Power grid transmission channel micro-flick device and unmanned aerial vehicle combined deployment method and system

Abstract

The application discloses a power grid transmission channel micro-photograph device and unmanned aerial vehicle combined deployment method and system. First, four sets of transmission channel targets, micro-photograph candidate points, unmanned aerial vehicle take-off and landing points, and candidate tasks are built, and risk scores of each target are obtained by combining fault probability prediction. For fixed micro-photograph devices and unmanned aerial vehicle inspection tasks, component defect identification effective coverage determination rules are established, multi-dimensional indexes such as observation distance, line-of-sight condition, observation angle and image interpretability are fused, and invalid coverage misjudgment of visible but unable to identify defects is eliminated. A collaborative decision-making model is constructed, the existing coverage, budget, optional deployment actions and engineering restrictions are comprehensively considered in each round of decision-making, and micro-photograph layout points or unmanned aerial vehicle inspection tasks are step by step optimized; a closed-loop optimization target is constructed by taking the newly added risk coverage benefit, standardized deployment cost and constraint violation penalty as the input. Finally, the micro-photograph layout scheme, unmanned aerial vehicle inspection task planning and coverage, and cost statistical data are output.

Description

Technical Field

[0001] This invention relates to the field of power transmission channel monitoring and inspection technology, specifically to a method and system for the joint deployment of micro-photography equipment and drones in power grid transmission channels. Background Technology

[0002] Power transmission corridors are crucial infrastructure for power transmission systems, typically traversing mountainous areas, forested regions, waterways, major transportation routes, and urban-rural fringe areas in the form of ribbon-like corridors. These corridors face complex environments, significant topographical variations, and numerous external disturbances. During operation, transmission corridors not only face environmental risks such as vegetation encroachment, wildfires, external damage, geological disasters, and severe weather, but also equipment risks such as cracks, damage, loosening, corrosion, and flashover hazards in critical components like insulators, fittings, clamps, and crossarm connections. Timely identification and mitigation of these risks are directly related to the safe and stable operation of transmission lines.

[0003] Current power transmission channel monitoring and inspection work typically combines various methods, including manual inspection, fixed video monitoring equipment, micro-video equipment, and drone inspection. Fixed video or micro-video equipment is suitable for long-term, continuous, and routine observation of key areas, capable of continuously collecting image or video information and providing high-frequency monitoring capabilities for critical local areas. Drone inspection, on the other hand, has advantages such as high mobility, flexible deployment, variable viewing angle, and wide coverage, enabling supplementary inspections of high-risk areas that are difficult for fixed monitoring equipment to cover long-term or that are temporarily added.

[0004] However, in engineering practice, the deployment of fixed video recording equipment and drone inspection tasks are usually planned using different processes. Fixed video recording equipment deployment often focuses on installation feasibility, experience in key areas, and visual coverage; drone inspection, on the other hand, focuses more on flight path planning, operation scheduling, endurance constraints, and mission feasibility under no-fly or altitude-restricted conditions. The lack of a unified evaluation standard, a unified benefit measurement standard, and a unified resource allocation method between the two types of resources easily leads to the following problems: On the one hand, fixed equipment may already cover some target objects, while drone tasks repeatedly inspect the same areas, resulting in resource redundancy; on the other hand, some high-risk target objects may have insufficient coverage or coverage gaps due to the lack of coordinated planning between fixed deployment and drone tasks.

[0005] Furthermore, for component-type targets such as insulators, fittings, and clamps, simply using "visibility" as a coverage criterion is often insufficient. Component defect identification typically has high requirements for observation distance, observation angle, line-of-sight conditions, image resolution, and image quality. Although some candidate monitoring points may be geometrically observable of the target object, they may not actually support defect identification and fault diagnosis due to excessive distance, severe obstruction, poor viewing angle, insufficient imaging pixels, or low image interpretability. Therefore, in the allocation of monitoring resources for transmission channels, relying solely on nominal visual coverage or empirical rules to configure monitoring resources is insufficient to meet the actual engineering needs for component-type defect identification.

[0006] Meanwhile, the risk levels of target objects along the transmission line are not uniform. Different tower locations, equipment types, and environmental sections typically differ in their failure probabilities, consequences, and maintenance priorities. If the risk differences among target objects are not fully reflected during resource allocation, the limited budget may be evenly distributed between high-risk and low-risk targets, thereby reducing the overall deployment efficiency.

[0007] Furthermore, the deployment of monitoring resources for power transmission channels is also affected by various engineering constraints, including budget, installation feasibility, safe distance from energized areas, power supply and communication conditions, drone endurance, return safety margin, no-fly zones, height restrictions, and the availability of take-off and landing points. If the deployment model selects monitoring resources solely based on coverage or theoretical benefits without simultaneously filtering out unfeasible actions during the decision-making process, it may output micro-photography point deployment schemes or drone inspection task schemes that are difficult to implement, thus impacting practical engineering applications.

[0008] Therefore, how to integrate the deployment of fixed micro-photography equipment and the configuration of UAV inspection tasks into the same technical framework under unified budget and engineering constraints, while simultaneously considering the differences in risk of target objects, the effective coverage conditions required for component defect identification, and the engineering feasibility of deployment actions, so as to achieve the coordinated deployment of power transmission channels with priority given to high-risk targets, priority given to effective coverage, and controllable overall costs, has become an urgent technical problem to be solved. Summary of the Invention

[0009] The purpose of this invention is to provide a method and system for the joint deployment of micro-photography equipment and drones in power grid transmission channels, so as to solve the problems of the existing configuration of power transmission channel monitoring resources, such as the separation between the deployment of fixed micro-photography equipment and the configuration of drone inspection tasks, the lack of a unified and effective coverage evaluation mechanism, the difficulty in prioritizing the coverage of high-risk targets, and the insufficient engineering feasibility of the deployment plan.

[0010] To achieve the above objectives, a method for jointly deploying micro-photography equipment and drones in power grid transmission channels is provided, the method comprising the following steps: S100: Acquire basic data of the power transmission channel, and construct a set of target objects, a set of candidate micro-photographing points, a set of UAV take-off and landing points, and a set of candidate UAV tasks based on the tower-level monitoring objects; S200. Based on the target object set, historical defect records, historical alarm statistics, environmental characteristics, and meteorological characteristics, determine the risk score for each target object; S300: Based on the risk score of each target object, calculate the effective coverage relationship of the candidate micro-photograph points to the fixed side of the target object, and the effective coverage relationship of the candidate UAV missions to the mission side of the target object. S400, based on the effective coverage relationship between the fixed side and the mission side, combined with budget constraints and engineering constraints, constructs a set of actionable actions for candidate micro-photography point deployment actions and UAV mission selection actions; S500: Construct a collaborative decision-making environment, combine the set of actionable actions, and use the action masking mechanism to shield non-executable actions in the unified action space; S600 invokes the trained collaborative decision-making model, selects the current deployment action based on the current state, target object risk score, remaining budget, and actionable action set after action masking, and updates the selected micro-photography point set, selected drone mission set, target object coverage status, and remaining budget based on the current deployment action; S700 When the termination condition is met, output the updated set of selected micro-photography points, the set of selected drone missions, the coverage status of the target object, and the remaining budget.

[0011] Furthermore, the method for constructing the target object set, candidate micro-photography location set, UAV take-off and landing point set, and UAV candidate task set in step S100 of the present invention includes: The method for constructing the target object collection is as follows: The monitoring objects in the power transmission channel are abstracted into target objects at the tower level, forming a set of target objects. Each target object corresponds to a key part of the tower in the power transmission channel that is related to defect monitoring; Among them, the key parts of the tower include at least one of the following: tower head area, crossarm area, insulator string area, hardware connection area, and clamp connection area; Each target object must include at least the target object number, the tower number to which it belongs, and its spatial location. Component type label Recommended observation direction vector Historical defect records, historical alarm statistics, equipment attributes, environmental characteristics, and meteorological characteristics; Among them, spatial location Represented using a unified projected coordinate system; The method for constructing the candidate micro-photography location set is as follows: Based on the tower structure, installation conditions, and monitoring requirements, a set of candidate micro-monitoring points is generated from the locations of installable components on the tower, crossarm locations, maintenance platform locations, tower body hanging points, and other locations that meet the installation conditions. ; Each candidate micro-photography location should include at least the location number and the location's spatial location. Installation height range, adjustable orientation range, lens parameters, power supply method, communication and data transmission method, and deployment cost. Installation feasibility markers ; Among them, candidate micro-photography locations Deployment costs It consists of the following parts:

[0012] in, For equipment costs, To maintain costs, For communication backhaul costs, For power supply costs; when At that time, it indicates the candidate micro-photograph point. The installation conditions are met and the system can proceed to the next set of feasible actions; when At that time, it indicates the candidate micro-photograph point. It does not meet the requirements for structural installation, electrical safety distance, power supply and communication conditions, or construction accessibility, and is therefore eliminated in the subsequent decision-making process; The method for constructing the set of drone take-off and landing points is as follows: A set of drone take-off and landing points is generated based on maintenance sites, power supply stations, inspection stations, or temporary take-off and landing areas. Each UAV take-off and landing point should include at least the take-off and landing point number, spatial location, site availability, service radius, take-off and landing time window, and corresponding airspace restriction information. Construct a set of candidate drone tasks: Based on the set of UAV take-off and landing points, the set of target objects, the spatial distribution of power transmission channels, and UAV operational constraints, a set of candidate UAV tasks is generated. ; Each drone candidate task At least include the mission number and the landing / takeoff point. target object subset Track or waypoint sequence, flight altitude parameters, flight speed parameters, hovering strategy parameters, estimated flight time Task cost ; Task cost of drone mission u Represented as:

[0013] in, To reduce fixed costs, This is the unit time cost coefficient. For the estimated flight time; When the mission flight time is determined by the track length, flight speed, and hovering duration, it is expressed as:

[0014] in, The length of the flight path. For flight speed, For the first The hovering time at each hovering point; The set of candidate drone tasks can be constructed by pre-generating a set of candidate tasks, or by generating new candidate drone tasks online based on the currently uncovered target objects, real-time airspace conditions, and remaining budget when the candidate tasks cannot meet the current coverage requirements.

[0015] Furthermore, the method for determining the risk score of each target object in step S200 of the present invention includes: For the target object collection Each target object in Construct the feature vector of the target object ; Wherein, the target object feature vector It should include at least one or more of the following: the number of historical defects of the target object, the severity of the defects, the frequency of historical alarms, the service life of the equipment, the component type, the pollution level, the channel environmental risk factors, the meteorological risk factors, and the operation and maintenance priority information; The feature vector of the target object Input the fault probability prediction model to obtain the target object Failure probability :

[0016] in, For failure probability prediction models, ; Based on the failure probability Determine the risk score wt of the target object t:

[0017] in, It is a monotonically increasing mapping function; target object The overall risk score is expressed as follows:

[0018] in, and For preset weighting coefficients, This is the normalization function; Based on risk score The target objects are sorted, and those with risk scores higher than a preset risk threshold are identified as high-risk target objects, or a preset number of target objects with the highest risk scores are identified as high-risk target objects. The high-risk target objects are used for subsequent effective coverage statistics, reward function calculation, and deployment action selection, so that fixed micro-photography equipment and drone missions prioritize coverage of target objects with higher risk scores.

[0019] Furthermore, the calculation method for the effective coverage relationship on the fixed side and the effective coverage relationship on the mission side in step S300 of the present invention includes: For the candidate micro-photography point set S For each candidate micro-shot point s in the target object set T and for each target object t in the target object set T, calculate the candidate micro-shot point. For the target object Fixed-side effective coverage indication ; Fixed side effective coverage indication Determined based on distance conditions, line-of-sight conditions, viewing angle conditions, and image interpretation conditions; The distance conditions include:

[0020]

[0021] in, Candidate micro-photography locations Spatial location, For the target object Spatial location, To be with the target object The maximum effective observation distance threshold corresponding to the component type; Line-of-sight conditions include: candidate micro-photograph locations With the target object There is no obstruction between them, or the degree of obstruction does not exceed a preset obstruction threshold; among them, the following is adopted. Indicates candidate micro-photograph points To the target object The line-of-sight result, when the line-of-sight requirements are met. ;otherwise, ; Viewpoint conditions include: based on the target object Recommended observation direction vector and candidate micro-photography locations Point to target object line of sight direction vector Calculate the angle of observation And determine the included angle of observation. Is it not greater than the preset observation angle threshold? ; Among them, the line-of-sight direction vector Represented as:

[0022] The observed angle Ang(s,t) is expressed as:

[0023] when At that time, it was considered that the perspective condition was met; Image interpretation criteria include: target object The key area is the candidate micro-photography location. The number of imaging pixels, image clarity, or readability in the acquired image shall not be lower than the corresponding preset threshold. Based on candidate micro-photo locations Lens parameters, candidate macro shooting points With the target object The observation distance between them and the target object The actual dimensions of the key areas are used to calculate the target object. The number of imaging pixels in the key area; when the number of imaging pixels in the key area is not lower than a preset pixel threshold, the image interpretation condition is considered to be met. When candidate micro-photo locations With the target object When the distance condition, line-of-sight condition, viewing angle condition, and image interpretation condition are all satisfied simultaneously, the following definition applies:

[0024] Otherwise, define:

[0025] For each UAV candidate task in the UAV candidate task set U u and each target object in the target object set T Calculate candidate tasks for unmanned aerial vehicles u For the target object Task-side effective coverage indication ; When drone candidate tasks u There must be at least one shooting location in the flight path or waypoint sequence such that the shooting location is within view of the target object. When the conditions of mission observation distance, camera field of view, observation angle, occlusion ratio, and image resolution are simultaneously satisfied, the following definition applies:

[0026] Otherwise, define:

[0027] The task observation distance condition refers to the distance between the UAV's shooting position and the target object. The distance between them does not exceed the preset task observation distance threshold; Effective coverage indication via fixed side and task-side effective coverage indication In the same coverage evaluation system, the effective coverage contribution of candidate micro-photograph sites and candidate UAV missions to the target object is uniformly measured.

[0028] Furthermore, in step S400 of the present invention, the method for constructing a set of actionable actions for candidate micro-photography site deployment actions and UAV mission selection actions based on budget constraints and engineering constraints includes: Based on the current remaining budget Brem and the set of selected micro-photography locations S k Selected UAV mission set U k Given the candidate micro-photo location set S, the candidate U drone mission set U, and engineering constraints, we construct the feasible micro-photo location deployment action set and the feasible U drone mission selection action set, respectively. Among them, the engineering constraints on the fixed micro-video side include one or more of the following: installation feasibility constraints, live safety distance constraints, power supply accessibility constraints, communication backhaul availability constraints, and construction accessibility constraints. Installation feasibility constraints are determined by installation feasibility markers. It means that when At that time, it indicates the candidate micro-photograph point. Installation conditions are met; when At that time, it indicates the candidate micro-photograph point. Installation conditions not met; The set of feasible micro-photography point deployment actions is represented as follows: , ,and Satisfy the engineering constraints of the fixed micro-photograph side; in, This represents the current state at the k-th decision step. Indicates the deployment of candidate micro-photograph locations The action, Indicates candidate micro-photograph points The deployment cost, Brem represents the current remaining budget. Indicates the first The set of micro-shot points selected at each decision step; Unmanned aerial vehicle (UAV) side engineering constraints include one or more of the following: mission cost constraints, endurance constraints, return-to-home margin constraints, no-fly zone constraints, altitude restriction constraints, take-off and landing point availability constraints, and mission time window constraints; the endurance constraints and return-to-home margin constraints are expressed as follows:

[0029] in, This indicates the estimated flight time of the candidate drone mission u. This indicates the return time required for the drone candidate mission u. Indicates the maximum permissible flight time for the drone; No-fly zone constraints are represented as follows:

[0030] in, For the mission trajectory, Gathering in the no-fly zone; Altitude restrictions are a candidate task for unmanned aerial vehicles (UAVs). u The mission execution altitude is within the permissible flight altitude range and does not exceed the maximum flight altitude set by regulatory or operational requirements; The set of feasible drone mission action selections is represented as follows: ,and

[0031] in, Indicates the selection of drone candidate tasks The action, Indicates candidate drone missions Task cost, Indicates the first k The set of drone missions has been selected at each decision-making step; Based on the set of feasible micro-photography location deployment actions and the set of feasible UAV mission selection actions, the first... k The overall set of actionable actions for each decision step:

[0032] in, Indicates the set of actions to be terminated; Add the stop action. The remaining budget is insufficient to execute any non-terminating action; the feasible micro-photograph point deployment action set and the feasible drone mission selection action set are both empty; the overall coverage rate has reached the preset overall coverage rate threshold; the coverage rate of high-risk targets has reached the preset high-risk coverage threshold; the additional risk coverage benefit of consecutive preset number decision steps is lower than the preset benefit threshold; or the current number of decision steps has reached the preset maximum number of decision steps. Before action selection, non-executable actions in the unified action space are filtered based on the overall set of actionable actions, so that the subsequent collaborative decision-making model selects the currently deployed action only from those actions that meet budget constraints and engineering constraints. Before or during the execution of an action, the selected action is dynamically verified based on real-time airspace status, changes in equipment availability, changes in weather conditions, or task execution feedback. If the selected action does not meet the engineering constraints during the dynamic verification phase, a constraint violation penalty is triggered, and a new deployment action is selected based on the updated set of actionable actions.

[0033] Furthermore, the method for constructing a collaborative decision-making environment in step S500 of the present invention includes: The state space includes a high-risk target explicit representation block, a global coverage statistics block, a budget and schedule block, and a summary block of the set of actionable actions; Among them, the high-risk target explicit representation block is used to record the current coverage status and normalized risk score of the target objects ranked first in risk score; The global coverage statistics block should include at least the overall coverage rate, the coverage rate of high-risk targets, and the gap ratio; The budget and schedule block should at least include the remaining budget percentage, the selected micro-photography location percentage, and the selected drone mission percentage; The action set summary block should include at least the proportion of action to deploy feasible micro-photo locations and the proportion of action to select feasible drone missions; The unified action space A includes a set of candidate micro-photography point deployment actions, a set of UAV mission selection actions, and a termination action; among which, the candidate micro-photography point deployment actions are represented as follows: The drone mission selection action is represented as The action to be terminated is indicated by "stop"; Based on the overall set of actionable actions Apply an action mask to actions in the unified action space A. , among which, when hour, =1; when hour, =0; When there are no feasible non-terminating actions in the set of feasible actions, add the terminating action stop to the set of feasible actions, so that the action probability distribution after action masking contains at least one executable action.

[0034] The original action probability distribution output by the collaborative decision-making model By performing action masking, we obtain the normalized probability distribution of possible actions:

[0035] The reward function is determined based on the additional risk coverage benefits brought by the current deployment action, the normalized deployment cost, and the penalty for constraint violations, and is expressed as:

[0036] in, This represents the additional risk coverage return at the k-th decision step. This represents the normalized deployment cost of the k-th decision step. This indicates restrictions and penalties for violations. and Preset weighting coefficients; The additional risk coverage benefit is determined based on the target object's risk score and the change in the target object's coverage status; when a deployment action causes a target object with a high risk score to change from an uncovered state to an effectively covered state, the corresponding additional risk coverage benefit is obtained. The normalized deployment cost is determined based on the ratio between the deployment cost of the micro-photography point corresponding to the current deployment action or the cost of the drone mission and the total budget. When expanding the candidate drone tasks using an online generation method, the online-generated candidate drone tasks are added to a candidate task pool with a preset capacity. When the number of candidate tasks exceeds the preset capacity, they are sorted and truncated according to the risk coverage benefit per unit cost. When the number of candidate tasks is less than the preset capacity, they are filled with empty tasks, and empty and non-executable tasks are masked by action masks to maintain the action output of the collaborative decision-making model.

[0037] Furthermore, the method for updating the selected micro-photography point set, the selected UAV mission set, the target object coverage status, and the remaining budget according to the current deployment action in step S600 of the present invention includes: Collaborative decision-making models include policy networks and value networks; The policy network is used to output the original action probability distribution of each action in the unified action space based on the current state, and the value network is used to estimate the value of the current state. In each training round, a complete joint deployment process of fixed micro-photography equipment and UAV mission is considered as a training round. The environment is initialized based on the basic data of the power transmission channel, the risk score of the target object, the set of candidate micro-photography points, the set of candidate UAV missions, the budget level, and the engineering constraints. In the k-th decision step, the policy network determines the current state. Output the original action probability distribution, and obtain the action probability distribution based on the action masking mechanism. Select the current deployment action from the possible action probability distribution. ; when Update the set of selected micro-photo locations at that time. Update remaining budget and according to Indicator Update the coverage status of each target object; when Update the selected drone mission set at that time. Update remaining budget And based on the effective coverage indication on the mission side. Update the coverage status of each target object; when If necessary, terminate the current training round or the current deployment process; The training process of reinforcement learning models forms a training trajectory based on immediate rewards, state transition results, and termination judgments, and updates the parameters of collaborative decision-making models through policy optimization.

[0038] The present invention also provides a joint deployment system of micro-photography equipment and drones for power grid transmission channels. The system includes a data acquisition module, a target object and resource set construction module, a risk scoring module, an effective coverage calculation module, an engineering constraint processing module, a collaborative decision-making environment construction module, a deployment action selection and status update module, and a result output module.

[0039] The data acquisition module is used to acquire basic data of the power transmission channel. The target object and resource set construction module is used to construct a target object set, a candidate micro-photograph location set, a UAV take-off and landing point set, and a UAV candidate task set based on the basic data of the power transmission channel. The risk scoring module is used to determine the risk score of each target object based on the target object set, historical defect records, historical alarm statistics, environmental characteristics, and meteorological characteristics. The effective coverage calculation module is used to calculate the fixed-side effective coverage relationship of the candidate micro-photograph locations to the target objects, and the task-side effective coverage relationship of the UAV candidate tasks to the target objects. The engineering constraint processing module is used to construct a set of feasible actions for candidate micro-photograph location deployment actions and UAV task selection actions based on budget constraints and engineering constraints. The collaborative decision-making environment construction module is used to construct a state space, a unified action space, a reward function, and an action masking mechanism. The deployment action selection and state update module is used to call the trained collaborative decision-making model to select the current deployment action, and update the selected micro-photograph location set, the selected UAV task set, the target object coverage status, and the remaining budget based on the current deployment action. The result output module is used to output the fixed micro-photography point deployment plan, the drone inspection task plan, and the coverage and cost statistics when the termination conditions are met.

[0040] The present invention also provides a computer device, including a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, it implements the aforementioned method for jointly deploying micro-photography equipment and drones along power grid transmission channels.

[0041] The present invention also provides a computer-readable storage medium having a computer program stored thereon. When the computer program is executed by a processor, it implements the above-described method for jointly deploying micro-photography equipment and drones in power grid transmission channels.

[0042] Compared with the prior art, the present invention has at least the following beneficial effects: This invention incorporates target object risk scoring into the calculation of new coverage revenue, so that the selection of fixed micro-photography equipment and drone missions is no longer based solely on spatial distance or human experience rules, but can prioritize the coverage of target objects with high failure probability and high failure consequences, thereby improving the effective coverage rate of high-risk target objects under limited budget.

[0043] This invention incorporates observation distance, line-of-sight relationship, observation angle, and image interpretation conditions into the fixed-side effective coverage judgment, and incorporates observation distance, field of view, observation angle, occlusion ratio, and image resolution into the task-side effective coverage judgment. This can avoid determining coverage solely from the geometric visual angle, thereby reducing false judgments of nominal coverage that are "visible but not clear" and improving the effectiveness of coverage results in defect identification scenarios.

[0044] This invention constructs a unified action space that includes candidate micro-photography site deployment actions, UAV mission selection actions, and termination actions. By using a set of feasible actions and an action masking mechanism, actions that do not meet budget constraints and engineering constraints are masked. This reduces the risk of the collaborative decision-making model outputting unexecutable deployment schemes and improves the engineering executability of deployment schemes.

[0045] This invention considers the benefits of new risk coverage, normalized deployment costs, and penalties for violations in the reward function, which can form a unified evaluation between the long-term monitoring advantages of fixed micro-video equipment and the flexible supplementary advantages of UAV missions, so that the deployment results take into account coverage benefits, cost control, and engineering constraints.

[0046] This invention can simultaneously output fixed micro-photography point deployment schemes, UAV inspection task schemes, overall coverage, high-risk target coverage, weighted coverage, list of uncovered target objects, and budget utilization rate, providing a unified basis for subsequent power transmission channel monitoring resource allocation, inspection task adjustment, scheme implementation, and operation and maintenance effect evaluation. Attached Figure Description

[0047] Figure 1 This is an overall flowchart of the method for jointly deploying micro-photography equipment and drones in power grid transmission channels according to the present invention; Figure 2 A schematic diagram is constructed for the target object, candidate micro-photography locations, UAV take-off and landing points, and candidate task set; Figure 3 A schematic diagram illustrating the determination of effective coverage relationships between the fixed side and the mission side. Figure 4 A schematic diagram illustrating the unified action space, set of possible actions, and action masking mechanism; Figure 5 Flowchart for action selection, status update, and result output. Detailed Implementation

[0048] To make the objectives, technical solutions, and beneficial effects of this invention clearer, the technical solutions of this invention will be described below in conjunction with the accompanying drawings and specific embodiments. It should be understood that the following embodiments are for illustrative purposes only and are not intended to limit the scope of protection of this invention. Where there is no conflict, the embodiments and technical features in the embodiments of this invention can be combined with each other.

[0049] like Figure 1 As shown in the figure, the method for joint deployment of micro-photography equipment and drones in power grid transmission channels in this embodiment includes steps such as data and resource set construction, target object risk scoring, effective coverage relationship calculation, feasible action set construction, collaborative decision-making environment construction, deployment action selection and status update, and result output.

[0050] S100: Obtain basic data of the power transmission channel and construct a set of target objects, a set of candidate micro-photography locations, a set of UAV take-off and landing points, and a set of candidate UAV tasks.

[0051] like Figure 2 As shown, in step S100, the basic data of the power transmission channel is used for constructing the target object set, screening candidate micro-photographing points, generating UAV take-off and landing points, and generating UAV candidate tasks; among them, the UAV candidate task set is generated based on the UAV take-off and landing point set, the target object set, the spatial distribution of the power transmission channel, and the UAV operation constraints.

[0052] Specifically, the basic data of the power transmission channel includes tower ledger data, equipment asset data, power transmission channel GIS data, historical defect data, historical alarm data, environmental data, meteorological data, candidate micro-photography location data, UAV take-off and landing point data, and UAV inspection operation data.

[0053] The target object set consists of tower-level monitoring objects in the transmission line. Each target object corresponds to a key tower part in the transmission line related to defect monitoring; the key tower parts include at least one of the following: tower head area, crossarm area, insulator string area, hardware connection area, and clamp connection area. Each target object includes at least the target object number, its tower number, spatial location, component type label, recommended observation direction vector, historical defect records, historical alarm statistics, equipment attributes, environmental characteristics, and meteorological characteristics.

[0054] The candidate micro-monitoring point set is generated based on the tower structure, installation conditions and monitoring requirements. Each candidate micro-monitoring point includes at least the point number, spatial location, installation height range, adjustable orientation range, lens parameters, power supply method, communication feedback method, point deployment cost and installation feasibility mark.

[0055] Deployment cost of candidate micro-photography sites The costs can be comprised of equipment costs, maintenance costs, communication backhaul costs, and power supply costs, as detailed below:

[0056] in, For equipment costs, To maintain costs, For communication backhaul costs, For power supply costs; The set of UAV take-off and landing points is generated based on maintenance sites, power supply stations, inspection stations, or temporary take-off and landing areas. The set of UAV candidate tasks is generated based on the set of UAV take-off and landing points, the set of target objects, the spatial distribution of power transmission channels, and UAV operational constraints. Each UAV candidate task includes at least a task number, its associated take-off and landing point, a subset of target objects to be inspected, a flight path or waypoint sequence, flight altitude parameters, flight speed parameters, hovering strategy parameters, estimated flight time, and task cost.

[0057] In one implementation, the drone mission Task cost Represented as:

[0058] in, To reduce fixed costs, This is the unit time cost coefficient. For the estimated flight time; When the mission flight time is determined by the track length, flight speed, and hovering duration, it is expressed as:

[0059] in, The length of the flight path. For flight speed, For the first The hovering time at each hovering point; S200. Based on the target object set, historical defect records, historical alarm statistics, environmental characteristics, and meteorological characteristics, determine the risk score for each target object.

[0060] For each target object in the target object set T Construct the feature vector of the target object The target object feature vector It should include at least one or more of the following: number of historical defects, severity of defects, frequency of historical alarms, service life of equipment, component type, pollution level, channel environmental risk factors, meteorological risk factors, and operation and maintenance priority information.

[0061] The feature vector of the target object Input the fault probability prediction model to obtain the target object Failure probability :

[0062] in, For failure probability prediction models, Failure probability prediction models can employ logistic regression models, random forest models, gradient boosting tree models, neural network models, or risk prediction models based on expert rules.

[0063] Based on the failure probability Determine the target object Risk score for:

[0064] in, It is a monotonically increasing mapping function.

[0065] In one implementation, risk scoring It can also be combined with the target object Fault consequence weight Sure:

[0066] Where α and β are preset weighting coefficients, and Norm(·) is the normalization function. A higher risk score indicates that the target object should be given priority for effective coverage by fixed micro-capture equipment or drone missions.

[0067] S300: Based on the risk scores of each target object, calculate the effective coverage relationship of the candidate micro-photograph points to the target object on the fixed side, and the effective coverage relationship of the candidate UAV missions to the target object on the mission side.

[0068] like Figure 3 As shown, in step S300, the effective coverage relationship on the fixed side is determined based on distance conditions, line-of-sight conditions, viewing angle conditions, and image interpretation conditions; the effective coverage relationship on the mission side is determined based on mission observation distance conditions, camera field of view conditions, observation viewing angle conditions, occlusion ratio conditions, and image resolution conditions.

[0069] By using both fixed-side and mission-side effective coverage indicators, the effective coverage contribution of candidate micro-photograph sites and UAV candidate missions to the target object is uniformly measured within the same coverage evaluation system.

[0070] For the effective coverage relationship on the fixed side, for each candidate micro-shot point in the candidate micro-shot point set S... and each target object in the target object set T Determine candidate micro-photograph locations For the target object Whether it constitutes effective coverage on the fixed side. The effective coverage relationship on the fixed side is determined based on distance conditions, line-of-sight conditions, viewing angle conditions, and image interpretation conditions.

[0071] Candidate micro-photography locations With the target object Observation distance between Represented as:

[0072] When the following conditions are met:

[0073] When this condition is met, the distance requirement is considered satisfied. Candidate micro-photography locations Spatial location, For the target object Spatial location, To be with the target object The maximum effective observation distance threshold corresponding to the component type.

[0074] The line-of-sight condition is the candidate micro-photograph location. With the target object There is no obstruction between them, or the degree of obstruction does not exceed the preset obstruction threshold. This can be achieved using... This indicates the line-of-sight result; when the line-of-sight requirements are met, ;otherwise, .

[0075] The viewing angle condition is the candidate micro-photograph location. For the target object The observation angle is not greater than the preset observation angle threshold. Based on the target object... Recommended observation direction vector and candidate micro-photography locations Point to target object line of sight direction vector Calculate the angle of observation :

[0076]

[0077] When the following conditions are met:

[0078] At that time, it is considered that the perspective condition is met.

[0079] Image interpretation criteria are target objects The key area is the candidate micro-photography location. The number of imaging pixels, image clarity, or readability in the acquired image are not lower than the corresponding preset threshold. In one implementation, based on candidate micro-capture points... Lens parameters, candidate macro shooting points With the target object The observation distance between them and the target object The actual dimensions of the key areas are used to calculate the target object. Number of pixels in key areas .when Not lower than the preset pixel threshold At that time, the image interpretation conditions are deemed met.

[0080] When candidate micro-photo locations With the target object When distance, line-of-sight, viewing angle, and image interpretation conditions are all met, the effective coverage indicator for the fixed side is defined. Otherwise, define:

[0081] Represented as:

[0082] Where 1(·) represents the indicator function, where, Represents the target object The key area is the candidate micro-photography location. The number of imaging pixels in the acquired image. This indicates the preset pixel threshold.

[0083] For the effective coverage relationship on the task side, for each UAV candidate task in the UAV candidate task set U... u and each target object in the target object set T Determine candidate drone tasks u Does the target object t constitute effective coverage on the mission side? (When a drone candidate mission...) u There must be at least one shooting location in the flight path or waypoint sequence such that the shooting location is within view of the target object. When the conditions of mission observation distance, camera field of view, observation angle, occlusion ratio, and image resolution are simultaneously satisfied, the following definition is made: Otherwise, define: .

[0084] Effective coverage indication via fixed side and task-side effective coverage indication In the same coverage evaluation system, the effective coverage contribution of candidate micro-photograph sites and candidate UAV missions to the target object is uniformly measured.

[0085] Based on the effective coverage relationship between the fixed side and the mission side, and combined with budget constraints and engineering constraints, S400 constructs a set of actionable actions for deploying candidate micro-photography points and selecting UAV missions.

[0086] Based on the current remaining budget Brem and the set of selected micro-photography locations S k Selected UAV mission set U k The set of candidate micro-photography locations S, the set of candidate U-drone tasks U, and engineering constraints are used to construct the set of feasible micro-photography location deployment actions and the set of feasible U-drone task selection actions, respectively.

[0087] Fixed micro-video camera side engineering constraints include one or more of the following: installation feasibility constraints, energized safety distance constraints, power supply accessibility constraints, communication backhaul availability constraints, and construction accessibility constraints. Feasible micro-video camera deployment action set. Represented as: , and Satisfying the engineering constraints of the fixed micro-photograph side} in, Indicates the deployment of candidate micro-photograph points s i The action, Indicates candidate micro-photograph point s i Deployment costs, S k This represents the set of micro-shot points selected at the k-th decision step; Unmanned aerial vehicle (UAV) side engineering constraints include one or more of the following: mission cost constraints, endurance constraints, return margin constraints, no-fly zone constraints, altitude restrictions, take-off and landing point availability constraints, and mission time window constraints.

[0088] Among them, the range constraint and return margin constraint can be expressed as:

[0089] in, Indicates candidate drone missions u The estimated flight time Indicates candidate drone missions u The time required for the return flight, Indicates the maximum permissible flight time for the drone; No-fly zone constraints are represented as follows: ,in, For the mission trajectory, Gathering in the no-fly zone; Feasible drone mission selection action set Represented as: and Satisfy the engineering constraints on the UAV side; in, Indicates the selection of drone candidate tasks The action, Indicates candidate drone missions Task cost, This represents the set of drone missions selected at the k-th decision step; Based on the set of feasible micro-photography location deployment actions and the set of feasible UAV mission selection actions, construct the overall set of feasible actions for the current decision step. :

[0090] in, This indicates the termination action set. The `stop` action can be added when the remaining budget is insufficient, the action set is empty, the overall coverage or high-risk target coverage reaches a preset threshold, the incremental revenue for several consecutive steps is lower than a preset threshold, or the current decision step count reaches a preset maximum decision step count. .

[0091] S500: Construct a collaborative decision-making environment, combine the set of actionable actions, and use an action masking mechanism to shield non-executable actions in a unified action space.

[0092] like Figure 4As shown, in step S500, this embodiment applies action masks to non-executable actions based on the overall set of actionable actions generated in step S400, on the basis of a unified action space. Through the action masking mechanism, actions with insufficient budget, unmet engineering constraints, or those already selected can be masked, allowing the collaborative decision-making model to select the currently deployed action only from the set of actionable actions.

[0093] The collaborative decision-making environment includes a state space, a unified action space, a reward function, and an action masking mechanism; The state space includes high-risk target coverage status, normalized risk score, overall coverage rate, high-risk target coverage rate, remaining budget ratio, selected micro-photography location ratio, selected drone mission ratio, and summary information of actionable action set.

[0094] The unified action space A includes the deployment action of candidate micro-photography points, the drone mission selection action, and the termination action, which can be represented as:

[0095] Action masking is used to determine the overall set of possible actions. Mask non-executable actions in the unified action space A. Action mask Represented as: =1,

[0096] =0,

[0097] The original action probability distribution output by the collaborative decision-making model By performing action masking, we obtain the normalized probability distribution of possible actions:

[0098] Through the aforementioned action masking mechanism, the collaborative decision-making model selects the currently deployed action only from actions that satisfy budget constraints and engineering constraints.

[0099] The reward function is determined based on the increased risk coverage benefits brought by the current deployment action, the normalized deployment cost, and the penalty for violating constraints, and can be expressed as:

[0100] in, This represents the additional risk coverage return at the k-th decision step. This represents the normalized deployment cost of the k-th decision step. This indicates restrictions and penalties for violations. and Preset weighting coefficients; The return from the increased risk coverage can be expressed as:

[0101] in, Represents the target object Risk score, Indicates the target object at the k-th decision step. The coverage status. The normalized deployment cost can be expressed as:

[0102] in, Indicates the current deployment action The corresponding cost, B represents the total budget.

[0103] When expanding the candidate drone tasks using an online generation method, the online-generated candidate drone tasks are added to a candidate task pool with a preset capacity. When the number of candidate tasks exceeds the preset capacity, they are sorted and truncated according to the risk coverage benefit per unit cost. When the number of candidate tasks is less than the preset capacity, they are filled with empty tasks, and empty and non-executable tasks are masked by action masks to maintain the stability of the action output dimension of the collaborative decision-making model.

[0104] S600 invokes the trained collaborative decision-making model, selects the current deployment action based on the current state, target object risk score, remaining budget, and actionable action set after action masking, and updates the selected micro-photography point set, selected drone task set, target object coverage status, and remaining budget based on the current deployment action.

[0105] like Figure 5 As shown, in step S600, the collaborative decision-making model selects the current deployment action at each decision step. When the current deployment action is the candidate micro-photograph point deployment action, the selected micro-photograph point set, remaining budget, and target object coverage status are updated; when the current deployment action is the UAV mission selection action, the selected UAV mission set, remaining budget, and target object coverage status are updated; when the current deployment action is the termination action, the current deployment process is stopped, and the process proceeds to the result output stage.

[0106] In the k Each decision step, the collaborative decision-making model, is based on the current state. Output the action probability distribution and select the currently deployed action from the set of available actions under the action mask constraint. .

[0107] when At that time, the corresponding candidate micro-photograph points will be... Add the selected micro-auction locations to the set, and update the remaining budget and target object coverage status:

[0108]

[0109]

[0110] when At that time, the corresponding candidate drone mission will be... Add to the selected drone mission set and update the remaining budget and target coverage status:

[0111]

[0112]

[0113] when If this happens, the current deployment process will be stopped.

[0114] In one implementation, the collaborative decision-making model is implemented using a reinforcement learning model, preferably trained using the PPO algorithm. The collaborative decision-making model includes a policy network and a value network. The policy network outputs the original action probability distribution of each action in a unified action space based on the current state, and the value network estimates the value of the current state. The PPO algorithm is only one preferred implementation of the collaborative decision-making model; in other implementations, other reinforcement learning models, heuristic search models, greedy policy models, or mixed-integer programming models can also be used for action selection.

[0115] S700 When the termination condition is met, output the updated set of selected micro-photography points, the set of selected drone missions, the coverage status of the target object, and the remaining budget.

[0116] The termination conditions include at least one of the following: the remaining budget is insufficient to execute any non-termination action; there are no non-termination actions in the set of actionable actions that satisfy the engineering constraints; the overall coverage rate reaches a preset threshold; the coverage rate of high-risk targets reaches a preset threshold; the preset maximum number of decision steps is reached; or the collaborative decision-making model outputs a termination action.

[0117] Output fixed micro-photography point deployment plan, drone inspection task plan, and coverage and cost statistics; The fixed micro-photography point deployment plan includes the selected candidate micro-photography point number, spatial location, installation height, recommended orientation, deployment cost, and a list of corresponding target objects to be covered.

[0118] The UAV inspection mission plan includes the selected UAV candidate mission number, take-off and landing point, flight path or waypoint sequence, subset of inspection target objects, estimated flight time and mission cost.

[0119] The coverage and cost statistics include at least the overall coverage rate, high-risk target coverage rate, weighted coverage rate, list of uncovered target objects, deployment cost of fixed micro-photography sites, drone mission execution cost, and budget utilization rate.

[0120] The present invention will be further described below with reference to specific embodiments. It should be noted that the following embodiments are only used to illustrate the technical solutions of the present invention, and are not intended to limit the scope of protection of the present invention.

[0121] Taking a section of 28 consecutive towers in a 220kV transmission line as an example, the monitoring targets along the line mainly include the insulator area and the hardware connection area. Based on the tower ledger data, the transmission line GIS data, and the configuration rules of key tower parts, the monitoring targets in this section are abstracted into tower part-level target objects, and a total of 56 target objects are constructed, including 34 insulator target objects and 22 hardware target objects.

[0122] Based on the installation locations on the tower body, crossarm locations, maintenance platform locations, and other tower component locations that meet the installation conditions, 18 candidate micro-photography locations were generated; based on the locations of maintenance stations and power supply stations, 2 drone take-off and landing points were determined; further combining the drone take-off and landing points, the spatial distribution of transmission channels, the location of target objects, and drone operation constraints, 24 candidate drone tasks were generated. The total project budget for this embodiment was set at 280,000 yuan.

[0123] For each target object, features such as historical defect frequency, defect severity, equipment service life, pollution level, surrounding vegetation, meteorological risk level, and maintenance priority are extracted to construct a target object feature vector. This vector is then input into a fault probability prediction model to obtain the fault probability of each target object. Subsequently, the fault probabilities and the weights of the fault consequences for each target object are normalized to form a risk score for the target object. A higher risk score indicates that the target object should be given priority for effective coverage by fixed micro-photography equipment or drone inspection missions. Based on calculations, this embodiment identified 15 high-risk targets, 24 medium-risk targets, and 17 low-risk targets.

[0124] In this embodiment, for fixed micro-photography equipment, the maximum effective observation distance for insulator-type targets is set to 18 meters, and the maximum effective observation distance for hardware-type targets is set to 12 meters, with an observation angle threshold of 35 degrees. The effective coverage relationship on the fixed side is no longer judged solely by "visibility," but simultaneously considers distance conditions, line-of-sight conditions, viewing angle conditions, and image interpretation conditions.

[0125] Specifically, when the observation distance between the candidate micro-capture point and the target object does not exceed the maximum effective observation distance threshold for the corresponding component type, the two meet the line-of-sight requirement, the included angle is no greater than 35 degrees, and the number of imaging pixels, image clarity, or readability of the key area of ​​the target object in the acquired image is not lower than a preset threshold, the candidate micro-capture point is deemed to constitute effective fixed-side coverage of the target object; otherwise, it is deemed not to constitute effective fixed-side coverage. This avoids the situation where the object is only "visible" in a geometric sense, but the image quality is insufficient to support defect identification.

[0126] For candidate UAV missions, the inspection altitude range is set to 15m to 35m, the inspection speed to 8m / s, the maximum allowable flight time per mission to 45min, the return safety margin to 8min, the effective observation distance threshold to 20m, and the requirement for effective image interpretation that the minimum imaging width of the key target area is not less than 100 pixels. When at least one shooting position exists in the track or waypoint sequence of a candidate UAV mission, such that this shooting position simultaneously satisfies the mission observation distance condition, camera field of view condition, observation angle condition, occlusion ratio condition, and image resolution condition for the target object, the candidate UAV mission is considered to have effective coverage of the target object on the mission side.

[0127] In this embodiment, the effective coverage contribution of candidate micro-photographing points and candidate UAV missions to the target object is uniformly measured in the same coverage evaluation system by using fixed-side effective coverage indicators and mission-side effective coverage indicators. For target objects that have been effectively covered by fixed micro-photographing points, the additional coverage benefit is not recalculated when a UAV mission covers the target object again; for target objects that cannot meet the effective coverage requirements of fixed micro-photographing equipment, UAV missions can participate in collaborative deployment as supplementary coverage resources.

[0128] Furthermore, a set of feasible actions is constructed based on the total project budget, deployment costs of candidate micro-photography sites, UAV mission costs, and engineering constraints. Engineering constraints on the fixed micro-photography side include installation feasibility constraints, energized safety distance constraints, power supply accessibility constraints, communication backhaul availability constraints, and construction accessibility constraints; engineering constraints on the UAV side include mission cost constraints, endurance constraints, return margin constraints, no-fly zone constraints, altitude restriction constraints, takeoff and landing point availability constraints, and mission time window constraints.

[0129] In each decision-making step, the system generates a set of feasible micro-photography site deployment actions and a set of feasible drone mission selection actions based on the current remaining budget, the set of selected micro-photography sites, the set of selected drone missions, and the aforementioned engineering constraints. It then further constructs an overall set of feasible actions. Candidate micro-photography site deployment actions and drone mission selection actions that do not meet the budget or engineering constraints are masked using an action masking mechanism, ensuring that the collaborative decision-making model selects only from the range of executable actions.

[0130] In this embodiment, a state representation method combining explicit representation of high-risk targets and global statistical summaries is adopted. State information includes the current coverage status of high-risk targets, normalized risk score, overall coverage rate, high-risk target coverage rate, remaining budget percentage, selected micro-capture point percentage, selected UAV mission percentage, and summary information of the set of actionable actions. The number K of explicitly represented high-risk targets is set to 10. The unified action space includes 18 micro-capture point deployment actions, 24 UAV mission selection actions, and 1 termination action, totaling 43 actions.

[0131] The collaborative decision-making model can be implemented using a reinforcement learning model, preferably trained using the PPO algorithm. It should be noted that the PPO algorithm is only a preferred implementation in this embodiment; in other embodiments, other reinforcement learning models, heuristic search models, greedy policy models, or mixed-integer programming models can also be used for action selection.

[0132] In this embodiment, the PPO model training parameters are set as follows: learning rate of 0.0001, discount factor of 0.99, dominance smoothing coefficient of 0.95, pruning coefficient of 0.2, batch size of 256, and maximum training epochs of 3000. During training, a complete joint deployment process of a fixed micro-capture device and a drone mission is considered as one training epoch. At each decision step, the agent outputs the action probability distribution based on the current state, and selects the current deployment action after action masking. The environment updates the coverage state, remaining budget, set of selected micro-capture locations, and set of selected drone missions based on the selected action, and returns an immediate reward.

[0133] The immediate reward is determined based on the increased risk coverage benefit, normalized deployment cost, and constraint violation penalty brought about by the current deployment action. Specifically, the increased risk coverage benefit is determined by the target object's risk score and changes in coverage status; the normalized deployment cost is determined by the ratio between the current action cost and the total project budget; and the constraint violation penalty is used to punish actions that do not meet engineering constraints or fail dynamic validation. Through this reward mechanism, the model can prioritize deployment actions that generate new effective coverage for high-risk target objects, have low unit costs, and meet engineering constraints.

[0134] During the online deployment phase, the trained collaborative decision-making model is loaded, and the risk score, coverage status, remaining budget, and set of actionable actions for the current channel scenario are input. The model progressively outputs either a fixed micro-capture point deployment action or a drone mission selection action. If the current action is a fixed micro-capture point deployment action, the corresponding candidate micro-capture point is added to the set of selected micro-capture points, and the coverage status of the target object is updated according to the fixed-side effective coverage relationship of the micro-capture point. If the current action is a drone mission selection action, the corresponding candidate drone mission is added to the set of selected drone missions, and the coverage status of the target object is updated according to the mission-side effective coverage relationship of the drone mission. If the current action is a termination action, the deployment process is stopped.

[0135] In this embodiment, the deployment process terminates when one of the following conditions is met: the remaining budget is insufficient to execute any non-terminating action; there are no non-terminating actions in the set of available actions that satisfy the engineering constraints; the overall coverage or high-risk target coverage reaches a preset threshold; the maximum number of decision steps is reached; or the collaborative decision-making model outputs a termination action.

[0136] After model-based decision-making, this embodiment ultimately selected 6 fixed micro-photography locations and 4 drone missions. The system finally outputs a deployment list of fixed micro-photography locations, a list of drone inspection missions, total cost statistics, weighted coverage statistics, high-risk target coverage statistics, and a list of uncovered target objects.

[0137] According to statistics, the total cost in this embodiment is 266,440 yuan, the weighted coverage rate reaches 0.884, the high-risk target coverage rate reaches 0.9333, and the budget utilization rate reaches 95.16%. Among them, the weighted coverage rate is used to reflect the overall effective coverage level after considering the risk score of the target object, the high-risk target coverage rate is used to reflect the proportion of high-risk target objects that have achieved effective coverage, and the budget utilization rate is used to reflect the degree to which the deployment plan uses the project budget.

[0138] As can be seen from this embodiment, the present invention can integrate fixed micro-photography equipment and UAV inspection tasks into the same collaborative decision-making framework under budget constraints and engineering constraints. Through risk scoring-driven mechanisms, effective coverage determination, feasible action set construction, and action masking, it achieves priority and effective coverage of high-risk targets and outputs a deployment plan with engineering feasibility. In summary, without departing from the overall technical concept of the present invention, those skilled in the art can appropriately replace or adjust the risk scoring generation method, effective coverage determination method, candidate resource construction method, state and action expression method, collaborative decision-making algorithm, constraint handling mechanism, and online update mechanism.

[0139] Although these changes differ in their specific implementation paths, they still revolve around the core idea of ​​risk-driven, effective coverage determination, unified budget, and collaborative deployment of fixed micro-photography equipment and UAV missions under engineering constraints. They can all achieve the technical objectives of this invention to improve the effective coverage of high-risk targets, control overall costs, and enhance the engineering feasibility of deployment schemes. Therefore, they should be considered to fall within the protection scope of this invention.

[0140] In another aspect, the present invention also discloses a computer-readable storage medium storing a computer program, which, when executed by a processor, causes the processor to perform the steps of the method described above.

[0141] In another aspect, the present invention also discloses a computer device, including a memory and a processor, wherein the memory stores a computer program, and when the computer program is executed by the processor, the processor performs the steps of the method described above.

[0142] In another embodiment provided in this application, a computer program product containing instructions is also provided, which, when run on a computer, causes the computer to execute any of the above embodiments of the method for joint deployment of micro-photography equipment and drones in power grid transmission channels.

[0143] It is understood that the systems, devices, and storage media provided in the embodiments of the present invention correspond to the methods provided in the embodiments of the present invention, and the explanations, examples, and beneficial effects of the relevant content can be referred to the corresponding parts of the above methods.

[0144] In the above embodiments, implementation can be achieved entirely or partially through software, hardware, firmware, or any combination thereof. When implemented using software, it can be implemented entirely or partially in the form of a computer program product. The computer program product includes one or more computer instructions. When the computer program instructions are loaded and executed on a computer, all or part of the processes or functions described in the embodiments of this application are generated. The computer can be a general-purpose computer, a special-purpose computer, a computer network, or other programmable device. The computer instructions can be stored in a computer-readable storage medium or transmitted from one computer-readable storage medium to another. For example, the computer instructions can be transmitted from one website, computer, server, or data center to another website, computer, server, or data center via wired (e.g., coaxial cable, fiber optic, digital subscriber line (DSL)) or wireless (e.g., infrared, wireless, microwave, etc.) means. The computer-readable storage medium can be any available medium that a computer can access or a data storage device such as a server or data center that integrates one or more available media. The available medium can be a magnetic medium (e.g., floppy disk, hard disk, magnetic tape), an optical medium (e.g., DVD), or a semiconductor medium (e.g., solid-state disk (SSD)).

[0145] It should be noted that, in this document, relational terms such as "first" and "second" are used only to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Without further limitations, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes said element.

[0146] The various embodiments in this specification are described in a related manner. Similar or identical parts between embodiments can be referred to mutually. Each embodiment focuses on describing the differences from other embodiments. In particular, the system embodiments are basically similar to the method embodiments, so the description is relatively simple; relevant parts can be referred to the descriptions of the method embodiments.

[0147] The above embodiments are only used to illustrate the technical solutions of the present invention, and are not intended to limit it. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A method for jointly deploying micro-photography equipment and drones in power grid transmission channels, characterized in that, The method includes the following steps: S100: Acquire basic data of the power transmission channel, and construct a set of target objects, a set of candidate micro-photographing points, a set of UAV take-off and landing points, and a set of candidate UAV tasks based on the tower-level monitoring objects; S200. Based on the target object set, historical defect records, historical alarm statistics, environmental characteristics, and meteorological characteristics, determine the risk score for each target object; S300: Based on the risk score of each target object, calculate the effective coverage relationship of the candidate micro-photograph points to the fixed side of the target object, and the effective coverage relationship of the candidate UAV missions to the mission side of the target object. S400, based on the effective coverage relationship between the fixed side and the mission side, combined with budget constraints and engineering constraints, constructs a set of actionable actions for candidate micro-photography point deployment actions and UAV mission selection actions; S500: Construct a collaborative decision-making environment, combine the set of actionable actions, and use the action masking mechanism to shield non-executable actions in the unified action space; S600 invokes the trained collaborative decision-making model, selects the current deployment action based on the current state, target object risk score, remaining budget, and actionable action set after action masking, and updates the selected micro-photography point set, selected drone mission set, target object coverage status, and remaining budget based on the current deployment action; S700 When the termination condition is met, output the updated set of selected micro-photography points, the set of selected drone missions, the coverage status of the target object, and the remaining budget.

2. The method for jointly deploying micro-photography equipment and drones in power grid transmission channels according to claim 1, characterized in that, The methods for constructing the target object set, candidate micro-photography location set, UAV take-off and landing point set, and UAV candidate task set in step S100 include: The method for constructing the target object collection is as follows: The monitoring objects in the power transmission channel are abstracted into target objects at the tower level, forming a set of target objects. Each target object corresponds to a key part of the tower in the power transmission channel that is related to defect monitoring; Among them, the key parts of the tower include at least one of the following: tower head area, crossarm area, insulator string area, hardware connection area, and clamp connection area; Each target object must include at least the target object number, the tower number to which it belongs, and its spatial location. Component type label Recommended observation direction vector Historical defect records, historical alarm statistics, equipment attributes, environmental characteristics, and meteorological characteristics; Among them, spatial location Represented using a unified projected coordinate system; The method for constructing the candidate micro-photography location set is as follows: Based on the tower structure, installation conditions, and monitoring requirements, a set of candidate micro-monitoring points is generated from the locations of installable components on the tower, crossarm locations, maintenance platform locations, tower body hanging points, and other locations that meet the installation conditions. ; Each candidate micro-photography location should include at least the location number and the location's spatial location. Installation height range, adjustable orientation range, lens parameters, power supply method, communication and data transmission method, and deployment cost. Installation feasibility markers ; Among them, candidate micro-photography locations Deployment costs It consists of the following parts: in, For equipment costs, To maintain costs, For communication backhaul costs, For power supply costs; when At that time, it indicates the candidate micro-photograph point. The installation conditions are met and the system can proceed to the next set of feasible actions; when At that time, it indicates the candidate micro-photograph point. It does not meet the requirements for structural installation, electrical safety distance, power supply and communication conditions, or construction accessibility, and is therefore eliminated in the subsequent decision-making process; The method for constructing the set of drone take-off and landing points is as follows: A set of drone take-off and landing points is generated based on maintenance sites, power supply stations, inspection stations, or temporary take-off and landing areas. Each UAV take-off and landing point should include at least the take-off and landing point number, spatial location, site availability, service radius, take-off and landing time window, and corresponding airspace restriction information. Construct a set of candidate drone tasks: Based on the set of UAV take-off and landing points, the set of target objects, the spatial distribution of power transmission channels, and UAV operational constraints, a set of candidate UAV tasks is generated. ; Each drone candidate task At least include the mission number and the landing / takeoff point. target object subset Track or waypoint sequence, flight altitude parameters, flight speed parameters, hovering strategy parameters, estimated flight time Task cost ; Task cost of drone mission u Represented as: in, To reduce fixed costs, This is the unit time cost coefficient. For the estimated flight time; When the mission flight time is determined by the track length, flight speed, and hovering duration, it is expressed as: in, The length of the flight path. For flight speed, For the first The hovering time at each hovering point; The set of candidate drone tasks can be constructed by pre-generating a set of candidate tasks, or by generating new candidate drone tasks online based on the currently uncovered target objects, real-time airspace conditions, and remaining budget when the candidate tasks cannot meet the current coverage requirements.

3. The method for jointly deploying micro-video equipment and drones in power grid transmission channels according to claim 1, characterized in that, The methods for determining the risk score of each target object in step S200 include: For the target object collection Each target object in Construct the feature vector of the target object ; Wherein, the target object feature vector It should include at least one or more of the following: the number of historical defects of the target object, the severity of the defects, the frequency of historical alarms, the service life of the equipment, the component type, the pollution level, the channel environmental risk factors, the meteorological risk factors, and the operation and maintenance priority information; The feature vector of the target object Input the fault probability prediction model to obtain the target object Failure probability : in, For failure probability prediction models, ; Based on the failure probability Determine the risk score wt of the target object t: in, It is a monotonically increasing mapping function; target object The overall risk score is expressed as follows: in, and For preset weighting coefficients, This is the normalization function; Based on risk score The target objects are sorted, and those with risk scores higher than a preset risk threshold are identified as high-risk target objects, or a preset number of target objects with the highest risk scores are identified as high-risk target objects. The high-risk target objects are used for subsequent effective coverage statistics, reward function calculation, and deployment action selection, so that fixed micro-photography equipment and drone missions prioritize coverage of target objects with higher risk scores.

4. The method for jointly deploying micro-photography equipment and drones in power grid transmission channels according to claim 1, characterized in that, The calculation methods for the effective coverage relationship on the fixed side and the effective coverage relationship on the mission side in step S300 include: For the candidate micro-photography point set S For each candidate micro-shot point s in the target object set T and for each target object t in the target object set T, calculate the candidate micro-shot point. For the target object Fixed-side effective coverage indication ; Fixed side effective coverage indication Determined based on distance conditions, line-of-sight conditions, viewing angle conditions, and image interpretation conditions; The distance conditions include: in, Candidate micro-photography locations Spatial location, For the target object Spatial location, To be with the target object The maximum effective observation distance threshold corresponding to the component type; Line-of-sight conditions include: candidate micro-photograph locations With the target object There is no obstruction between them, or the degree of obstruction does not exceed a preset obstruction threshold; among them, the following is adopted. Indicates candidate micro-photograph points To the target object The line-of-sight result, when the line-of-sight requirements are met. ;otherwise, ; Viewpoint conditions include: based on the target object Recommended observation direction vector and candidate micro-photography locations Point to target object line of sight direction vector Calculate the angle of observation And determine the included angle of observation. Is it not greater than the preset observation angle threshold? ; Among them, the line-of-sight direction vector Represented as: The observed angle Ang(s,t) is expressed as: when At that time, it was considered that the perspective condition was met; Image interpretation criteria include: target object The key area is the candidate micro-photography location. The number of imaging pixels, image clarity, or readability in the acquired image shall not be lower than the corresponding preset threshold. Based on candidate micro-photo locations Lens parameters, candidate macro shooting points With the target object The observation distance between them and the target object The actual dimensions of the key areas are used to calculate the target object. The number of imaging pixels in the key area; when the number of imaging pixels in the key area is not lower than a preset pixel threshold, the image interpretation condition is considered to be met. When candidate micro-photo locations With the target object When the distance condition, line-of-sight condition, viewing angle condition, and image interpretation condition are all satisfied simultaneously, the following definition applies: Otherwise, define: For each UAV candidate task in the UAV candidate task set U u and each target object in the target object set T Calculate candidate tasks for unmanned aerial vehicles u For the target object Task-side effective coverage indication ; When drone candidate tasks u There must be at least one shooting location in the flight path or waypoint sequence such that the shooting location is within view of the target object. When the conditions of mission observation distance, camera field of view, observation angle, occlusion ratio, and image resolution are simultaneously satisfied, the following definition applies: Otherwise, define: The task observation distance condition refers to the distance between the UAV's shooting position and the target object. The distance between them does not exceed the preset task observation distance threshold; Effective coverage indication via fixed side and task-side effective coverage indication In the same coverage evaluation system, the effective coverage contribution of candidate micro-photograph sites and candidate UAV missions to the target object is uniformly measured.

5. The method for jointly deploying micro-video equipment and drones in power grid transmission channels according to claim 1, characterized in that: In step S400, the method for constructing a set of actionable actions for candidate micro-photography site deployment actions and UAV mission selection actions based on budget constraints and engineering constraints includes: Based on the current remaining budget Brem and the set of selected micro-photography locations S k Selected UAV mission set U k Given the candidate micro-photo location set S, the candidate U drone mission set U, and engineering constraints, we construct the feasible micro-photo location deployment action set and the feasible U drone mission selection action set, respectively. Among them, the engineering constraints on the fixed micro-video side include one or more of the following: installation feasibility constraints, live safety distance constraints, power supply accessibility constraints, communication backhaul availability constraints, and construction accessibility constraints. Installation feasibility constraints are determined by installation feasibility markers. It means that when At that time, it indicates the candidate micro-photograph point. Installation conditions are met; when At that time, it indicates the candidate micro-photograph point. Installation conditions not met; The set of feasible micro-photography point deployment actions is represented as follows: , ,and Satisfy the engineering constraints of the fixed micro-photograph side; in, This represents the current state at the k-th decision step. Indicates the deployment of candidate micro-photograph locations The action, Indicates candidate micro-photograph points The deployment cost, Brem represents the current remaining budget. Indicates the first The set of micro-shot points selected at each decision step; Unmanned aerial vehicle (UAV) side engineering constraints include one or more of the following: mission cost constraints, endurance constraints, return-to-home margin constraints, no-fly zone constraints, altitude restriction constraints, take-off and landing point availability constraints, and mission time window constraints; the endurance constraints and return-to-home margin constraints are expressed as follows: in, This indicates the estimated flight time of the candidate drone mission u. This indicates the return time required for the drone candidate mission u. Indicates the maximum permissible flight time for the drone; No-fly zone constraints are represented as follows: in, For the mission trajectory, Gathering in the no-fly zone; Altitude restrictions are a candidate task for unmanned aerial vehicles (UAVs). u The mission execution altitude is within the permissible flight altitude range and does not exceed the maximum flight altitude set by regulatory or operational requirements; The set of feasible drone mission action selections is represented as follows: ,and in, Indicates the selection of drone candidate tasks The action, Indicates candidate drone missions Task cost, Indicates the first k The set of drone missions has been selected at each decision-making step; Based on the set of feasible micro-photography location deployment actions and the set of feasible UAV mission selection actions, the first... k The overall set of actionable actions for each decision step: in, Indicates the set of actions to be terminated; Add the stop action. The remaining budget is insufficient to execute any non-terminating action; the feasible micro-photograph point deployment action set and the feasible drone mission selection action set are both empty; the overall coverage rate has reached the preset overall coverage rate threshold; the coverage rate of high-risk targets has reached the preset high-risk coverage threshold; the additional risk coverage benefit of consecutive preset number decision steps is lower than the preset benefit threshold; or the current number of decision steps has reached the preset maximum number of decision steps. Before action selection, non-executable actions in the unified action space are filtered based on the overall set of actionable actions, so that the subsequent collaborative decision-making model selects the currently deployed action only from those actions that meet budget constraints and engineering constraints. Before or during the execution of an action, the selected action is dynamically verified based on real-time airspace status, changes in equipment availability, changes in weather conditions, or task execution feedback. If the selected action does not meet the engineering constraints during the dynamic verification phase, a constraint violation penalty is triggered, and a new deployment action is selected based on the updated set of actionable actions.

6. The method for jointly deploying micro-video equipment and drones in power grid transmission channels according to claim 1, characterized in that, The method for constructing the collaborative decision-making environment in step S500 includes: The state space includes a high-risk target explicit representation block, a global coverage statistics block, a budget and schedule block, and a summary block of the set of actionable actions; Among them, the high-risk target explicit representation block is used to record the current coverage status and normalized risk score of the target objects ranked first in risk score; The global coverage statistics block should include at least the overall coverage rate, the coverage rate of high-risk targets, and the gap ratio; The budget and schedule block should at least include the remaining budget percentage, the selected micro-photography location percentage, and the selected drone mission percentage; The action set summary block should include at least the proportion of action to deploy feasible micro-photo locations and the proportion of action to select feasible drone missions; The unified action space A includes a set of candidate micro-photography point deployment actions, a set of UAV mission selection actions, and a termination action; among which, the candidate micro-photography point deployment actions are represented as follows: The drone mission selection action is represented as The action to be terminated is indicated by "stop"; Based on the overall set of actionable actions Apply an action mask to actions in the unified action space A. , among which, when hour, =1; when hour, =0; When there are no feasible non-terminating actions in the set of feasible actions, add the terminating action stop to the set of feasible actions, so that the action probability distribution after action masking contains at least one executable action. The original action probability distribution output by the collaborative decision-making model By performing action masking, we obtain the normalized probability distribution of possible actions: The reward function is determined based on the additional risk coverage benefits brought by the current deployment action, the normalized deployment cost, and the penalty for constraint violations, and is expressed as: in, This represents the additional risk coverage return at the k-th decision step. This represents the normalized deployment cost of the k-th decision step. This indicates restrictions and penalties for violations. and Preset weighting coefficients; The additional risk coverage benefit is determined based on the target object's risk score and the change in the target object's coverage status; when a deployment action causes a target object with a high risk score to change from an uncovered state to an effectively covered state, the corresponding additional risk coverage benefit is obtained. The normalized deployment cost is determined based on the ratio between the deployment cost of the micro-photography point corresponding to the current deployment action or the cost of the drone mission and the total budget. When expanding the candidate drone tasks using an online generation method, the online-generated candidate drone tasks are added to a candidate task pool with a preset capacity. When the number of candidate tasks exceeds the preset capacity, they are sorted and truncated according to the risk coverage benefit per unit cost. When the number of candidate tasks is less than the preset capacity, they are filled with empty tasks, and empty and non-executable tasks are masked by action masks to maintain the action output of the collaborative decision-making model.

7. The method for jointly deploying micro-photography equipment and drones in power grid transmission channels according to claim 1, characterized in that, The method for updating the selected micro-photography point set, the selected drone mission set, the target object coverage status, and the remaining budget based on the current deployment action in step S600 includes: Collaborative decision-making models include policy networks and value networks; The policy network is used to output the original action probability distribution of each action in the unified action space based on the current state, and the value network is used to estimate the value of the current state. In each training round, a complete joint deployment process of fixed micro-photography equipment and UAV mission is considered as a training round. The environment is initialized based on the basic data of the power transmission channel, the risk score of the target object, the set of candidate micro-photography points, the set of candidate UAV missions, the budget level, and the engineering constraints. In the k-th decision step, the policy network determines the current state. Output the original action probability distribution, and obtain the action probability distribution based on the action masking mechanism. Select the current deployment action from the possible action probability distribution. ; when Update the set of selected micro-photo locations at that time. Update remaining budget and according to Indicator Update the coverage status of each target object; when Update the selected drone mission set at that time. Update remaining budget And based on the effective coverage indication on the mission side. Update the coverage status of each target object; when If necessary, terminate the current training round or the current deployment process; The training process of reinforcement learning models forms a training trajectory based on immediate rewards, state transition results, and termination judgments, and updates the parameters of collaborative decision-making models through policy optimization.

8. A system for the joint deployment of micro-photography equipment and drones in power grid transmission channels, characterized in that, include: The data acquisition module is used to acquire basic data of the power transmission channel; The target object and resource set construction module is used to construct a target object set, a candidate micro-photography location set, a UAV take-off and landing point set, and a UAV candidate task set based on the power transmission channel basic data. The risk scoring module is used to determine the risk score of each target object based on the target object set, historical defect records, historical alarm statistics, environmental characteristics, and meteorological characteristics. The effective coverage calculation module is used to calculate the fixed-side effective coverage relationship of candidate micro-photograph points to the target object, and the task-side effective coverage relationship of candidate UAV missions to the target object, respectively. The engineering constraint processing module is used to construct a set of actionable actions for candidate micro-photography point deployment actions and UAV mission selection actions based on budget constraints and engineering constraints. The collaborative decision-making environment construction module is used to construct the state space, unified action space, reward function, and action masking mechanism. The deployment action selection and status update module is used to call the trained collaborative decision-making model to select the current deployment action, and update the selected micro-photography point set, the selected drone mission set, the target object coverage status and the remaining budget based on the current deployment action; The results output module is used to output the fixed micro-photography point deployment plan, the drone inspection task plan, and the coverage and cost statistics when the termination conditions are met.

Images