sufficient to support the completion of any segment of flight.

[0058] The formal expression of this constraint is: For any UAV, let the sequence of flight edges it executes in this round of mission be... ,in This represents the total number of flight segments for the drone in this mission, interspersed with several resupply stops (i.e., resetting remaining battery power via cooperative edges). For each flight segment in the sequence... ( All of the following must be satisfied: and

[0059] in For the first Predicted flight energy consumption for the segment of flight. This represents the drone's remaining battery power before the start of this flight segment. This represents the flight time consumed during this flight segment. Remaining battery power. For details on the step-by-step verification process of the recursive calculation rules and constraints, please refer to step S3.2.

[0060] S2: In one implementation, refer to Figure 2 Based on the spatiotemporal network model constructed in S1, the target dimensions that need to be optimized for the inspection path planning problem are clarified, and the multi-dimensional optimization objectives that are mutually constrained are defined.

[0061] This embodiment defines four optimization objectives, including: maximizing the integrity of inspection coverage, minimizing the total system energy consumption, minimizing the total task completion time, and maximizing the load balancing of multiple drones.

[0062] 2.1: In a further implementation, the optimization objective—maximizing the integrity of inspection coverage—is explained; The physical meaning of inspection coverage integrity is: among all power distribution equipment to be inspected, the cumulative importance score of the successfully inspected equipment out of the total importance score of all equipment. This goal aims to cover as many high-importance equipment as possible within limited battery life and time.

[0063] Inspection coverage integrity Defined as: ; in, This represents the total number of power distribution devices to be inspected within the inspection task area. Index for power distribution equipment; Indicating the first step in the inspection task The inspection status variable for each power distribution device is 1, which indicates that the power distribution device has been covered by the inspection of at least one drone flight edge in the path planning scheme, and 0 indicates that it has not been covered. For power distribution equipment The device importance weight is a positive real number that is actually assigned.

[0064] In a further implementation, the inspection state variable The calculation method is as follows: traverse all selected drone flight edges in the path planning scheme. Extract the list of inspection equipment identifiers for each flight edge, if the power distribution equipment If the index number appears in the inspection equipment identifier list of at least one flight edge, then ,otherwise In mathematics, this is represented by the logical OR operation: ; in Indicates flight edge The list of inspection equipment identifiers, with the condition in square brackets, has a value of 1 when true and 0 when false.

[0065] It should be noted that in the optimization objective formula... It is not an arbitrary subset of flying edges in the spatiotemporal network, but rather a set of feasible flying edges obtained through optimization in step S3, which simultaneously satisfies the connectivity constraints and flight time constraints in step S1.4 and passes the resource feasibility check in step S3.2. Therefore, the value of the objective function is naturally based on the physical feasibility of the solution.

[0066] In a further implementation, the device importance weight Based on the importance of power distribution equipment in the power system, they are divided into three levels, and the specific assignment rules are as follows: Level 1 critical equipment refers to key node equipment such as distribution transformers, ring main units, and switch stations. These devices undertake the core functions of power distribution and voltage transformation; a failure in these devices would cause widespread power outages or serious safety accidents. The weight of Level 1 critical equipment is assigned within the range [2.0, 3.0) based on the specific device's criticality within the system.

[0067] Level 2 critical equipment refers to important power distribution line towers, pole-mounted switches, cable distribution boxes, and other equipment that plays a vital role in the normal operation of power distribution lines. Failures of these devices will affect the power supply reliability of local areas. The weights of Level 2 critical equipment are assigned within the interval [1.0, 2.0).

[0068] Level 3 critical equipment refers to a large number of conventional equipment such as power line towers, low-voltage distribution devices, and insulators, where the impact of a single equipment failure is limited. The weight of Level 3 critical equipment is assigned within the range (0, 1.0).

[0069] 2.2: In a further implementation, the optimization objective—minimizing the total system energy consumption—is explained; The physical meaning of total system energy consumption is: the weighted sum of energy consumed by the inspection vehicles and all drones to complete the entire inspection task. The purpose of setting this target is to save energy consumption costs as much as possible while ensuring inspection quality.

[0070] Total system energy consumption Defined as: ; in, It is the set of all selected vehicle movement edges in the path planning scheme. Which vehicle movement edges are selected is determined by the vehicle path after encoding and decoding in the upper layer. For set One of the vehicle movement edges in the middle; Vehicle movement side The mobile energy consumption attribute value, in kilowatt-hours, is given by the definition of the vehicle movement edge in step S1.3; The set of all selected drone flight edges in the path planning scheme. It is a collection of all drone flight paths. A subset. Which drone flight edges are selected is determined by the drone task allocation and flight path after lower-level encoding and decoding. For set A drone flying along the edge of the road, Drone flight edge The predicted flight energy consumption attribute value, in ampere-hours. This value is calculated using the energy consumption model formula in step S3.

[0071] This is the vehicle energy consumption weighting coefficient, used to adjust the relative importance of vehicle energy consumption in the total energy consumption target. In this embodiment... Set to 0.3. This is the drone energy consumption weighting coefficient, used to adjust the relative importance of drone energy consumption in the total energy consumption target. In this embodiment... Set to 0.7.

[0072] In a further reference implementation, the rationale for setting the two types of energy consumption weight coefficients is given. In the power grid inspection scenario, the energy density of drone batteries is limited, and the time and management costs of each charging or battery replacement are relatively high; while the charging and refueling of inspection vehicles are relatively convenient. Therefore, a higher weight is given to the energy consumption of drones in the objective function, making the optimization algorithm more inclined to select the solution with lower drone energy consumption during the search process.

[0073] 2.3: In a further implementation, the optimization objective—minimizing the total task completion time—is explained; The physical meaning of the total mission completion time is: the total length of time elapsed from the start of the inspection mission to the moment when the last drone completes all its tasks and returns to the inspection vehicle.

[0074] Total task completion time for: ; Indicates the flight path of the drone The timestamp attribute value of the endpoint drone node, that is, the time when the flight edge is completed. The calculation takes the maximum value among all the selected drone flight edges at that moment, which is the moment when the last drone completes its final flight. This indicates the planned start time of the inspection mission.

[0075] It should be noted that the maximum return time of all selected flight edges is used to calculate the total mission completion time because, in the model, the inspection vehicle's operation time is completely covered by the drone's operation time (the inspection vehicle always arrives at the docking point before the drone). Therefore, the end of the entire mission is marked by the moment when the last drone safely returns to the inspection vehicle.

[0076] 2.4: In a further implementation, the optimization objective—maximizing the load balancing of multiple drones—is explained.

[0077] The physical meaning of multi-drone load balancing is: the degree to which the cumulative flight workload of multiple drones is similar. The purpose of introducing this objective is to avoid situations where some drones bear excessive workloads (frequent flights, rapid battery depletion) while other drones have very little workload, thereby extending the average lifespan of drones and improving the overall efficiency of multi-drone parallel operations.

[0078] Load balancing Defined as: ; is the coefficient of variation, where This represents the average total flight distance of all drones. The standard deviation of the total flight distance of each UAV is calculated as follows: Let the total number of drones participating in this inspection mission be... In this embodiment The index for each drone is used express, .

[0079] No. Cumulative mission volume of drones Defined as: the sum of the spatial distances of all drone flight edges to which the drone is assigned to perform its missions.

[0080] in The selected path planning scheme and the one chosen by the first The flight path of the drone is a collection of data. For flight edge The spatial distance attribute value, in kilometers.

[0081] all Average number of missions per drone for:

[0082] all Standard deviation of cumulative missions of drones for:

[0083] Will and Substituting into the definition of load balancing: The range of this formula is [0,1]. This is explained as follows: when all drones have exactly the same workload, the standard deviation... , To achieve maximum balance; when the workload of each drone differs more significantly, the standard deviation increases. The larger the ratio The closer to 1, The closer it gets to 0.

[0084] S3: In one implementation, based on the spatiotemporal network model of S1 and the optimization objective function of S2, the Pareto front is solved, and the path planning scheme is output.

[0085] 3.1: This embodiment adopts a two-layer coding structure. The upper layer coding describes the docking point access scheme of the inspection vehicle, and the lower layer coding describes the inspection task allocation scheme of the UAV starting from each docking point. (1) The upper-layer encoding describes the sequence of vehicle stop points, using a length of... sequence of integers This indicates that the vehicle is... The order of visits to each stop; sequence length This indicates the total number of stops planned for this inspection mission. In a typical 50km transmission line inspection scenario... The value range is 20 to 35. The specific value is set by the operator during algorithm initialization based on the size of the inspection area and the number of devices to be inspected, and serves as the input parameter of the algorithm.

[0086] Sequence elements ( ) is an integer representing the first digit of the first digit. The vehicle node corresponding to each stop is in the vehicle node set. The index number in the sequence. The order of the sequence elements is the order in which the inspection vehicles visit the stops.

[0087] For example: upper-level encoding This means that the inspection vehicle will first go to the vehicle node with index number 5 to stop, then go to the vehicle node with index number 12 to stop, then go to the vehicle node with index number 3, and so on.

[0088] During decoding, the upper-layer encoder connects adjacent vehicle nodes sequentially according to the sequence. Specifically, starting from the initial vehicle node at the task's starting point (usually the initial parking position of the inspection vehicle), it searches for vehicle movement edge sequences in the spatiotemporal network that connect the current vehicle node to the next sequence of vehicle nodes, generating a set of selected vehicle movement edges. At the same time, each vehicle node in the sequence corresponds to a stop, thus determining... The spatial location and arrival time window of each stop.

[0089] (2) The lower-level encoding describes the flight mission allocation matrix, with a OK column binary matrix Represented by; where the number of rows in the matrix The number of columns in the matrix is ​​consistent with the total number of docking points in the upper-level encoding. The total number of electrical devices to be inspected as defined in step S2, objective one; elements in the matrix For 0-1 variables, when Indicates from the first The drones departing from the first docking point are responsible for inspecting the first... One electrical device, This indicates a lack of responsibility. Matrix A constraint must be met: each column must have exactly one element that is 1, i.e. For each All of these conditions are met. The physical meaning of this constraint is that each power device must be assigned to one and only one docking point, from which a drone is dispatched to conduct inspections, thus avoiding duplication or omission of tasks.

[0090] During decoding, the lower-level encoder performs decoding for each stop. Gathering all satisfaction electrical equipment This represents the set of equipment that the drone needs to inspect starting from this docking point. Then, combining the location of this docking point and the location of the next docking point, the drone's flight path is generated using the following path construction algorithm: Algorithm input: Vehicle node at the starting stop The vehicle node at the next stop The current docking point contains a collection of electrical equipment that needs to be inspected. .

[0091] Algorithm output: From Departure, Coverage All equipment in the middle, finally arriving The drone flight sequence.

[0092] Step 1: Set the drone's current location as The corresponding spatial grid coordinates will be used to set the list of remaining equipment to be inspected. The flight edge sequence is initialized to an empty set.

[0093] Step 2: From the list of remaining devices to be inspected, select the grid containing the device that is closest to the current location of the drone as the next target grid, generate the drone flight edge from the current location to the target grid, add the flight edge to the flight edge sequence, update the current location of the drone to the target grid, and remove the device from the list of remaining devices to be inspected.

[0094] Step 3: Repeat step 2 until the list of remaining devices to be inspected is empty.

[0095] Step 4: Generate a path from the grid containing the last device to... The drone flight edges in the grid are added to the flight edge sequence.

[0096] Step 5: Output the complete drone flight side sequence.

[0097] The set of drone flight edges generated at each docking point is the sum of the selected drone flight edges. The upper-layer and lower-layer coding work together: the upper-layer coding determines when and where the inspection vehicle will arrive; the lower-layer coding determines which power equipment the drone needs to fly to for inspection at each stop. Together, the two layers of coding uniquely determine a complete vehicle-machine collaborative path scheme.

[0098] 3.2: In a further implementation, a resource feasibility check is performed on each UAV flight path output in 3.1 to determine whether the path is physically safe to fly, i.e. whether it meets the constraints of UAV endurance and flight time.

[0099] For each drone flight edge obtained from decoding Its predicted flight energy consumption is calculated using the following energy consumption model formula: ; in, Indicates the flight path of the drone The predicted flight energy consumption is expressed in ampere-hours (AH). This value is a theoretical value calculated based on planning parameters during the path planning phase, not an actual measured value. This represents the spatial distance of the flight edge, in kilometers. It is read from the attributes of the flight edge. The wind speed is measured in meters per second. This data is obtained from meteorological forecasts and is used as a prediction parameter during the path planning phase. The total weight of the mission payload carried by the drone, in kilograms. From the drone node... Attribute reading. This represents the flight time consumption of this flight edge, in hours. It is read from the flight edge's attributes and converted to hours. Note that all values ​​used in the energy consumption model calculation are normalized values.

[0100] This is the distance energy consumption coefficient, measured in ampere-hours per kilometer. In this embodiment, it is set to 0.2. The wind speed energy consumption coefficient is expressed in ampere-hours per meter per second. In this embodiment, the value is taken as 0.15. The load energy consumption coefficient is expressed in ampere-hours per kilogram. In this embodiment, it is set to 0.3. The above three coefficients are determined by fitting the flight performance parameters provided by the UAV manufacturer or actual field flight calibration test data.

[0101] Furthermore, a resource feasibility check is performed on each individual (i.e., the drone path plan): (1) Assume that a drone performs the following tasks in this round of missions sequentially. Segment flight, number Segment flight corresponds to the flight path of the drone , The predicted flight energy consumption for each flight edge is calculated using the energy consumption model described above. .

[0102] (2) Determine the initial state. Before the UAV departs from the first docking point, the battery is replaced via a collaborative process, and the remaining power is set to full capacity. In this embodiment The value is 10 amp-hours.

[0103] (3) Track the changes in electricity consumption segment by segment. For the first segment... Duan Feixing: Record the remaining battery power of the drone at the start of this flight segment. .

[0104] If flying side If the starting point is a docking point (i.e., the starting drone node is connected to a vehicle node through a collaborative edge), then the remaining battery power is first reset to full capacity. .

[0105] If flying side If the starting point is not a docking point (i.e., the drone flies directly to the next device after flying to multiple devices), then the remaining battery power will be the remaining value after the previous flight segment ends. .

[0106] To more clearly illustrate the power tracking process in continuous flight scenarios, a typical example is given below. Suppose a drone departs from docking point A, plans to inspect device 1, device 2, and device 3 in sequence, and finally flies to docking point B. This flight sequence contains four drone flight edges: Fly from docking point A to equipment 1. Fly from device 1 to device 2. Fly from device 2 to device 3. Fly from device 3 to docking point B.

[0107] First flight The starting point is stop point A, and the remaining battery power is reset to [value missing]. Anshi. Assume The remaining battery power after the flight is 10 - 1.5 = 8.5 amp-hours.

[0108] Second flight The starting point is not a stop point; the initial remaining battery power is... Anshi. Assume The remaining battery level after the flight was 7.7 amp-hours.

[0109] Third stage of flight The starting point is not a stop point; the initial remaining battery power is... Anshi. Assume The remaining battery level after the flight was 6.5 amp-hours.

[0110] Fourth stage of flight The starting point is not a stop point; the initial remaining battery power is... Anshi. Assume After the flight concludes and the aircraft arrives at docking point B, the remaining battery level is 4.5 amp-hours. At docking point B, the remaining battery level is reset via a cooperative edge. An Shi.

[0111] Before each flight begins, a violation check is performed as follows (4).

[0112] (4) For the first Normalization violation of segment flight calculation :

[0113] The first term in the formula is the energy consumption violation rate, which measures whether the predicted energy consumption for this flight segment exceeds the current remaining battery power (by what percentage); the second term is the duration violation rate, which measures the duration of this flight segment. Has the maximum battery life been exceeded? . The operation indicates that the term is 0 if no constraint is violated.

[0114] (5) All of the drone's total The violation rate of each flight segment is used to obtain the sum of the violation rates of the drone. (6) Sum the violation scores of all UAVs to obtain the total violation score of the individual (path planning scheme). :

[0115] (7): Assess individual feasibility. If This indicates that the path planning scheme represented by this individual violates the endurance constraint or flight time constraint. Therefore, this individual is determined to be an infeasible solution and will be directly eliminated in subsequent algorithmic selection operations, not entering the next generation of the population. Individuals are considered as feasible solutions and enter into non-dominated sorting and Pareto front construction.

[0116] 3.3: In a further implementation, to ensure the safety of multiple drones flying simultaneously, this embodiment performs spatial partitioning and collision avoidance distance constraint checks before individual decoding.

[0117] Furthermore, when decoding the lower-level encoding, a quadtree spatial partitioning algorithm is invoked. This algorithm recursively divides the spatial grid of the inspection area into several non-overlapping flight corridors based on equipment distribution and terrain features. The termination condition for the quadtree recursive partitioning is: the number of equipment to be inspected in each flight corridor is basically balanced (the difference in the number of equipment between corridors does not exceed 20%), or the recursion depth reaches a preset upper limit (set to 5 layers in this embodiment). The number of flight corridors is equal to the number of UAVs. Each flight corridor consists of a set of continuous spatial grid cells, allocated to a single UAV for independent use throughout its mission cycle. A buffer zone with a width of at least 20 meters is established between the boundaries of different flight corridors to reduce the risk of collisions from a spatial structural perspective.

[0118] During the execution of the path construction algorithm in S3.1, the flight corridors pre-divided in this step serve as spatial constraints, and each UAV can only generate flight edges within the grid cells contained in the flight corridor assigned to it.

[0119] It should be noted that the quadtree spatial partitioning algorithm in this step differs from the quadtree adaptive partitioning algorithm used in step S1.1 for determining the grid resolution in terms of purpose and partitioning object: S1.1 partitions the space itself into layers to determine the grid density, while this step groups the partitioned grid cells to form flight corridors.

[0120] Furthermore, based on spatial partitioning, collision detection is performed on the temporal dimension of all drone flight paths after decoding. The detection rule is: for any two drones... and ( , If in the same time slice If both are in flight (i.e., both have drones flying by during that time slice), then the spatial coordinates are read from their respective drone node attributes. and Calculate the Euclidean distance:

[0121] Requirements must be met:

[0122] Safety distance in this embodiment The value is 5 meters.

[0123] If the distance between two drones within a certain time slice is detected to be less than the safe distance, the flight plan is deemed to have a collision risk. The solution is as follows: the algorithm attempts to adjust the flight sequence of one of the drones (delay takeoff or adjust the flight altitude). If the adjustment still cannot eliminate the conflict, that drone is eliminated.

[0124] 3.4: In view of the special characteristics of the two-layer coding structure, this embodiment designs corresponding crossover operators and mutation operators for the upper and lower layers respectively.

[0125] (1) Genetic operators encoded in the upper layer; The upper-level encoding is an ordered integer sequence, which is suitable for genetic operators with permutation encoding.

[0126] Crossover operator – Partial matching crossover: Two crossover positions are randomly selected, and gene segments between the parent individuals P1 and P2 are interchanged. After the interchange, duplicate genes may appear in the offspring individuals (i.e., the same vehicle node index appears twice in the sequence). A mapping repair strategy is used to eliminate duplicates: a gene correspondence is established between the interchanged segments, and duplicate genes in the offspring sequence other than the interchanged segments are replaced according to the correspondence to ensure that the final sequence is from 1 to... An effective permutation that does not repeat within the permutation.

[0127] Mutation operator – Reverse mutation: Randomly selects a continuous segment from the upper-level encoded sequence with a preset probability (set to 0.05 in this embodiment) and reverses the order of elements within that segment. For example, the sequence... Select subfield After mutation, it becomes This operation alters the order in which inspection vehicles visit intermediate stops, thereby exploring different vehicle routing options.

[0128] (2) Genetic operators encoded in the lower layer The lower-level encoding is a binary matrix, and it satisfies the constraint that each column contains exactly one 1.

[0129] Crossover operator – uniform column swap: For the task allocation matrix of the two parent individuals, each column is judged with a preset probability (set to 0.5 in this embodiment): if the random number is greater than 0.5, then all elements in that column are swapped. Since only the entire column is swapped, the constraint that there is exactly one 1 in the column is naturally preserved and no additional repair is required.

[0130] Mutation Operator – Bit Flip and In-Point Repair: Randomly select an element from the matrix with a preset probability (0.02 in this embodiment). If the element's current value is 0, flip it to 1, and simultaneously set all other elements in the same column to 0 (ensuring each column still has exactly one 1). If the element's current value is 1, flip it to 0, and randomly select an element currently 0 from that column and flip it to 1. This operation adjusts the designated docking point of a power device, thereby changing the task assignment of the drone.

[0131] 3.5: In a further implementation, the specific solution process and output are given. The initialization phase generates an initial population of 100 individuals. The upper-level encoding of each individual is randomly generated into a sequence of length [length missing]. Given a sequence of unique integers, the lower-level encoding randomly generates a sequence where each column contains exactly one 1. Binary matrix.

[0132] (2) Generation iteration; the algorithm runs for a total of 200 generations. In each generation, the following operations are performed: Step 1: Decode each individual in the population. During decoding, firstly, the spatial partitioning algorithm in step S3.3 is executed to divide flight corridors for each UAV. Then, under the spatial constraints of the flight corridors, the path construction algorithm in step S3.1 is used to generate the UAV flight edge sequence, resulting in... and .

[0133] Step 2: Perform the resource feasibility check in step S3.2 and calculate the total violation degree for each individual. .Will Individuals are marked as infeasible solutions and removed from the population. The number of removals is counted. The same random generation method as the initial population is used to generate the population. A new individual is added to the population, restoring the population size to 100. A resource feasibility check is performed on the added individual; if infeasible individuals still exist, the process of generating and replacing them is repeated until all 100 individuals are feasible solutions.

[0134] Step 3: For feasible individuals, calculate the inspection coverage integrity one by one according to the four formulas in step S2. Total system energy consumption Total task completion time Multi-drone load balancing Four objective function values.

[0135] Step 4: Perform non-dominated ranking and crowding distance calculation on feasible individuals based on the four objective function values.

[0136] Step 5: Generate a new generation of population through binary tournament selection, crossover, and mutation operations.

[0137] (3) After 200 generations of evolution, the algorithm terminates. Output the Pareto optimal solution set, which is the set of all feasible individuals that are not mutually dominant in the four objective dimensions.

[0138] Each solution in the Pareto optimal solution set contains the following complete information: Inspection vehicle stop access sequence and the arrival time window for each stop; Task allocation matrix ; Selected vehicle movement edge set It includes the start and end node indices, movement time consumption, and movement energy consumption for each vehicle movement edge; Selected drone flight edge collection It includes the start and end node indexes, flight time consumption, predicted flight energy consumption, and inspection equipment identification list for each flight edge; Four objective function values: , , , .

[0139] Decision-makers evaluate and select a path planning scheme from the Pareto optimal solution set based on the priorities of the current task (e.g., prioritizing inspection coverage or prioritizing shortening completion time) as the input scheme for the subsequent execution phase.

[0140] S4: In one implementation, during the actual execution of the path planning scheme selected in S3, a monitoring and feedback adjustment mechanism is established to address possible deviations between the actual environment and the assumptions made in the planning stage, and to ensure the safe completion of the inspection task under dynamically changing conditions.

[0141] During the execution of the path planning scheme, the system continuously collects flight status parameters in real time through the onboard sensors of the drone. The parameters that need to be monitored include: the drone's current remaining battery power, the drone's current flight speed, the ambient wind speed at the drone's location, and the real-time current value of the drone's motors.

[0142] The monitoring data is collected every 2 minutes. Each collected data is transmitted back to the ground monitoring center via a wireless communication link.

[0143] Unlike the predicted flight energy consumption calculated by the energy consumption model formula in step S3.2, step S4 focuses on the actual energy consumed by the UAV during actual flight. Actual energy consumption The calculation method is as follows: The motor power supply current is measured in real time using the UAV's onboard current sensor, in amperes, and then numerically integrated over the flight time.

[0144] in for The current sensor sampling value at a given time (in amperes), with the integration start and end time being the time from the takeoff of this flight segment. up to the current moment In engineering implementation, the current sensor performs discrete sampling every 2 minutes to obtain a time series. The above integral is approximated using the trapezoidal rule. The integral result is calculated online by the onboard flight control computer, and the unit is converted to ampere-hours.

[0145] Strict distinction is required: It is the actual energy consumption value obtained by measuring with sensors and calculating through integration. Step S3, the planning stage, is based on the energy consumption model formula. The predicted energy consumption value is calculated by substituting the planning parameters. These are two independent variables with different physical meanings and different acquisition methods.

[0146] S4.2: Deviation Detection and Trigger Judgment Each time a drone completes a flight mission (corresponding to a drone flight path planned in step S3), After returning to the inspection vehicle—according to the connectivity constraints in step S1.4, the endpoint of each UAV flight edge is the location of the inspection vehicle. Therefore, at the end of the flight segment, the UAV naturally has the conditions for energy consumption comparison—the system extracts the predicted energy consumption value corresponding to that flight segment. and actual energy consumption value Calculate the relative deviation :

[0147] in For the flight edge in the output scheme of step S3 Predicted flight energy consumption attribute values, This is the actual energy consumption value for this flight segment obtained through current integration in S4.1.

[0148] This embodiment presets a deviation trigger threshold. When the calculation yields At this point, a significant deviation between actual and planned energy consumption is detected. Possible reasons include: ambient wind speed significantly higher than forecast, the drone encountering turbulence requiring additional power to maintain attitude, and the actual power consumption of the mission payload exceeding the nominal value. If the subsequent flight plan is not adjusted in this situation, the drone may be unable to safely return to the inspection vehicle due to premature battery depletion.

[0149] when When the conditions are met, the system will automatically trigger the online replanning process.

[0150] S4.3 Online Replanning: During task execution, the remaining task arrangements are re-solved based on the current real state of the system.

[0151] (1) Constructing a local spatiotemporal network The first step in online reprogramming is to extract a local subnetwork from the original global spatiotemporal network to reduce the problem size and solution time. The extraction rules are as follows: Time range: Only retain data from the current moment. All spatiotemporal nodes and edges from the beginning to the next 2 hours.

[0152] Spatial range: Only the spatial grid area that may be reached by inspection vehicles and drones within the time range is retained.

[0153] Task scope: Remove the devices that have already been inspected from the original set of devices to be inspected, and keep the remaining devices that have not yet been inspected. The set is marked as follows: .

[0154] State initialization: Set the state parameters such as the drone's current remaining battery power, current location, the inspection vehicle's current location and remaining battery power as the initial conditions of the local spatiotemporal network.

[0155] After this truncation, the size of the nodes and edges in the local network is reduced to 30% to 50% of the original global network, ensuring that the replanning can be completed within a limited time.

[0156] (2) Re-plan and optimize objectives; The optimization objective of online replanning differs from that of global planning. During the task execution phase, the most important consideration has shifted from "multi-objective balance" to "completing as many inspections of remaining critical equipment as possible while ensuring safety." Therefore, the optimization objective of replanning is set as a single objective—maximizing the weighted completion rate of remaining tasks.

[0157] in For the remaining equipment to be inspected; Index the remaining devices; For equipment Importance weights (following the definition from step S2); For inspection status variables, Indicates equipment It was covered in the revised plan. This indicates that the content is not covered.

[0158] At the same time, safety constraints must be met:

[0159] in This represents the drone's current actual remaining battery power (obtained in real-time by onboard sensors). To adjust the predicted total energy consumption for all subsequent flight missions in the plan, i.e. ,in This is the set of all newly planned flight edges for the drone in the online replanning scheme, and the predicted energy consumption of each flight edge. Calculate according to the energy consumption model in step S3.2; As a safety margin, the value is set at 15% of the full capacity, or 1.5 amp-hours. This constraint ensures that the drone still has a safety margin to deal with emergencies after completing all subsequent flights.

[0160] (3) Reprogramming solution method Reprogramming has strict requirements on solution speed; therefore, the non-dominated sorting genetic algorithm used in global optimization is not adopted. Instead, a greedy random adaptive search strategy is used. The steps of this strategy are as follows: Step 1: Greedily construct the initial solution. Assign the remaining devices according to their importance weights. Sort the devices in descending order, starting with the device with the highest weight, and try assigning them one by one to the drones departing from the nearest available docking point. For each device assigned, check if the safety constraints are still satisfied; if so, accept the assignment; otherwise, skip the device. After traversing all devices, a feasible initial solution is obtained.

[0161] Step 2: Random Adaptive Search Improvement. Multiple iterative searches are performed within the neighborhood of the initial solution. In each iteration, one of the following three neighborhood operations is randomly selected: task reassignment (transferring a device from its original docking point to another), time window adjustment (adjusting the arrival time window of a docking point to a subsequent time slice), or flight path reorganization (adjusting the order in which UAVs access inspection devices departing from a docking point). If the new solution covers more important devices (with higher weighted completion rates) while satisfying safety constraints, the new solution is accepted; otherwise, it is accepted with a certain probability.

[0162] Step 3: Complete the search within the preset maximum number of iterations or time limit (in this embodiment, the maximum replanning time is set to 30 seconds) and output the optimal adjustment solution. If no feasible solution that meets the safety constraints is found within the maximum number of iterations or the 30-second time limit, the system triggers the safe return mechanism, ordering all UAVs to immediately interrupt the inspection mission and return to the nearest available docking point along the shortest path or directly return to the location of the inspection vehicle.

[0163] (4) Plan update and implementation The adjusted plans output from the online replanning are directly distributed to inspection vehicles and drones, replacing the parts of the original plan that have not yet been executed. Inspection vehicles and drones continue to perform subsequent inspection tasks according to the new plan.

[0164] Example 2: Full Process Demonstration of a Typical Scenario To further illustrate the complete implementation process of this invention, a full-process demonstration is provided below using a specific power inspection scenario. In this embodiment, the power consumption values ​​are converted according to the formula 1 ampere-hour equals 1000 milliampere-hours.

[0165] Scene setting An inspection task was carried out on a 50-kilometer-long 110 kV transmission line. Along the line, 86 electrical equipment items were to be inspected, including 17 first-level important equipment, 34 second-level important equipment, and 35 third-level important equipment. Available resources included one inspection vehicle and... drones ( The drone has a maximum flight time of 45 minutes and a full battery capacity of [missing information]. The payload capacity is 10 amp-hours (i.e., 10,000 milliamp-hours). The UAV is equipped with a visible light camera and an infrared thermal imager, with a total payload weight of 1.5 kilograms.

[0166] Phase 1: Spatiotemporal Network Construction The system imports geographic information system data for transmission lines, equipment ledger information, and daily weather forecast data. It then calculates the distribution of equipment across the entire area and the maximum equipment density. It is 1.0 units per square kilometer.

[0167] Calculate the initial grid resolution: Select a local area in the inspection zone with a device density of 0.8 per square kilometer, and substitute it into the formula to obtain... Meters. After checking the terrain complexity factor, the terrain of this area is flat. No threshold, no encryption required, final grid resolution Take 12 meters.

[0168] The inspection area is divided into 150 spatial grids using a 12-meter grid. The planned time span for this inspection mission is 12 hours (i.e., 720 minutes from mission start to the latest planned end time), with time slices set at 5 minutes each, resulting in 144 time slices.

[0169] Construct a spatiotemporal network: Automatically generate approximately 750 vehicle nodes and approximately 2250 drone nodes, as well as corresponding vehicle movement edges, drone flight edges, and collaborative edges.

[0170] Phase Two: Multi-objective Optimization Solution Run the improved non-dominated sorting genetic algorithm. Population size: 100, number of generations: 200.

[0171] Let's take a candidate solution from one of the iterations as an example to illustrate the calculation of each index: The total task completion time of the candidate solution The timeframe is 4.2 hours (from startup to the return of the last drone).

[0172] Total system energy consumption calculation: The inspection vehicle traveled a total distance of 82 kilometers. The energy consumption of the vehicle was calculated and weighted by a weighting factor. The predicted cumulative energy consumption of all flight sides of the drone is 8.22 amp-hours. The total system energy consumption is obtained after weighting. An Shi.

[0173] Inspection coverage completeness: The list of inspection equipment identifiers for all flight edges in the solution covers 84 out of 86 devices. The sum of the weights of these 84 covered devices accounts for 98% of the total weight, i.e. .

[0174] Load balancing: The total flight distances of the three drones were 48 km, 51 km, and 45 km, respectively. (Average) kilometers, variance Standard deviation km, .

[0175] The degree of violation of the scheme The feasibility check was passed.

[0176] After 200 generations of evolution, a Pareto optimal solution set is output, containing 15 non-dominant path solutions. The decision-maker selects the solution that prioritizes complete coverage as the execution plan based on task requirements.

[0177] Phase 3: Execution and Triggering Replanning The system dispatches vehicles and drones according to the selected plan to begin operations. The inspection vehicles travel along the highway to the designated stopping points, and the three drones perform flight inspections segment by segment as planned.

[0178] In the second hour of the mission, the system detected that the local wind speed had increased from the predicted 3 meters per second to 6 meters per second. During the recently completed flight segment, the onboard current sensor of a drone sampled and integrated the data using the trapezoidal rule to obtain the actual energy consumption. An Shi.

[0179] The planning parameters for this flight segment are: flight distance. kilometers, flight time Forecast wind speed every minute (i.e., 0.133 hours) meters per second, load Kilograms. Predicted energy consumption is calculated according to the energy consumption model in step S3.2: Anshi. Considering the actual wind speed increases to 6 meters per second, if the calculation is recalculated based on the actual wind speed, it would be... An-hour. The actual energy consumption obtained by integrating the current sensor during this flight segment. Anshi, and the predicted energy consumption during the planning phase Compare ampere-hours and calculate the deviation: This figure far exceeds the 20% threshold. The system immediately triggers online replanning.

[0180] Phase Four: Replanning and Task Completion The replanning module constructs a local spatiotemporal network based on the current state (62 remaining uninspected devices, current drone battery level, and current location), covering a range for the next 2 hours.

[0181] With the goal of maximizing the remaining weighted completion amount, the greedy random adaptive search strategy generates an adjustment plan within 18 seconds: reallocate the remaining inspection tasks, add a temporary docking point to shorten the round-trip distance of the drones, and arrange equipment groups with lighter energy consumption for drones affected by wind speed.

[0182] After the adjusted plan was issued, the system continued to execute according to the new plan. The entire inspection task was completed within 4.5 hours, covering all 86 devices, with the actual total energy consumption controlled within 9.0 amp-hours.

[0183] This typical scenario demonstrates the complete technical chain of the method of the present invention, from modeling and optimization to dynamic adjustment, and verifies the feasibility and effectiveness of the solution under real working conditions and sudden changes.

[0184] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and not to limit it. Although the present invention has been described in detail with reference to the above embodiments, those skilled in the art should understand that modifications or equivalent substitutions can still be made to the specific implementation of the present invention. Any modifications or equivalent substitutions that do not depart from the spirit and scope of the present invention should be covered within the protection scope of the claims of the present invention.

Claims

1. A vehicle-to-machine cooperative power inspection path optimization method considering range constraints, characterized in that, include: Construct a vehicle-machine collaborative spatiotemporal network model and formulate spatiotemporal constraints; The inspection task is discretized in both spatial and temporal dimensions, and a spatiotemporal coordinate system is constructed using the adaptive grid and time slice obtained after discretization. Define vehicle nodes, drone nodes, and edge attributes to construct the vehicle-drone inspection process as a graph structure; Based on the vehicle-machine collaborative spatiotemporal network model, the target dimensions of inspection path planning are clarified, and multi-dimensional optimization objectives are formulated, including maximizing the integrity of inspection coverage, minimizing the total system energy consumption, minimizing the total task completion time, and maximizing the load balancing of multiple drones. Based on the spatiotemporal network model and multidimensional optimization objective, a two-layer coding structure is constructed; for each UAV, spatial partitioning and collision avoidance distance constraint checks are performed, flight energy consumption is calculated and predicted, and resource feasibility checks are performed; the two-layer coding structure is iteratively solved and a Pareto optimal solution set is output. The final path planning scheme is selected from the Pareto optimal solution set and executed; during the execution process, the flight status parameters of the UAV are monitored and the actual energy consumption of the flight mission is calculated; the actual energy consumption of the flight mission is compared with the predicted flight energy consumption to calculate the relative deviation; when the relative deviation is greater than the corresponding threshold, the replanning is triggered to update the vehicle-machine cooperative spatiotemporal network model and optimization objective, and the local path planning scheme is solved again.

2. The vehicle-machine cooperative power inspection path optimization method considering range constraints according to claim 1, characterized in that, The inspection task is discretized in both spatial and temporal dimensions, including: In the spatial dimension, the initial grid size is calculated based on the distribution density of power distribution equipment in the inspection area; the terrain complexity factor is calculated using a digital elevation model, and the initial grid size is corrected to obtain the grid resolution. The grid resolution is set to meet the field-of-view coverage constraints of the UAV sensor, which is described as follows: the diagonal length of the grid resolution does not exceed the diameter of the UAV's field of view.

3. The vehicle-machine cooperative power inspection path optimization method considering range constraints according to any one of claims 1 or 2, characterized in that, Discretizing the inspection task in both spatial and temporal dimensions also includes: In terms of time, the entire inspection task time series is divided into equidistant time slices; the time slice length is reduced when any of the following conditions are met, taking into account real-time monitoring data and task priority settings. Condition 1: When the current time window is traced back... Inside, the difference between the maximum and minimum wind speeds exceeds the set value; Condition 2: The current inspection task priority attribute value is 1.

4. The vehicle-machine cooperative power inspection path optimization method considering range constraints according to claim 1, characterized in that, The construction process of the spatiotemporal network model of the vehicle system includes: Vehicle node Represented as the vehicle in time slice Time located in grid coordinates A new vehicle node is generated at each time slice as the vehicle moves. drone nodes Indicates the drone in time slice Time located in grid coordinates Energy is The load is As the drone flies, a new drone node is generated at the new location after each time slice.

5. The vehicle-machine cooperative power inspection path optimization method considering range constraints according to claim 1, characterized in that, The construction process of the vehicle system spatiotemporal network model also includes: Edge types include vehicle-moving edges, drone-flying edges, and collaborative edges; A vehicle movement edge connects two vehicle nodes of the same inspection vehicle in adjacent time slices; the attributes of a vehicle movement edge include the spatial distance between the two vehicle nodes at the two ends of the edge, the movement time consumption, and the movement energy consumption. The drone flight edge connects two drone nodes of the same drone in adjacent time slices; the attributes of the drone flight edge include the spatial distance between the drone nodes at both ends of the edge, flight time consumption, predicted flight energy consumption, and a list of inspection equipment identifiers; A collaborative edge connects a vehicle node and a drone node located in the same time slice and on the same spatial grid. The attributes of the collaborative edge include service time and energy replenishment. Service time describes the total time taken for the drone from landing on the inspection vehicle to completing takeoff preparation. Energy replenishment describes the additional available energy obtained by the drone at the docking point.

6. The vehicle-machine cooperative power inspection path optimization method considering range constraints according to claim 1, characterized in that, The inspection coverage integrity includes: The completeness of inspection coverage is defined as the proportion of the cumulative importance score of the inspected equipment to the total importance score of all equipment in the task area. Mathematically, it can be described as a fraction, with the denominator being the sum of the equipment importance of all the power distribution equipment to be inspected; and the numerator being the sum of the product of the equipment importance of all the equipment to be inspected and the corresponding inspection state variables. The inspection status variable is used to indicate whether the power distribution equipment is covered by at least one inspection by a drone in the path planning scheme. 1 indicates that it is covered and 0 indicates that it is not covered.

7. The vehicle-machine cooperative power inspection path optimization method considering range constraints according to claim 1, characterized in that, The total energy consumption of the system includes: The total energy consumption of the system is defined as the weighted sum of the energy consumed by the inspection vehicle and the drone in order to complete the inspection task; In mathematical terms, the path planning scheme involves calculating the sum of the movement energy consumption of all vehicle movement edges and the sum of the predicted flight energy consumption of all drone flight edges, and then weighted summing of the two sums.

8. The vehicle-machine cooperative power inspection path optimization method considering range constraints according to claim 1, characterized in that, The total task completion time includes: The total task completion time is defined as the total time elapsed from the start of the inspection task to the moment when the last drone completes all tasks and returns to the inspection vehicle. In mathematical terms, the path planning scheme calculates the time difference between the return time of each UAV and the start time of the inspection task; the maximum time difference is selected as the total task completion time.

9. The vehicle-machine cooperative power inspection path optimization method considering range constraints according to claim 1, characterized in that, The multi-drone load balancing includes: The multi-drone load balancing degree is defined as the degree of balance in the cumulative flight missions of multiple drones. Mathematically, it is described as 1 minus the coefficient of variation, which is the ratio of the average to the standard deviation of the total flight distance of each UAV.

10. The vehicle-machine cooperative power inspection path optimization method considering range constraints according to claim 1, characterized in that, The two-layer coding structure includes: The upper-level encoding is a sequence of vehicle stop points, and the lower-level encoding is a flight mission allocation matrix; the elements in the flight mission allocation matrix... Indicates from the first Are drones departing from each docking point responsible for inspecting the [number]th [location]? One electrical device, This indicates a lack of responsibility. This indicates that they are taking responsibility.

11. The vehicle-machine cooperative power inspection path optimization method considering range constraints according to claim 1, characterized in that, The spatial partitioning and collision avoidance distance constraint checks include: The spatial partitioning is as follows: corresponding to the number of drones, the spatial grid of the inspection area is divided into multiple flight corridors, and each drone performs its mission in one flight corridor. Based on spatial partitioning, the Euclidean distance between any two UAVs in the same time slice is calculated. When the Euclidean distance is greater than or equal to the safe distance, the collision avoidance distance constraint check is passed. The Euclidean distance is calculated based on the spatial coordinates of the drone nodes in the solved drone individual.

12. The vehicle-machine cooperative power inspection path optimization method considering range constraints according to claim 1, characterized in that, The resource feasibility check process includes: The predicted flight energy consumption is obtained by weighting the spatial distance of the flight edge, the ambient wind speed, and the total weight of the mission payload carried by the UAV, and then multiplying it by the flight time consumption of the flight edge. Calculate the violation degree for each flight edge in an individual, and sum the violation degrees for all flight edges to obtain the total violation degree of the individual; When the total violation degree of an individual is equal to 0, the individual is determined to be a feasible solution and retained, and participates in the iterative solution of the path planning scheme.

13. The vehicle-machine cooperative power inspection path optimization method considering range constraints according to claim 12, characterized in that, The calculation process for the degree of violation includes: The violation degree is the sum of the energy consumption violation degree and the duration violation degree; Energy consumption violation degree: The difference between the predicted flight energy consumption of the flight edge and the current remaining power is divided by the full power capacity. The quotient is compared with 0, and the larger value is taken as the energy consumption violation degree. Duration violation score: The difference between the flight time consumption of the flight edge and the maximum flight time of the drone is divided by the maximum flight time of the drone. The quotient is compared with 0, and the larger value is taken as the duration violation score.

14. The vehicle-machine cooperative power inspection path optimization method considering range constraints according to claim 12, characterized in that, The replanning includes: Based on the current moment, a local spatiotemporal network is extracted from the spatiotemporal network model, and the current state parameters of the UAV are used as the initial conditions of the local spatiotemporal network. The update and optimization objective is to maximize the weighted completion rate of the remaining tasks, which is expressed as the sum of the product of the inspection status variables of the remaining equipment to be inspected and their importance weights.

15. A vehicle-machine cooperative power inspection path optimization system considering range constraints, executing the inspection path optimization method as described in any one of claims 1-14, characterized in that, The system includes: The collaborative spatiotemporal network model construction module is used to construct a vehicle-machine collaborative spatiotemporal network model and formulate spatiotemporal constraints. The inspection task is discretized in the spatial and temporal dimensions, and the spatiotemporal coordinate system is constructed with the adaptive mesh and time slice obtained after discretization. Vehicle nodes, UAV nodes and edge attributes are defined to construct the vehicle-machine inspection process as a graph structure. Multi-objective definition module: Based on the vehicle-machine collaborative spatiotemporal network model, the objective dimensions of inspection path planning are clarified, and multi-dimensional optimization objectives are formulated, including maximizing the integrity of inspection coverage, minimizing the total system energy consumption, minimizing the total task completion time, and maximizing the load balancing of multiple drones; The Pareto solution module is used to construct a two-layer coding structure based on the spatiotemporal network model and multidimensional optimization objective; for each UAV, it performs spatial partitioning and collision avoidance distance constraint checks, calculates and predicts flight energy consumption, and performs resource feasibility checks; it iteratively solves the two-layer coding structure and outputs the Pareto optimal solution set. The local replanning module is used to select and execute the final path planning scheme from the Pareto optimal solution set. During execution, it monitors the UAV flight status parameters and calculates the actual energy consumption of the flight mission. It compares the actual energy consumption of the flight mission with the predicted flight energy consumption to calculate the relative deviation. When the relative deviation is greater than the corresponding threshold, it triggers replanning to update the vehicle-machine cooperative spatiotemporal network model and optimization objective, and re-solves the local path planning scheme.

16. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the computer program is loaded into the processor, it implements the inspection path optimization method according to any one of claims 1-15.

17. A computer-readable storage medium storing a computer program, characterized in that, When the computer program is executed by the processor, it implements the inspection path optimization method according to any one of claims 1-15.

Images