Multi-rotor aircraft cooperative inspection method and system for distribution network based on improved epsilon-constraint optimization under typhoon influence

Abstract

This invention discloses a method and system for collaborative inspection of power distribution networks by multiple unmanned aerial vehicles (UAVs) under typhoon influence, based on improved epsilon-constraint optimization. The method includes: first, designing a node risk assessment method, which integrates physical, informational, and social risks and introduces an electrical quantity correction factor to accurately identify key nodes; second, designing a path planning method, which aims to minimize the number of UAVs and uses the longest inspection time as a constraint to construct a trajectory planning model, and employs a two-stage solution strategy consisting of mixed-integer linear programming and a backoff greedy algorithm to obtain the optimal flight plan; finally, designing a take-off and landing point selection method, which comprehensively considers the typhoon path and node priority distribution, establishes a dynamic take-off and landing point optimization model, and completes the selection of the best location. The system includes: a data acquisition and input module, a dynamic priority assessment module, a collaborative trajectory planning module, a dynamic take-off and landing point optimization module, and a scheme output and scheduling module.

Description

Technical Field

[0001] This invention relates to the fields of power system emergency management, UAV collaborative control and operations optimization technology, specifically to a risk assessment and improvement method based on cyber-physical-social system coupling applicable to extreme weather disaster scenarios such as typhoons. - A method and system for multi-UAV collaborative inspection of power distribution networks based on constrained optimization theory. Background Technology

[0002] In recent years, extreme weather events such as typhoons have occurred frequently, posing a serious threat to the safe and stable operation of power distribution networks. Rapid and accurate fault inspection after a disaster is crucial for restoring power supply. Unmanned aerial vehicle (UAV) technology, due to its flexibility, efficiency, and lack of ground-based limitations, has been widely used in power grid inspections. However, achieving efficient collaborative inspections of multiple UAVs in the highly dynamic and destructive specific disaster scenario of typhoons still faces significant challenges.

[0003] 1. Insufficient dynamic assessment of node risks: Existing inspection strategies often rely on static network topology importance or simple physical damage probability to assess node priority. They lack integrated consideration of multi-dimensional risks such as information layer (e.g., communication interruption) and social layer (e.g., user impact). Furthermore, they fail to effectively account for the dynamic changes in the electrical operation status of the distribution network (e.g., voltage overruns, line overloads) during typhoons. This results in the inability to accurately identify the critical nodes that truly need inspection, leading to the misallocation of limited emergency resources.

[0004] 2. Complex Multi-UAV Cooperative Path Planning: Multi-UAV cooperative inspection is a typical multi-objective, strongly constrained combinatorial optimization problem. It requires balancing multiple objectives, such as the number of UAVs, total inspection time, and coverage of high-priority nodes, within an extremely short time window. Existing methods, such as traditional vehicle pathing problem algorithms, often suffer from low computational efficiency, susceptibility to local optima, or difficulty in simultaneously satisfying multiple conflicting objectives when solving such large-scale, highly dynamic optimization problems, failing to provide high-quality solutions within the tight timeframes of emergency response.

[0005] 3. Static and Rigid Take-off and Landing Site Selection: Most studies assume that UAVs take off and land from one or more fixed bases. However, typhoon paths are dynamically changing, and fixed take-off and landing sites may quickly enter dangerous wind circles and become unusable, or be too far from high-risk areas that change over time, resulting in low inspection efficiency. Existing research lacks in-depth discussion of dynamic, safe, and efficient UAV take-off and landing site selection and recovery strategies.

[0006] Therefore, there is an urgent need for a comprehensive solution that can integrate dynamic risk perception, multi-objective resource collaborative optimization, and dynamic environmental adaptive decision-making to improve the overall efficiency and system resilience of drone-based collaborative inspections of power distribution networks during typhoon disasters.

[0007] A search revealed that Chinese invention patent CN115265486A discloses a method for autonomous drone-based surveying of power transmission lines after a typhoon, based on a typhoon monitoring system. After receiving typhoon warning information, the typhoon warning platform for power transmission lines intelligently dispatches drones near the geographical location indicated in the warning information. The dispatched drones automatically receive dispatch information and initiate a differentiated disaster survey mode to conduct surveys, returning the results to a remote command center. During the survey, the drones continuously monitor network signal coverage. If no network signal is detected, the survey results are automatically stored in the drone's internal storage. If a network signal is detected, the results are transmitted to the remote command center promptly or in real-time. This invention intelligently dispatches drones for surveys based on information from typhoons, reducing human intervention and significantly improving the efficiency of disaster emergency response.

[0008] The technical comparison between this application and the aforementioned patent is as follows:

[0009] 1. The patent "Autonomous Survey Method of Transmission Lines after Typhoon Disaster by UAV Based on Typhoon Monitoring System" discloses an autonomous survey method of transmission lines after typhoon disaster by UAV based on typhoon monitoring system. It can intelligently call nearby UAVs through typhoon early warning platform, automatically start the wind disaster differentiated survey mode to conduct survey, and return the survey results to the remote command center.

[0010] This invention focuses not only on automatic scheduling of UAVs, but also on the implementation of a series of collaborative optimization schemes, such as dynamic node priority assessment that integrates information, physical and social multi-dimensional risks, multi-UAV collaborative trajectory planning constrained by minimizing the number of UAVs and the longest single-UAV inspection time, and take-off and landing point selection optimization in dynamic environments.

[0011] There are fundamental differences between the two in terms of implementing technical solutions and solving problems.

[0012] 2. The patent "Autonomous Survey Method of Transmission Lines after Typhoon Disaster by UAV Based on Typhoon Monitoring System" mainly provides a solution for the autonomous survey of transmission lines by a single UAV after a typhoon. It mainly addresses the problem of "how to quickly view the site" and replaces manual labor to improve the timeliness of information collection.

[0013] This invention addresses the scenario of multi-drone collaborative inspection under the influence of typhoons, and focuses on how to accurately identify key nodes, efficiently allocate limited resources, and dynamically adapt to environmental changes in dynamic typhoon environments. It not only meets the needs of single-drone automated inspection, but also realizes the intelligent decision-making needs of complex scenarios such as multi-drone collaboration, path optimization, and dynamic adaptation of take-off and landing points.

[0014] The two differ fundamentally in their application scenarios and research objectives.

[0015] A search revealed that Chinese invention patent CN116009584A discloses an automatic route conversion method for UAV inspection of power transmission lines, comprising the following steps: Step S1: Obtain the route scene to be inspected and determine the photo point information based on the scene; Step S2: Extract the corresponding waypoint information based on the photo points; Step S3: Reassemble and stitch the waypoint information to obtain a new autonomous inspection route; Step S4: Perform flight safety verification based on the route path length and the number of photo points of the new autonomous inspection route; Step S5: Save the verified route to the database for use by the on-site inspection terminal. This invention utilizes existing refined routes and automatically converts routes based on different types of inspection needs by extracting relevant waypoint information. The converted routes can be used in scenarios such as channel inspection, infrared inspection, and emergency inspection (typhoons, floods, ice storms, wildfires), achieving fully automated UAV inspection of overhead power transmission lines in different scenarios.

[0016] The technical comparison between this application and the aforementioned patent is as follows:

[0017] 1. The patent "An Automatic Conversion Method for UAV Inspection Routes of Transmission Lines" discloses an automatic conversion method for UAV inspection routes of transmission lines. By selecting and reorganizing photo points from an existing refined route database based on the inspection scenario, a new single-machine autonomous inspection route is generated.

[0018] This invention focuses not only on the template-based conversion of flight routes, but also on the implementation of a series of collaborative optimization schemes, such as dynamic node priority assessment that integrates information, physical and social multidimensional risks, multi-aircraft collaborative trajectory planning constrained by minimizing the number of UAVs and the longest single-aircraft inspection time, and take-off and landing point selection optimization in dynamic environments.

[0019] There are fundamental differences between the two in terms of implementing technical solutions and solving problems.

[0020] 2. The patent "An Automatic Conversion Method for UAV Inspection Flight Routes of Transmission Lines" mainly provides a solution for the rapid reuse of flight routes in different inspection scenarios. It mainly addresses the problem of "how to improve the efficiency of flight route generation" and realizes the automatic generation of flight routes for channel inspection, infrared inspection and emergency inspection through template conversion.

[0021] This invention addresses the scenario of multi-UAV collaborative inspection under the influence of typhoons. It aims to solve the problems of accurately identifying key nodes, efficiently allocating limited resources, and dynamically adapting to environmental changes in dynamic disaster environments. It not only meets the requirements of single-UAV route reuse, but also realizes the intelligent decision-making needs of complex scenarios such as multi-UAV collaboration, path optimization, and dynamic adaptation of take-off and landing points.

[0022] The two differ fundamentally in their application scenarios and research objectives.

[0023] A search revealed a Chinese invention patent with publication number CN120323377A, which discloses a method for typhoon disaster prevention and rapid post-disaster recovery in deep-sea cage aquaculture. Before a disaster, a drone equipped with a high-definition camera is used to take aerial photographs of the cage's surface structure. An underwater robot uses sonar for layered scanning and the camera to collect underwater images, generating pre-disaster structural and fish population data. Three-dimensional modeling is used to analyze and repair weak points in the frame, netting, and anchor chains. Combined with typhoon information, stocking density is optimized to reduce the risk of facility damage and fish loss. After the disaster, the structural and fish population data are repeated, using the pre-disaster baseline as a reference to compare and assess the degree of frame damage, netting breakage, anchor chain breakage, and fish escape or mortality, generating a graded repair plan and disaster relief plan. This method improves detection accuracy and assessment efficiency through multi-device collaboration and data fusion, quickly locating damaged areas and optimizing recovery strategies, providing an efficient solution for disaster prevention, mitigation, and rapid post-disaster recovery in deep-sea cage aquaculture.

[0024] The technical comparison between this invention and the aforementioned patent is as follows:

[0025] 1. The patent "A method for preventing typhoon disasters and rapid recovery after disasters in deep-sea cage aquaculture" discloses a method for preventing typhoon disasters and rapid recovery after disasters in deep-sea cage aquaculture. The method uses drones and underwater robots to acquire data on the above-water and underwater structures of the cages and the fish population, respectively. The method compares and evaluates the damage to the facilities and the loss of the fish population through pre-disaster and post-disaster 3D modeling, and generates a repair plan.

[0026] This invention does not involve multi-device disaster assessment, but rather the implementation of a series of collaborative optimization schemes, including dynamic node priority assessment that integrates information-physical-social multi-dimensional risks, multi-drone collaborative trajectory planning constrained by minimizing the number of drones and the longest single-drone inspection time, and take-off and landing point selection optimization in dynamic environments.

[0027] There are fundamental differences between the two in terms of implementing technical solutions and solving problems.

[0028] 2. The patent "A method for typhoon disaster prevention and rapid post-disaster recovery in deep-sea cage aquaculture" mainly provides solutions for disaster assessment and post-disaster recovery in deep-sea cage aquaculture. It mainly addresses the problem of "how to detect and assess disaster losses" and provides decision-making basis for post-disaster repair by collecting data through multi-device collaboration.

[0029] This invention addresses the scenario of multi-UAV collaborative inspection under the influence of typhoons. It aims to solve the problems of accurately identifying key nodes, efficiently allocating limited resources, and dynamically adapting to environmental changes in dynamic disaster environments. It not only meets the needs of disaster assessment but also enables intelligent decision-making in complex scenarios such as multi-UAV collaboration, path optimization, and dynamic adaptation of take-off and landing points.

[0030] The two differ fundamentally in their application scenarios and research objectives. Summary of the Invention

[0031] This invention aims to overcome the shortcomings of the prior art and provide a multi-UAV collaborative inspection method and system for power distribution networks under typhoon conditions based on improved epsilon-constraint optimization. This method constructs a three-tiered closed-loop framework of "dynamic risk assessment - collaborative trajectory planning - dynamic takeoff and landing adaptation" to achieve precise deployment, efficient collaboration, and safe operation of inspection resources.

[0032] To achieve the above objectives, the technical solution adopted by the present invention is as follows:

[0033] A multi-UAV collaborative inspection method for power distribution networks based on improved epsilon-constraint optimization under typhoon influence includes the following steps: S1, Node Dynamic Priority Evaluation Phase: Input typhoon wind field data, distribution network topology and operating parameters, equipment ledger and social attribute data, construct a comprehensive risk assessment model that integrates physical risk, information risk and social risk, and introduce an electrical quantity correction factor based on forward and backward power flow calculation for dynamic correction, output the dynamic priority score and level classification of all nodes. S2, Multi-UAV Collaborative Flight Path Planning Phase: With minimizing the number of drones in use as the core optimization objective, a maximum allowable inspection time per drone is set. for Constraints, constructing based on improvement - The constraint method path optimization model adopts a two-stage hybrid solution strategy: the first stage clusters nodes based on their geographical location and priority to generate a high-quality candidate path set; the second stage uses an integer linear programming model or a backoff greedy algorithm to select paths and optimize coverage, and finally outputs the inspection paths of each UAV. S3, Dynamic Takeoff and Landing Point Joint Optimization Phase: Based on the typhoon's real-time location, radius of influence, and spatial distribution of high-priority nodes, the starting point with the highest comprehensive score is dynamically selected from the preset candidate points. After the mission is completed, based on the spatiotemporal accessibility judgment, the landing points adjacent to the terminal are intelligently merged, and a landing point scheme that can effectively improve recovery efficiency while ensuring safety is output.

[0034] As a preferred technical solution of the present invention: In step S1: A comprehensive risk assessment model integrating physical risk, information risk, and social risk is constructed, as follows: Comprehensive Risk of Typhoon Nodes Due to physical risks Information risk Social risks It consists of three linearly superimposed dimensions, following the basic framework of "risk R = failure probability Pr × failure loss S": (1); (2); in, The weighting coefficients for the three-dimensional risk of load nodes and satisfying ; To characterize the differentiated responses of different types of loads to typhoon disturbances, the nodes are first defined. Typhoon distance attenuation function from the typhoon center: (3); in, For nodes Distance to the center of the typhoon It is the attenuation constant; The risk calculations for each dimension are as follows: Physical risks under normal operating conditions Used to measure the probability of physical failure under normal operating conditions of equipment structural vulnerability and environmental threats. By comprehensively considering factors such as equipment aging, current load levels, environmental risks, and historical failure rates, it was found that the probability of failure during a typhoon is amplified due to environmental degradation, thus yielding the probability of physical failure under typhoon conditions. : (4); in This is the physical failure probability amplification factor. This is the typhoon distance attenuation function; Physical failure losses under normal operating conditions This includes equipment damage and repair costs, economic losses to users due to power outages, and maintenance operation costs. The increased difficulty of emergency repairs under typhoon conditions amplifies these losses, resulting in physical damage losses due to typhoons. : (5); in This is the amplification factor for physical failure losses. This is the typhoon distance attenuation function; Information risks under normal operating conditions This reflects the vulnerability of communication systems and information security, specifically the probability of information failure under normal operating conditions. The probability of information failure under typhoon conditions is obtained by assessing information security protection levels, historical information attack frequencies, and the probability of communication outages caused by typhoons. This is taken into account the physical damage to communication facilities caused by strong winds, which increases the risk of outages. : (6); in This is the information failure probability amplification factor. This is the typhoon distance attenuation function; Information failure loss under normal operating conditions This includes increased data leakage costs due to the difficulty of system recovery and amplified cascading effects, monitoring and operational losses caused by control system failures, the scope of communication interruptions, and information security repair costs. Under the influence of typhoons, the increased difficulty of repairs leads to amplified losses, resulting in information failure losses during typhoons. : (7); in This is the amplification factor for information failure loss. This is the typhoon distance attenuation function; Social risks under normal working conditions Used to measure the impact of node failure on social operation and public safety, it represents the probability of social failure under normal operating conditions. The social risk during a typhoon is related to the user sensitivity of the load supplied by the node and is affected by the regional emergency response capability. Due to user gathering and emergency demand, the social risk increases sharply. This leads to the probability of social failure during a typhoon. : (8); in This is the amplification factor for the probability of social failures. This is the typhoon distance attenuation function; Social failure losses under normal operating conditions Due to the risks of critical infrastructure service disruptions, public safety risks, and social order disruptions, when quantified using the expected load loss index, social losses during typhoons are exacerbated by public service disruptions and secondary disasters, thus yielding the social failure losses during typhoons. : (9); in This is the social failure loss amplification factor. This is the typhoon distance attenuation function.

[0035] As a preferred technical solution of the present invention: In step S1 By introducing a dynamic correction coefficient for load nodes Non-loaded node topology coefficients Build final priority The details are as follows: For load nodes, i.e., the baseline load : Dynamic load under typhoon action Based on the reference load After correction for fluctuation coefficient, the following is obtained: (10); In the formula, Let be the typhoon distance attenuation function. The load fluctuation type coefficient is a dimensionless adjustment factor used to characterize the differences in sensitivity or vulnerability of different types of power loads to typhoon weather disturbances. Can be used to construct nodes i Load dynamic correction factor ; Using the dynamic load obtained from equation (10) as input, the voltage at node i Transmission power of line ji Based on the dynamic load of the line The power flow calculation of a radial distribution network is obtained by using the forward-backward substitution method. For any line... ,in, As the parent node, If it is a child node, then: (11); (12); in, It represents the active power flowing from parent node j to child node i under the influence of a typhoon. It is the influence of the typhoon from the parent node j Flow to child nodes i reactive power, c parent node j child node index, Let j be the set of all child nodes of parent node j. parent node j Flow to child nodes c active power, parent node j Flow to child nodes c The reactive power, where c is the index of the child node of parent node j. , For the line impedance, For the line Reactive power, For the line Active power The voltage of parent node j, (13); Repeat the calculation in the order of priority. and Until the voltage change at all nodes satisfies: (14); in The number of iterations is the output after convergence. and To calculate the voltage deviation correction factor and power margin correction factor The final electrical quantity, This is the convergence threshold; The correction factor is obtained by simulating 100 typhoon attacks in Monte Carlo and combining them with the BATTS typhoon model. Power is calculated back from the end of the network to the root node, and voltage is calculated forward from the root node. The process is iterated until convergence, thereby obtaining the electrical state of the system under typhoon disturbance. To quantify the degree of disturbance that typhoons cause to the electrical operating status of the system, a voltage deviation correction factor is defined. With power margin correction factor : (15); (16); in, Rated voltage, The rated capacity of the line. and As voltage deviation correction factor and power margin correction factor, respectively characterize the degree of node voltage exceeding the limit and line load rate, the larger the value, the worse the electrical operating condition. To incorporate both electrical operating conditions and load fluctuations into priority dynamic adjustment, a load dynamic correction coefficient is constructed. : (17); In the formula, where This is a typhoon intensity adjustment factor, with values ​​determined based on the typhoon category and geographical location. Node type weights are assigned to differentiate between emergency, industrial, and residential loads, and the impact of distributed power source integration is considered, using a geometric mean. The correction factor comprehensively reflects the dual risks of voltage and power exceeding limits; the larger the correction factor, the greater the increase in node priority. The nodal load under typhoon conditions is determined by the baseline load. The correction factor was obtained after correction using the typhoon attenuation model. and Then by and Seek; To characterize the real-time load impact of typhoon dynamic disturbances on load nodes, a dynamic correction coefficient is introduced. and typhoon load nodesi Comprehensive risks The overall risk is multiplied by the correction factor to obtain the overall priority of the load nodes. : (18); Equation (18) embodies the combined effect of static vulnerability and real-time operational risk in a product form, thereby achieving dynamic correction of node importance; For non-load nodes, i.e., baseline load Load correction is not applicable. Its priority is primarily based on three factors: physical risk, information risk, and topological importance, supplemented by social risk for comprehensive assessment. Physical risk reflects the likelihood of physical damage to equipment; information risk reflects the degree of communication and control failure; topological importance reflects the criticality of a node in the power grid structure; and social risk reflects the indirect impact of interconnection node failure on downstream user power supply. Since non-load nodes do not directly supply power to users, non-load nodes under typhoon conditions... i Overall Priority Physical risks under typhoon Information risks during typhoons Social risks under typhoons Superimposed topological importance constitute: (19); in, The weighting coefficients for the three-dimensional risk of non-load nodes and satisfying ,and Indicate the topological importance of non-loaded node i: (20); in For node degree weight, The weights are the distances from the root node, and they satisfy... , For nodes The degree, that is, the number of lines directly connected to it. This represents the maximum degree of all nodes in the network. Characterizes the relative density of the connecting lines within a node itself. Let be the electrical distance from node i to the root node. For its network maximum value, It reflects the relative distance from the node to the root node of the distribution network. Adjustment coefficient for node function type; Ultimately, through The function will Limited to interval, This is the lower bound threshold for topological importance.

[0036] As a preferred technical solution of the present invention: In step S2 Based on improvements The objective function and constraints of the path optimization model using the constraint method are expressed as follows: Objective function setting for the path optimization model: Parameter: Let For the set of available drone IDs, For the set of all nodes to be inspected, The set of candidate take-off and landing points is generated and updated in real time by the dynamic take-off and landing point joint optimization stage in step S3 based on the typhoon location, node priority and safety constraints. k is the UAV index, representing the kth UAV, and d is the index of the candidate take-off and landing point, representing the dth candidate point. Decision variables: Indicates drone Is it enabled? Indicates drone Does the flight originate from node i and proceed to node j? Indicates whether node i is controlled by a drone. access; Indicates drone From candidate take-off and landing points take off; Indicates drone From candidate take-off and landing points landing; Objective: Minimize the number of drones. (twenty one); Application of the constraint method: The objective of minimizing the maximum inspection time is transformed into a constraint condition, setting a maximum allowable single-machine inspection time. As a parameter By By scanning with variable parameters and repeatedly solving the above single-objective path optimization model, the desired result can be obtained. The frontier solution set is used to quantify the trade-off between the number of drones and inspection time; Constraints of the path optimization model: Constraint 1: The time constraint requires that the total mission time of each drone must not exceed the maximum allowable single-drone inspection time. ,in, Let be the distance from node i to node j. For drones from nodes i To the node j Inspection time, To ensure the drone moves at a constant speed, : (twenty two); Constraint 2: The node full coverage constraint requires that each node must be visited by one and only one drone once. (twenty three); The takeoff and landing point selection constraint ensures that each activated drone must select one departure point and exactly one landing point: (twenty four); (25); Constraint 3: Segment flow balance constraint requirements: Starting point: If drone From candidate points Takeoff requires executing a procedure from... Flight segment departing for the node to be inspected: (26); Landing point inflow: If drone From candidate points For landing, a process of arriving at the node to be inspected must be executed. Flight segment: (27); Outflow from intermediate nodes: For each node to be inspected If it is attacked by drones The visit, then, happens to be from The outbound segment; otherwise, there is no outflow. (28); Inflow from intermediate nodes: For each node to be inspected If it is attacked by drones The visit, then, happens to be from Inflow segments, otherwise no inflow: (29); Constraint 4: Formula for UAV area boundary constraints, assuming the node to be inspected. The geographic coordinates are ( ): (30); (31); in , and , These are the horizontal and vertical coordinate boundaries of the distribution network inspection area, which are preset by the geographical scope of the inspection task.

[0037] As a preferred technical solution of the present invention: In step S2 The two-stage hybrid solution strategy is as follows: Phase 1: Candidate Path Generation Clustering is performed based on node geographic location and priority, with high-priority nodes as seeds; within each cluster, an initial path is constructed using a priority-based nearest neighbor insertion method, and the 2-opt algorithm is applied for local optimization; timed-out paths are split and a candidate path pool is output. Phase Two: Path Selection and Optimization Construct an integer linear programming model for the set covering problem, incorporating priority weights into the objective function. Prioritize covering key nodes; if the integer linear programming solution fails, backtrack to a greedy algorithm that prioritizes covering unassigned high-priority nodes to ensure a feasible solution is obtained.

[0038] As a preferred technical solution of the present invention: In step S3 The scoring function for dynamic selection of starting point is: (32); in, For the set of candidate take-off and landing points, and For the weighting coefficients, satisfying , This represents the normalized minimum distance from the candidate point to the typhoon's trajectory. Let represent the specific coordinates of the d-th candidate starting point, traj be the discrete point set of the typhoon trajectory, and let p be any point on the typhoon trajectory traj. Indicates the d-th candidate starting point The shortest Euclidean distance to all points on the typhoon's trajectory is calculated. The denominator uses `maxmin` for dynamic normalization, ensuring the safety metric is between [0,1]. For each candidate point... , Let T be the normalized weighted average distance from the candidate starting point to all high-priority nodes, where T is the set of nodes to be inspected. This indicates that all nodes to be inspected will be traversed, with i as the node index. Represents the coordinates of the node. The distance is the Euclidean distance between the candidate starting point and each node to be inspected. The priority weight of the node is set according to the node level. The level node is 3. The level node is 2. The level node is set to 1; As a safety constraint, candidate sites must be located outside the typhoon-affected area. (33); in This is the current radius of the typhoon's influence. For safety margin.

[0039] As a preferred technical solution of the present invention: In step S3 Landing points are intelligently merged as follows: First, collect all planned drone endpoints and set the endpoint set. ,in For the total number of drones, each endpoint Including coordinates and the planned arrival time of the drone ; Then based on the spatial proximity threshold Clustering of endpoints: If the Euclidean distance between two points is less than 1 / 2... Then it is considered mergeable, and thus Divide into several non-overlapping groups Each group End points within a group can be considered for merging. The size is ; Finally, define the binary decision variables. Indicate whether to group All endpoints within the range are merged to its geometric center. : (34); If merge ( If so, then the group contributes 1 landing point. If not merged ( ), then contribute Given 1 initial endpoint, the optimization objective is to minimize the total number of final landing points used: (35); The time it takes for the typhoon to reach the merge point must be later than the longest time it takes for all drones to fly from their original terminal points to the merge point, i.e., for each merge group. ,like Then, the spatiotemporal safety constraint must be met: the time required for the latest arriving drone to reach the merging point from its endpoint should be less than or equal to the time when the typhoon is expected to reach the merging point. (36); in The point of arrival of the typhoon center The estimated time, For the drone's cruising speed, For the drone to fly from endpoint e to the merging point The time required.

[0040] A multi-UAV collaborative inspection system for power distribution networks based on improved epsilon-constraint optimization under typhoon conditions, including Data acquisition and input module: used to acquire typhoon forecast data, power distribution network model data, equipment and social attribute data; Dynamic priority evaluation module: used to realize multi-dimensional risk fusion calculation and dynamic correction, and output node priority; Collaborative trajectory planning module: used to implement improved - Path optimization and solution using constraint methods; Dynamic take-off and landing point optimization module: used to realize the dynamic selection of the departure point and the intelligent merging of the landing point; Solution Output and Scheduling Module: This module integrates path and take-off / landing point information, generates executable collaborative inspection and scheduling instructions, and supports visual display.

[0041] Compared with the prior art, the beneficial effects of the present invention are as follows:

[0042] This invention addresses the three core challenges of multi-drone collaborative inspection of power distribution networks during typhoon disasters, proposing a systematic solution. First, by constructing a multi-dimensional dynamic risk assessment model integrating information, physical, and social dimensions, it achieves accurate and real-time identification of key nodes, solving the problem of inaccurate resource allocation. Second, through improvements... This strategy, combining constraint methods with mixed-integer programming heuristics, efficiently solves the collaborative path planning problem under resource constraints. It significantly reduces UAV usage while ensuring coverage of high-priority nodes, addressing the issue of inefficient resource utilization. Finally, by optimizing take-off and landing points dynamically incorporating typhoon paths, it ensures the safety and spatial efficiency of the operational base, resolving the problem of inflexible operations. Simulation experiments demonstrate that this invention significantly outperforms traditional static strategies in terms of node identification accuracy, UAV resource conservation, improved inspection efficiency, and system security, possessing strong engineering application value and promising prospects for widespread adoption. Attached Figure Description

[0043] Figure 1 The diagram shows the flowchart of the multi-UAV collaborative inspection method for power distribution networks based on improved epsilon-constraint optimization under the influence of typhoons in this invention.

[0044] Figure 2 The image shown is a scenario of multi-drone collaborative inspection of a power distribution network under the influence of a typhoon, according to the present invention.

[0045] Figure 3The diagram shows the node dynamic priority evaluation and electrical quantity correction factor calculation process of this invention.

[0046] Figure 4 The figure shown is based on the improvement of the present invention. -Flowchart of the first-stage trajectory planning solution strategy using the constraint method;

[0047] Figure 5 The figure shown is based on the improvement of the present invention. -Flowchart of the second-stage trajectory planning solution strategy for the constraint method;

[0048] Figure 6 The diagram shown is a flowchart of the dynamic take-off and landing point selection and merging optimization decision-making process of this invention.

[0049] Figure 7 These are experimental simulation diagrams of embodiments of the present invention;

[0050] Figure 8 The diagram shown is a performance comparison of three priority evaluation strategies: dynamic, static, and random, in an embodiment of the present invention.

[0051] Figure 9 The diagram shows the improved hybrid algorithm and the traditional greedy algorithm in different aspects of this invention. constraint values ​​( A comparison curve of the maximum number of drones required for inspection under ( );

[0052] Figure 10 The diagram shows the improved hybrid algorithm and the traditional greedy algorithm in different aspects of this invention. A comparison curve of the maximum time spent on drone inspections;

[0053] Figure 11 The diagram shown is a comparison of the dynamic take-off and landing point strategy and the fixed take-off and landing point strategy in an embodiment of the present invention. Detailed Implementation

[0054] The present invention will be further illustrated below with reference to the accompanying drawings and specific embodiments. It should be understood that the following specific embodiments are for illustrative purposes only and are not intended to limit the scope of the invention.

[0055] like Figure 1 As shown, the multi-UAV collaborative inspection method for power distribution networks based on improved epsilon-constraint optimization proposed in this invention under typhoon influence includes the following steps:

[0056] S1, Node Dynamic Priority Evaluation Phase:

[0057] Input typhoon wind field data, distribution network topology and operating parameters, equipment ledger and social attribute data, construct a comprehensive risk assessment model that integrates physical risk, information risk and social risk, and introduce an electrical quantity correction factor based on forward and backward power flow calculation for dynamic correction, output the dynamic priority score and level classification of all nodes.

[0058] Specifically as follows:

[0059] 1. Construct a comprehensive risk assessment model that integrates physical risk, information risk, and social risk, as detailed below:

[0060] Comprehensive Risk of Typhoon Nodes Due to physical risks Information risk Social risks It consists of three linearly superimposed dimensions, following the basic framework of "risk R = failure probability Pr × failure loss S":

[0061] (1);

[0062] (2);

[0063] in, The weighting coefficients for the three-dimensional risk of load nodes and satisfying ;

[0064] During a typhoon, user electricity consumption behavior changes significantly due to warnings, production shutdowns, or emergency demands. To characterize the differentiated responses of different types of loads to typhoon disturbances, nodes are first defined. Typhoon distance attenuation function from the typhoon center:

[0065] (3);

[0066] in, For nodes Distance to the center of the typhoon It is the attenuation constant;

[0067] The risk calculations for each dimension are as follows:

[0068] Physical risks under normal operating conditions Commonly used to measure the structural vulnerability and environmental threats of equipment, and the probability of physical failure under normal operating conditions. By comprehensively considering factors such as equipment aging (e.g., service life), current load level (normalized value), environmental risks (distance from the typhoon center and wind speed), and historical failure rates, it is found that during a typhoon, the probability of failure is amplified due to environmental degradation, thus yielding the probability of physical failure under typhoon conditions. :

[0069] (4);

[0070] in This is the physical failure probability amplification factor. This is the typhoon distance attenuation function;

[0071] Physical failure losses under normal operating conditions This includes equipment damage and repair costs, economic losses to users due to power outages, and maintenance operation costs. The increased difficulty of emergency repairs under typhoon conditions amplifies these losses, resulting in physical damage losses due to typhoons. :

[0072] (5);

[0073] in This is the amplification factor for physical failure losses. This is the typhoon distance attenuation function;

[0074] Information risks under normal operating conditions This reflects the vulnerability of communication systems and information security, specifically the probability of information failure under normal operating conditions. The probability of information failure under typhoon conditions is obtained by assessing information security protection levels, historical information attack frequencies, and the probability of communication outages caused by typhoons. This is taken into account the physical damage to communication facilities caused by strong winds, which increases the risk of outages. :

[0075] (6);

[0076] in This is the information failure probability amplification factor. This is the typhoon distance attenuation function;

[0077] Information failure loss under normal operating conditions This includes increased data leakage costs due to the difficulty of system recovery and amplified cascading effects, monitoring and operational losses caused by control system (such as SCADA) failures, the scope of communication interruptions, and information security repair costs. Under the influence of typhoons, the increased difficulty of repairs leads to amplified losses, resulting in information failure losses during typhoons. :

[0078] (7);

[0079] in This is the amplification factor for information failure loss. This is the typhoon distance attenuation function;

[0080] Social risks under normal working conditions Used to measure the impact of node failure on social operation and public safety, it represents the probability of social failure under normal operating conditions. The social risk during a typhoon is related to the user sensitivity of the load supplied by the node (e.g., hospitals and schools have high weight) and is also affected by the regional emergency response capability. Due to user gathering and emergency demand, the social risk increases sharply during a typhoon, thus yielding the probability of social failure during a typhoon. :

[0081] (8);

[0082] in This is the amplification factor for the probability of social failures. This is the typhoon distance attenuation function;

[0083] Social failure losses under normal operating conditions Due to the risks of critical infrastructure service disruptions, public safety risks, and social order disruptions, when quantified using indicators such as expected load loss, social losses during typhoons are exacerbated by public service disruptions and secondary disasters, thus yielding the social failure losses during typhoons. :

[0084] (9);

[0085] in This is the social failure loss amplification factor. This is the typhoon distance attenuation function.

[0086] The specific quantitative standards for probabilities Pr and losses S in each dimension, as well as the correction coefficients under typhoon conditions, are determined based on the hardening strategies of various equipment in the distribution network and the impact of distributed generation (DG).

[0087] 2. By introducing a dynamic correction coefficient for load nodes Non-loaded node topology coefficients Build final priority The details are as follows:

[0088] For load nodes such as users, i.e., the baseline load :

[0089] Dynamic load under typhoon action Based on the reference load After correction for fluctuation coefficient, the following is obtained:

[0090] (10);

[0091] In the formula, Let be the typhoon distance attenuation function. The load fluctuation type coefficient is a dimensionless adjustment factor used to characterize the differences in sensitivity or vulnerability of different types of power loads to typhoon weather disturbances.

[0092] Using the dynamic load obtained from equation (10) as input, the voltage at node i Transmission power of line ji Based on the dynamic load of the line The power flow calculation of a radial distribution network is obtained by using the forward-backward substitution method. For any line... ,in, As the parent node, If it is a child node, then:

[0093] (11);

[0094] (12);

[0095] in, It represents the active power flowing from parent node j to child node i under the influence of a typhoon. It is the influence of the typhoon from the parent node j Flow to child nodes i reactive power, c parent node j child node index, Let j be the set of all child nodes of parent node j. parent node j Flow to child nodes c active power, parent node j Flow to child nodes c The reactive power, where c is the index of the child node of parent node j. , For the line impedance, For the line Reactive power, For the line Active power Let be the voltage of the parent node j.

[0096] (13);

[0097] Repeat the calculation in the order of priority. and Until the voltage change at all nodes satisfies:

[0098] (14);

[0099] in The number of iterations is the output after convergence. and To calculate the correction factor and The final electrical quantity, This is the convergence threshold;

[0100] The correction factor is calculated by simulating 100 typhoon attacks using Monte Carlo simulation and combining it with the BATTS typhoon model. Power is calculated back from the network endpoint to the root node, and then voltage is calculated forward from the root node, iterating until convergence. This allows us to obtain the system's electrical status under typhoon disturbances, providing dynamic and quantitative operational risk inputs for node priority assessment;

[0101] To quantify the degree of disturbance that typhoons cause to the electrical operating status of the system, a voltage deviation correction factor is defined. With power margin correction factor :

[0102] (15);

[0103] (16);

[0104] in, Rated voltage, The rated capacity of the line. and These represent the degree of node voltage exceeding limits and the line load rate, respectively. The larger the value, the worse the electrical operating condition.

[0105] To incorporate both electrical operating conditions and load fluctuations into priority dynamic adjustment, a load dynamic correction coefficient is constructed. :

[0106] (17);

[0107] In the formula, where This is a typhoon intensity adjustment factor, with values ​​determined based on the typhoon category and geographical location. Node type weights are assigned to differentiate between emergency (hospital), industrial, and residential loads, and the impact of DG access is considered (DG weights can be reduced). A geometric mean is used. The correction factor comprehensively reflects the dual risks of voltage and power exceeding limits; the larger the correction factor, the greater the increase in node priority. The nodal load under typhoon conditions is determined by the baseline load. The correction factor was obtained after correction using the typhoon attenuation model. and Then by and Seek;

[0108] To characterize the real-time load impact of typhoon dynamic disturbances on load nodes, a dynamic correction coefficient is introduced. and typhoon load nodes i Comprehensive risks Multiplying the comprehensive risk by the correction factor, the final load node is obtained. Overall priority :

[0109] (18);

[0110] Equation (18) embodies the combined effect of "static vulnerability" and "real-time operational risk" in the form of a product, thereby realizing the dynamic correction of node importance;

[0111] For non-load nodes such as switching stations and tie nodes, i.e., the baseline load Load correction is not applicable. Its priority is primarily based on three factors: physical risk, information risk, and topological importance, supplemented by social risk for comprehensive assessment. Physical risk reflects the likelihood of physical damage to equipment; information risk reflects the degree of communication and control failure; topological importance reflects the criticality of a node in the power grid structure; and social risk reflects the indirect impact of interconnection node failure on downstream user power supply. Since non-load nodes do not directly supply power to users, non-load nodes under typhoon conditions... i Overall Priority Physical risks under typhoon Information risks during typhoons Social risks under typhoons Superimposed topological importance constitute:

[0112] (19);

[0113] in, The weighting coefficients for the three-dimensional risk of non-load nodes and satisfying ,and Indicate the topological importance of non-loaded node i:

[0114] (20);

[0115] in For node degree weight, The weights are the distances from the root node, and they satisfy... , For nodes The degree, that is, the number of lines directly connected to it. This represents the maximum degree of all nodes in the network. Characterizes the relative density of the connecting lines within a node itself. It is the electrical distance (sum of per-unit values) from node i to the root node (substation bus). For its network maximum value, It reflects the relative distance from the node to the root node of the distribution network. The node function type correction coefficient is set to 0.3 in this implementation to ensure that non-load nodes have a reasonable priority baseline and to meet the calculation requirements of the subsequent priority evaluation model.

[0116] Ultimately, through The function will Limited to interval, This is the lower bound threshold for topological importance.

[0117] S2, Multi-UAV Collaborative Flight Path Planning Phase:

[0118] With minimizing the number of drones in use as the core optimization objective, a maximum allowable inspection time per drone is set. for Constraints, constructing based on improvement - The constraint method path optimization model adopts a two-stage hybrid solution strategy: the first stage clusters nodes based on their geographical location and priority to generate a high-quality candidate path set; the second stage uses an integer linear programming model or a backoff greedy algorithm to select paths and optimize coverage, and finally outputs the inspection paths of each UAV.

[0119] Specifically as follows:

[0120] 1. Based on improvements The objective function and constraints of the path optimization model using the constraint method are expressed as follows:

[0121] Objective function setting for the path optimization model:

[0122] Parameter: Let For the set of available drone IDs, For the set of all nodes to be inspected, The set of candidate take-off and landing points is generated and updated in real time by the dynamic take-off and landing point joint optimization stage in step S3 based on the typhoon location, node priority and safety constraints. k is the UAV index, representing the kth UAV, and d is the index of the candidate take-off and landing point, representing the dth candidate point.

[0123] Decision variables: Indicates drone Is it enabled? Indicates drone Does the flight originate from node i and proceed to node j? Indicates whether node i is controlled by a drone. access; Indicates drone From candidate take-off and landing points take off; Indicates drone From candidate take-off and landing points landing;

[0124] Objective: Minimize the number of drones.

[0125] (twenty one);

[0126] Application of the constraint method: The objective of "minimizing the maximum inspection time" is transformed into a constraint condition, setting a maximum allowable single-machine inspection time. As a parameter By By scanning with variable parameters and repeatedly solving the above single-objective path optimization model, the desired result can be obtained. The frontier solution set is used to quantify the trade-off between the number of drones and inspection time;

[0127] Constraints of the path optimization model:

[0128] Constraint 1: The time constraint requires that the total mission time (flight and inspection) of each drone must not exceed the maximum allowable single-drone inspection time. ,in, Let be the distance from node i to node j. For drones from nodes i To the node j Inspection time, To ensure the drone moves at a constant speed, :

[0129] (twenty two);

[0130] Constraint 2: The node full coverage constraint requires that each node must be visited by one and only one drone once.

[0131] (twenty three);

[0132] The takeoff and landing point selection constraint ensures that each activated drone must select one departure point and exactly one landing point:

[0133] (twenty four);

[0134] (25);

[0135] Constraint 3: Segment flow balance constraint requirements:

[0136] Starting point: If drone From candidate points Takeoff requires executing a procedure from... Flight segment departing for the node to be inspected:

[0137] (26);

[0138] Landing point inflow: If drone From candidate points For landing, a process of arriving at the node to be inspected must be executed. Flight segment:

[0139] (27);

[0140] Outflow from intermediate nodes: For each node to be inspected If it is attacked by drones The visit, then, happens to be from The outbound flight segment (heading to other nodes awaiting inspection or take-off and landing points) will not be allowed; otherwise, there will be no outflow.

[0141] (28);

[0142] Inflow from intermediate nodes: For each node to be inspected If it is attacked by drones The visit, then, happens to be from Inbound flight segments (from other nodes or takeoff / landing points awaiting inspection); otherwise, no inbound flight segments.

[0143] (29);

[0144] Constraint 4: Formula for UAV area boundary constraints, assuming the node to be inspected. The geographic coordinates are ( ):

[0145] (30);

[0146] (31);

[0147] in , and , These are the horizontal and vertical coordinate boundaries of the distribution network inspection area, respectively. All of these coordinate boundaries are in km and are preset based on the geographical scope of the inspection task.

[0148] At the algorithmic level, priority is also deeply integrated into the entire model process as a soft constraint through a weighting system and dynamic adjustment and optimization strategies: using priority weights To ensure consistent metrics, a guiding role is continuously applied in dynamic starting point selection, node clustering guidance, path construction distance correction, local optimization reward mechanisms, and node preprocessing. This ensures rapid coverage of high-priority nodes without significantly impacting resource and time efficiency, thereby improving the system's responsiveness to critical nodes.

[0149] 2. Based on improvements The constraint-based path optimization model is a variant of the large-scale vehicle routing problem (VRP). Directly solving mixed-integer linear programming (MILP) faces computational difficulties when the number of nodes is large. Therefore, this invention designs a two-stage hybrid solution strategy, specifically as follows:

[0150] Phase 1: Candidate Path Generation

[0151] Clustering is performed based on node geographic location and priority, with high-priority nodes as seeds; within each cluster, an initial path is constructed using a priority-based nearest neighbor insertion method, and the 2-opt algorithm is applied for local optimization; timed-out paths are split and a candidate path pool is output.

[0152] Phase Two: Path Selection and Optimization

[0153] Construct an integer linear programming model for the set covering problem, incorporating priority weights into the objective function. Prioritize covering critical nodes; if the integer linear programming (ILP) solution fails, fall back to a greedy algorithm that prioritizes covering unassigned high-priority nodes to ensure a feasible solution is obtained.

[0154] S3, Dynamic Takeoff and Landing Point Joint Optimization Phase: Based on the typhoon's real-time location, radius of influence, and spatial distribution of high-priority nodes, the starting point with the lowest comprehensive score is dynamically selected from the preset candidate points. After the mission is completed, based on the spatiotemporal accessibility judgment, the landing points adjacent to the terminal are intelligently merged, and a landing point scheme that can effectively improve recovery efficiency while ensuring safety is output. Specifically as follows:

[0155] 1. The scoring function for the dynamic selection strategy of starting point is:

[0156] (32);

[0157] in, For the set of candidate take-off and landing points, and For the weighting coefficients, satisfying , This represents the normalized minimum distance from the candidate point to the typhoon's trajectory. Let represent the specific coordinates of the d-th candidate starting point, traj be the discrete point set of the typhoon trajectory, and let p be any point on the typhoon trajectory traj. Indicates the d-th candidate starting point The shortest Euclidean distance to all points on the typhoon's trajectory is calculated. The denominator uses `maxmin` for dynamic normalization, ensuring the safety metric is between [0,1]. For each candidate point... , Let T be the normalized weighted average distance from the candidate starting point to all high-priority nodes, where T is the set of nodes to be inspected. This indicates that all nodes to be inspected will be traversed, with i as the node index. Represents the coordinates of the node. The distance is the Euclidean distance between the candidate starting point and each node to be inspected. The priority weight of the node is set according to the node level. The level node is 3. The level node is 2. The level node is set to 1;

[0158] Candidate sites must be located outside the typhoon-affected area as a safety constraint.

[0159] (33);

[0160] in This is the current radius of the typhoon's influence. For safety margin.

[0161] 2. The strategy for intelligent merging of landing points is as follows:

[0162] First, collect all planned drone endpoints and set the endpoint set. ,in For the total number of drones, each endpoint Including coordinates and the planned arrival time of the drone ;

[0163] Then based on the spatial proximity threshold Clustering of endpoints: If the Euclidean distance between two points is less than 1 / 2... Then it is considered mergeable, and thus Divide into several non-overlapping groups Each group End points within a group can be considered for merging. The size is ;

[0164] Finally, define the binary decision variables. Indicate whether to group All endpoints within the range are merged to its geometric center. :

[0165] (34);

[0166] If merge ( If so, then the group contributes 1 landing point. If not merged ( ), then contribute Given 1 initial endpoint, the optimization objective is to minimize the total number of final landing points used:

[0167] (35);

[0168] The time it takes for the typhoon to reach the merge point must be later than the longest time it takes for all drones to fly from their original terminal points to the merge point, i.e., for each merge group. ,like Then, the spatiotemporal safety constraint must be met: the time required for the latest arriving drone to reach the merging point from its endpoint should be less than or equal to the time when the typhoon is expected to reach the merging point.

[0169] (36);

[0170] in The point of arrival of the typhoon center The estimated time, For the drone's cruising speed, For the drone to fly from endpoint e to the merging point The time required.

[0171] This strategy improves operational efficiency while strictly ensuring the safety of data recovery.

[0172] like Figure 2-6 As shown, this paper details the implementation process of a multi-UAV collaborative inspection method for power distribution networks based on improved epsilon-constraint optimization under typhoon influence, and verifies the method in a simulation environment based on the IEEE 33-node power distribution system and the BATTS typhoon model.

[0173] Specifically, the following steps are included:

[0174] Step S1: Dynamic priority evaluation of nodes, specifically:

[0175] 1. Input data preparation: Load the 24-hour typhoon wind field trajectory data (including center location, moving speed, and radius of influence) generated by the BATTS typhoon model; import the topology, line parameters, and baseline load data of the IEEE 33-node system; set the equipment attributes (type, age, historical failure rate) and social attributes (user type: residential, industrial, hospital, etc.) of each node.

[0176] 2. Multidimensional comprehensive risk calculation: For each node Calculate its comprehensive risk value under the typhoon using the following formula. :

[0177] First, it is defined as the typhoon distance decay function. ,in For nodes Distance to the center of the typhoon The attenuation constant is:

[0178] (3);

[0179] ;

[0180] in, The weighting coefficients for the three-dimensional risk of load nodes and satisfying .

[0181] 1) Probability of physical failures during typhoons :

[0182] ;

[0183] in Let be the typhoon distance attenuation function. This is the physical failure probability amplification factor. This represents the probability of physical failure under normal operating conditions.

[0184] 2) Physical failure losses during typhoons :

[0185] ;

[0186] in Let be the typhoon distance attenuation function. This is the amplification factor for physical failure losses. Losses due to physical failures under normal operating conditions.

[0187] 3) Probability of information failure during typhoons :

[0188] ;

[0189] in Let be the typhoon distance attenuation function. This is the information failure probability amplification factor. This represents the probability of information failure under normal operating conditions.

[0190] 4) Information loss during typhoons :

[0191] ;

[0192] in Let be the typhoon distance attenuation function. This is the amplification factor for information failure loss. This refers to information loss due to malfunctions under normal operating conditions.

[0193] 5) Probability of social disruptions during typhoons :

[0194] ;

[0195] in Let be the typhoon distance attenuation function. This is the amplification factor for the probability of social failures. This represents the probability of social failure under normal operating conditions.

[0196] 6) Social disruption losses during typhoons :

[0197] ;

[0198] in Let be the typhoon distance attenuation function. This is the social failure loss amplification factor. This refers to social failure losses under normal operating conditions.

[0199] 3. Dynamic correction of electrical quantities:

[0200] 1) Using the forward-backward substitution method, power flow calculations are performed on the distribution network under each Monte Carlo simulation typhoon scenario. Power is calculated back from the network end to the root node, and voltage is calculated forward from the root node. This calculation is repeated in sequence. and Until the voltage change at all nodes meets the requirements , The convergence threshold is denoted as , where The iteration number is the number of iterations until convergence. A total of 100 simulations were performed (each simulation was required to converge) to obtain the voltage at each node. and the transmission power of each line .

[0201] 2) Calculate the voltage deviation correction factor and power margin correction factor ,in Rated voltage, This refers to the power transmitted through the line.

[0202] 3) Calculate the dynamic correction coefficient for load nodes. ,in This is a typhoon intensity adjustment factor, with values ​​determined based on the typhoon level (super typhoon, strong typhoon, etc.) and geographical location (coastal / inland). Node type weights are assigned to differentiate between emergency (hospital) and industrial / residential loads, and the impact of DG (distributed generation) access is considered (DG weights can be reduced). Dynamic load under typhoon conditions. Based on the reference load Through load fluctuation coefficient The correction was obtained.

[0203] 4. Priority composition and classification:

[0204] 1) For load nodes (baseline load) ): ;

[0205] 2) For non-load nodes ( ): ,in, This indicates that the weight is based on the node degree. Distance weight from root node (Both conditions are met) The topological importance of computation, For nodes The degree, that is, the number of lines directly connected to it. This represents the maximum degree of all nodes in the network. It represents the relative density of the connecting lines of a node itself. For nodes Electrical distance to the root node (substation busbar) (sum of per-unit values). For its network maximum value, It reflects the relative distance from the node to the root node of the distribution network. This is a correction coefficient for the node's functional type. Wherein, The weighting coefficients for the three-dimensional risk of non-load nodes and satisfying .

[0206] 3) All nodes After normalization, the values ​​are divided into three levels based on quantiles: (High-level, top 20%) (Medium grade, middle 40%) (Low grade, last 40%).

[0207] Step S2: Multi-UAV Collaborative Flight Path Planning

[0208] Parameter settings: Set the drone's cruise speed Maximum permissible single-machine inspection time constraint for Interval is of each sequence point. The priority weight of the node is set according to the node level. The number of level nodes is 3. The number of level nodes is 2. The level node is 1.

[0209] Phase 1: Generating a candidate path pool

[0210] 1) Clustering: using all Nodes are used as initial cluster seeds. Combined with the K-means algorithm, clusters are formed based on node geographic coordinates and priority weights. Clusters.

[0211] 2) Intra-cluster path construction and optimization: For each cluster, starting from the dynamically selected starting point (see step 3), a Hamiltonian cycle covering all nodes in the cluster is constructed using the nearest neighbor insertion method (prioritizing the insertion of high-priority nodes). Subsequently, the 2-opt algorithm is immediately applied to the cycle for local optimization to reduce the total path length.

[0212] 3) Constraint checking and path splitting: Calculate the total time (flight + inspection) of the optimized path. If it exceeds... Then, the maximum gap method is used to split the path into two paths that satisfy... Constrained sub-paths. All generated paths (including split sub-paths) are added to the "candidate path pool".

[0213] Phase Two: Path Selection and Coverage Optimization

[0214] 1) Establish the ILP model: Construct a set coverage model with the goal of minimizing the total number of selected paths and penalizing uncovered high-priority nodes.

[0215] Parameter: Let For the candidate path set, For the set of all nodes to be inspected, This is a set of candidate take-off and landing points, which is generated and updated in real time by the dynamic take-off and landing point optimization module in step three based on the typhoon location, node priority, and safety constraints.

[0216] ;

[0217] in Indicates whether to select a path , Represents a node Is it by path cover, The priority weight of the node. This is the penalty coefficient. The model minimizes the total number of paths (i.e., the number of drones) while imposing a larger penalty on uncovered high-priority nodes, thus prioritizing coverage of important nodes.

[0218] 2) Solving and Backoff: Use an ILP solver (such as Gurobi) to solve the problem within a time limit (e.g., 120 seconds). If the optimal solution is found, output it directly. If the timeout occurs or the solution is not feasible, start the backoff greedy algorithm. The algorithm process is as follows:

[0219] a. Initialize the set of uncovered nodes .

[0220] b. Loop until... Empty: From the candidate path pool In the middle, select one that can cover The maximum sum of node weights ( (Maximum) path Add it to the solution.

[0221] c. Never covered node set Remove from All nodes covered.

[0222] d. Update the candidate path pool and remove paths that conflict with or are inefficient compared to the selected paths.

[0223] Step S3: Dynamic take-off and landing point joint optimization strategy;

[0224] 1. Dynamic selection strategy for starting point:

[0225] 1) Obtain the current location of the typhoon center. and the radius of influence of typhoons Pre-set a set of geographically dispersed safety candidate points :satisfy ,in This is the current radius of the typhoon's influence. For safety margin, This is the set of candidate take-off and landing points.

[0226] 2) Identify all If the set of level nodes is not entirely empty and all nodes are located outside the danger zone, then each candidate starting point... Calculate the safe distance metric: Calculate the weighted efficiency distance metric: Normalization and to The interval, in which For nodes Priority weights, This is the set of candidate take-off and landing points. Let d represent the specific coordinates of the d-th candidate starting point, and traj be a discrete set of points on the typhoon trajectory (generated by the BATTS typhoon model, each point containing the coordinates of the typhoon center). Let p be any point on the typhoon trajectory traj, then Indicates the d-th candidate starting point The shortest Euclidean distance to all points on the typhoon's trajectory is used. The denominator uses `maxmin` for dynamic normalization, ensuring the safety metric is within the range [0,1]. Here, `T` represents the set of nodes to be inspected. This indicates that all nodes to be inspected will be traversed, where i serves as both the node index and the coordinates of that node. This represents the Euclidean distance between the candidate starting point and each node to be inspected. This strategy ensures that, under safe conditions, the starting point can dynamically adapt to the spatial distribution of high-priority tasks.

[0227] 3) Calculate the overall score: for each candidate starting point , . ( , For the weighting coefficients to satisfy This embodiment adopts =0.7、 This strategy iterates through all feasible candidate points that satisfy the constraints and calculates a comprehensive score for each candidate point. , The highest candidate point is used as the starting point for the current inspection batch.

[0228] 4) If If the set of level nodes is empty or all are within the danger zone, select a location far from the typhoon center. The farthest candidate point is used as the starting point.

[0229] 2. Intelligent landing point merging strategy:

[0230] 1) Obtain the set of endpoints for all UAV inspection path plans (obtained from step 2). k is the total number of drones. Each endpoint Including coordinates and the planned arrival time of the drone Pre-set merging threshold (like ), the typhoon's current location and speed of movement.

[0231] 2) Based on spatial proximity threshold Clustering of endpoints: If the Euclidean distance between two points is less than 1 / 2... Therefore, it is considered mergeable. Divide into several non-overlapping groups Each group End points within a group can be considered for merging. The size is (i.e., the number of endpoints included).

[0232] 3) Identify connected components: Use breadth-first search (BFS) to find all merge groups. For each merge group... Calculate its geometric center point : .

[0233] 4) Judgment Is it within the current or predicted typhoon danger zone? If outside the zone, then directly... The unified landing point for this group is determined; if it is within the area, a spatiotemporal assessment is performed: the typhoon's arrival is linearly extrapolated based on its current position and speed. Time All drones within the computing group start from their respective endpoints. Fly to Time required .

[0234] 5) If the typhoon arrives at the merge point later than the longest time all drones have taken to fly from their original endpoints to the merge point: Then merge into Otherwise, do not merge, including all original endpoints within the group. Each is reserved as an independent landing point.

[0235] 6) Output the final set of landing points.

[0236] This invention also provides a multi-UAV collaborative inspection system for power distribution networks based on improved epsilon-constraint optimization under typhoon conditions, comprising:

[0237] Data acquisition and input module: used to acquire typhoon forecast data, power distribution network model data, equipment and social attribute data;

[0238] Dynamic priority evaluation module: used to execute step S1, realize multi-dimensional risk fusion calculation and dynamic correction, and output node priority;

[0239] Collaborative trajectory planning module: used to execute step S2, achieving improvement based on... - Path optimization and solution using constraint methods;

[0240] Dynamic take-off and landing point optimization module: used to execute step S3, realize the dynamic selection of the departure point and the intelligent merging of the landing point;

[0241] Solution Output and Scheduling Module: This module integrates path and take-off / landing point information, generates executable collaborative inspection and scheduling instructions, and supports visual display.

[0242] This embodiment describes the specific experimental design of a multi-UAV collaborative inspection method for power distribution networks based on improved epsilon-constraint optimization under typhoon conditions:

[0243] according to Figure 7 The scenario shown is simulated and verified. In this embodiment, an IEEE 33-node power distribution system (tree topology), multiple drones, and a typhoon scenario simulation system are included. According to the power distribution network inspection task requirements under typhoon disaster, the control center uses the collaborative optimization method proposed in this invention to dynamically evaluate node priorities, plan drone trajectories, and adjust take-off and landing point positions. The test area simulates the power distribution network area under the influence of a typhoon, including dynamic typhoon trajectories, different types of nodes, and fixed / dynamic take-off and landing point scenarios.

[0244] Parameter settings: The experiment is conducted on an IEEE 33-node distribution system, configured with 33 nodes in a tree topology. The distance between adjacent nodes on the backbone is 8 km, the furthest node (nodes 1 to 18) is 136 km, and the average node distance is 8 km. System reference voltage, reference capacity, total active load, and total reactive load all use standard test system data. The allowable operating range of node voltage is the standard distribution network operating threshold. Social impact weights are allocated according to node type: photovoltaic (PV), residential (PV), and so on. ),industry( ) and emergency response ( The data are categorized into four types, and assigned corresponding importance coefficients. , In this experiment, all values ​​are taken as 1, and the electrical quantity correction factor is... , The base value of the node physical failure probability was obtained by combining forward and backward substitution with Monte Carlo simulation. Based on the equipment's service life and historical failure rate, the equipment's service life follows a uniform distribution of 2–25 years. The historical failure rate is set to the range [0.0005, 0.01], which corresponds to the baseline failure probability, representing the probability of equipment failure per unit time (year) under normal operating conditions. The specific value is determined by linear interpolation within this range based on the node's service life; the longer the service life, the larger the value. The weighting coefficients for the three-dimensional risk of the load node are also considered. Take 0.4, Take 0.3, The weighting coefficient for the three-dimensional risk of non-loaded nodes is set to 0.3. Take 0.4, Take 0.4, The weight is set to 0.2 (because load distribution nodes indirectly affect users, thus reducing the weight of social impact). Typhoon intensity adjustment factor. Take 1.3. Attenuation constant. Take 0.15.

[0245] Initial parameters: The experiment used the Monte Carlo method to generate 100 typhoon scenarios, including 24-hour wind field trajectories. The typhoon path was generated by the BATTS model, and the wind speed influence factor was [not specified]. Calculated based on the wind speed-distance model. Overall risk. The risk is synthesized through a three-dimensional risk model encompassing physical, informational, and social factors, where the physical risk incorporates a wind speed amplification factor. Cost coefficient In node priority calculation, a load correction factor is introduced for load nodes. Topological importance Based on distance calculation between nodes and the system center. To verify the effectiveness of the dynamic priority evaluation model of this invention, a baseline true risk ranking needs to be pre-defined as the evaluation standard. This invention constructs this baseline based on comprehensive offline simulation: that is, considering the dynamic impact of typhoons throughout the entire period, complete power flow calculation, and equipment parameters, the node risk ranking obtained through offline calculation serves as the reference true value for measuring the accuracy of each scheme identification.

[0246] Experimental Scenario: Based on this, three comparison schemes were set up: the random scheme completely ignores system state and risk information, and only performs random sorting as the comparison benchmark; the static scheme calculates the comprehensive risk and fixes the priority based on the load snapshot after the typhoon impact, and although it incorporates disaster information, it does not consider real-time electrical status; the dynamic scheme is the core method of this invention, which further integrates the electrical status correction factor obtained from real-time power flow calculation on the basis of the static scheme, so that the priority can be dynamically updated with the system operating status. All strategies are executed under the same typhoon scenario to ensure comparability. The evaluation index uses the overlap rate between the Top-5 high-risk nodes output by each scheme and the baseline true value to measure the identification accuracy, and at the same time, the calculation time per operation is recorded.

[0247] Combination Figure 7 and Figure 8The data results show that the dynamic scheme achieves an accuracy rate of 90.5%, significantly higher than the static scheme (24.8%) and the random scheme (16.1%), indicating that the model of this invention can more accurately identify the truly important nodes under the influence of typhoons. The error bars show that the dynamic scheme has the smallest standard deviation, indicating its most stable performance under different typhoon scenarios; while the static and random schemes exhibit significantly larger fluctuations and poorer stability. Regarding computational efficiency, the average single-calculation time for all three schemes is within 1ms (random 0.562ms, static 0.551ms, dynamic 0.874ms). Although the dynamic scheme experiences a slight increase due to the introduction of tidal current calculations, the 0.874ms time is perfectly acceptable for typhoon emergency scenarios with minute-level updates, and latency can be further reduced through parallelization and convergence acceleration. Therefore, this invention, while ensuring dynamic accuracy and stability, possesses engineering-acceptable real-time computational capabilities. The advantages of this invention's model lie in its integration of information, physical, and social factors, as well as the introduction of real-time correction of electrical operating status. This not only provides a more comprehensive range of dimensions but also reduces the dependence on data volume, making it more suitable for real-time dynamic assessment during disasters. Furthermore, it emphasizes dynamic response and is applicable to real-time priority adjustment under typhoon path changes.

[0248] Initial parameters: The simulation area, typhoon model path, and power distribution system topology are set consistent with the above parameters, and the take-off and landing points are fixed. The experiment uses fixed take-off and landing points, with the UAV's cruising speed V = 36 km / h, a maximum flight time of 240 minutes, and a single-point inspection time of 5 minutes. Maximum permissible time... Scans are performed at 20-minute intervals within a timeframe of 120 to 300 minutes. Node priority weights. According to its level ( The values ​​are set to 3, 2, and 1 respectively. The ILP solution time limit is set to 120 seconds, and the relative tolerance is [missing value]. That is, when the relative deviation between the current solution and the objective function value of the theoretical optimal solution is less than 0.1%, the solver terminates prematurely.

[0249] Experimental Scenario: Under the typhoon path generated by the BATTS model, the hybrid algorithm proposed in this invention (integrating priority ILP exact solution and priority greedy backoff mechanism) is compared with the traditional greedy algorithm. The traditional greedy algorithm only iteratively constructs the path according to the nearest distance principle, without considering priority weights and global optimization. (Maximum number of drones...) The actual maximum execution time (Makespan) is used as a metric to record and compare the two algorithms under different conditions. Performance under constraints.

[0250] Combination Figure 9 and Figure 10 The experimental results show that, under the same conditions... Under constraints, the minimum number of drones required by the hybrid algorithm of this invention is always less than that of the traditional greedy algorithm. This is especially true when time constraints are tight (such as...). =140 minutes), the average drone usage can be reduced by up to 30%. At the same time, the actual maximum time (Makespan) of the hybrid algorithm scheme is only 1%-3% higher than that of the greedy algorithm, which shows that the hybrid algorithm achieves significant drone resource savings with minimal extension of task time.

[0251] Initial parameters: The simulation area, UAV cruise speed, UAV maximum endurance, typhoon model path, and power distribution system topology are set consistent with the parameters mentioned above. Landing point merging judgment threshold. Set the distance to 12 kilometers. Node priority weight. The levels are still set to 3, 2, and 1 respectively.

[0252] Experimental Scenario: The experimental system compared a fixed site strategy (center fixed at (75,75)) with a dynamic site strategy (center dynamic selection based on real-time typhoon location and node priority weighting). The typhoon's movement path is as follows: Figure 7 As shown, the drone will traverse the area where a fixed airfield is located to test the effectiveness of the two strategies under extreme threats. The average initial flight distance of the drone from its starting point to its first inspection node will be statistically analyzed under both strategies, and the entry and landing points into typhoon danger zones will be monitored.

[0253] Combination Figure 11 The experimental results show that the dynamic take-off and landing point strategy enables... The time required to complete the level node inspection was reduced by 16%. The dynamic strategy also achieved a slight reduction (8%) in total flight time. Throughout the typhoon's evolution, the dynamic strategy successfully ensured that all takeoff and landing points remained outside the typhoon's safe zone, while the fixed strategy experienced a sharp deterioration in safety during the typhoon's passage, leading to unreliable critical responses. Therefore, the dynamic strategy significantly outperforms the fixed strategy in terms of safety, inspection efficiency, and dynamic adaptability—it can adjust takeoff and landing points in real time according to the typhoon's location, avoiding the risk of fixed stations being covered by the typhoon, and improves recovery efficiency through terminal point merging, making it more suitable for emergency inspection tasks in dynamic disaster environments such as typhoons.

[0254] Initial parameters: Based on the IEEE 33-node system, the standard IEEE 69-node and 123-node test systems were further used as extended scenarios. Within the same 150km × 150km area, the spacing between adjacent nodes on the backbone was reduced from 8km to 4.5km and 3.2km, but the average node spacing remained stable at around 4.5-4.6km. Typhoon parameters, UAV performance parameters (speed 36km / h, maximum endurance 240min), and optimization algorithm parameters were kept unchanged. A unified... =240 minutes as the performance comparison benchmark.

[0255] Experimental Scenario: The proposed collaborative distribution network inspection method was executed in three different scale distribution network systems. The total inspection completion time, average number of drones used, proportion of nodes at each level prioritized for inspection, and algorithm solution time were recorded for each scale. To quantify the priority bias effect, a priority-weighted coverage rate was defined. ,in , , They are respectively , , Priority inspection rate of level nodes.

[0256] As shown in Table 1, as the system scale increases from 33 nodes to 123 nodes, the total inspection time increases accordingly, and the required number of drones also increases reasonably. The key point is... The priority inspection rate of nodes remained above 54% in all three types of systems, with priority-weighted coverage. The time taken in the three systems were 2.31, 2.37, and 2.38 respectively (the theoretical maximum is 3), indicating that the priority inspection of high-priority nodes is significantly effective. The algorithm's solution time was 2.31 minutes, 2.37 minutes, and 2.38 minutes in the 33, 69, and 123 node systems, respectively, meeting the real-time requirement of rolling replanning every 60 minutes in emergency scenarios. This invention introduces priority weights for the first time and quantifies the priority inspection rate of nodes at each level. The above statistical indicators intuitively demonstrate the strategy's effectiveness in prioritizing high-priority nodes.

[0257] Table 1

[0258]

[0259] The above-mentioned joint tests show that the method proposed in this invention can maintain priority inspection of high-priority nodes in systems with different node sizes, and also exhibits good scalability and a certain degree of adaptability to different scales.

[0260] Those skilled in the art will understand that embodiments of this application can be provided as methods, systems, or computer program products. Therefore, this application can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, this application can take the form of a computer program product embodied on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0261] This application is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of this application. It will be understood that each block of the flowchart illustrations and / or block diagrams, as well as combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart... Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0262] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0263] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0264] It should be noted that the above description is only a preferred embodiment of the present invention and is not intended to limit the present invention in any other way. Any modifications or equivalent changes made based on the technical essence of the present invention shall still fall within the scope of protection claimed by the present invention.

Claims

1. A method for multi-unmanned aerial vehicle (UAV) cooperative inspection and distribution network based on improved epsilon-constraint optimization under typhoon influence, characterized in that, Includes the following steps: S1, Node Dynamic Priority Evaluation Phase: Input typhoon wind field data, distribution network topology and operating parameters, equipment ledger and social attribute data, construct a comprehensive risk assessment model that integrates physical risk, information risk and social risk, and introduce an electrical quantity correction factor based on forward and backward power flow calculation for dynamic correction, output the dynamic priority score and level classification of all nodes. S2, Multi-UAV Collaborative Flight Path Planning Phase: With the core optimization goal of minimizing the number of unmanned aerial vehicles enabled, set the maximum allowed single machine inspection time For Constraints, build path optimization model based on improved - constraint method, using two-stage hybrid solution strategy: the first stage is based on the clustering of node geographical location and priority and generating high-quality candidate path set; the second stage is through integer linear programming model or backtracking greedy algorithm for path selection and coverage optimization, finally output the inspection path of each unmanned aerial vehicle; S3, Dynamic Takeoff and Landing Point Joint Optimization Phase: Based on the typhoon's real-time location, radius of influence, and spatial distribution of high-priority nodes, the starting point with the highest comprehensive score is dynamically selected from the preset candidate points. After the mission is completed, based on the spatiotemporal accessibility judgment, the landing points adjacent to the terminal are intelligently merged, and a landing point scheme that can effectively improve recovery efficiency while ensuring safety is output.

2. The multi-UAV collaborative inspection method for power distribution networks under typhoon influence based on improved epsilon-constraint optimization as described in claim 1, characterized in that, In step S1: A comprehensive risk assessment model integrating physical risk, information risk, and social risk is constructed, as follows: Comprehensive Risk of Typhoon Nodes Due to physical risks Information risk Social risks It consists of three linearly superimposed dimensions, following the basic framework of "risk R = failure probability Pr × failure loss S": (1)； (2)； in, The weighting coefficients for the three-dimensional risk of load nodes and satisfying ; To characterize the differentiated responses of different types of loads to typhoon disturbances, the nodes are first defined. Typhoon distance attenuation function from the typhoon center: (3)； in, For nodes Distance to the center of the typhoon It is the attenuation constant; The risk calculations for each dimension are as follows: Physical risks under normal operating conditions Used to measure the probability of physical failure under normal operating conditions of equipment structural vulnerability and environmental threats. By comprehensively considering factors such as equipment aging, current load levels, environmental risks, and historical failure rates, it was found that the probability of failure during a typhoon is amplified due to environmental degradation, thus yielding the probability of physical failure under typhoon conditions. : (4)； in This is the physical failure probability amplification factor. This is the typhoon distance attenuation function; Physical failure losses under normal operating conditions This includes equipment damage and repair costs, economic losses to users due to power outages, and maintenance operation costs. The increased difficulty of emergency repairs under typhoon conditions amplifies these losses, resulting in physical damage losses due to typhoons. : (5)； in This is the amplification factor for physical failure losses. This is the typhoon distance attenuation function; Information risks under normal operating conditions This reflects the vulnerability of communication systems and information security, specifically the probability of information failure under normal operating conditions. The probability of information failure under typhoon conditions is obtained by assessing information security protection levels, historical information attack frequencies, and the probability of communication outages caused by typhoons. This is taken into account the physical damage to communication facilities caused by strong winds, which increases the risk of outages. : (6)； in This is the information failure probability amplification factor. This is the typhoon distance attenuation function; Information failure loss under normal operating conditions This includes increased data leakage costs due to the difficulty of system recovery and amplified cascading effects, monitoring and operational losses caused by control system failures, the scope of communication interruptions, and information security repair costs. Under the influence of typhoons, the increased difficulty of repairs leads to amplified losses, resulting in information failure losses during typhoons. : (7)； in This is the amplification factor for information failure loss. This is the typhoon distance attenuation function; Social risks under normal working conditions Used to measure the impact of node failure on social operation and public safety, it represents the probability of social failure under normal operating conditions. The social risk during a typhoon is related to the user sensitivity of the load supplied by the node and is affected by the regional emergency response capability. Due to user gathering and emergency demand, the social risk increases sharply. This leads to the probability of social failure during a typhoon. : (8)； in This is the amplification factor for the probability of social failures. This is the typhoon distance attenuation function; Social failure losses under normal operating conditions Due to the risks of critical infrastructure service disruptions, public safety risks, and social order disruptions, when quantified using the expected load loss index, social losses during typhoons are exacerbated by public service disruptions and secondary disasters, thus yielding the social failure losses during typhoons. : (9)； in This is the social failure loss amplification factor. This is the typhoon distance attenuation function.

3. The multi-UAV collaborative inspection method for power distribution networks under typhoon influence based on improved epsilon-constraint optimization as described in claim 1, characterized in that, In step S1 By introducing a dynamic correction coefficient for load nodes Non-loaded node topology coefficients Build final priority The details are as follows: For load nodes, i.e., the baseline load : Dynamic load under typhoon action Based on the reference load After correction for fluctuation coefficient, the following is obtained: (10)； In the formula, Let be the typhoon distance attenuation function. The load fluctuation type coefficient is a dimensionless adjustment factor used to characterize the differences in sensitivity or vulnerability of different types of power loads to typhoon weather disturbances. Can be used to construct nodes i Load dynamic correction factor ; Using the dynamic load obtained from equation (10) as input, the voltage at node i Transmission power of line ji Based on the dynamic load of the line The power flow calculation of a radial distribution network is obtained by using the forward-backward substitution method. For any line... ,in, As the parent node, If it is a child node, then: (11)； (12)； in, It represents the active power flowing from parent node j to child node i under the influence of a typhoon. It is the influence of the typhoon from the parent node j Flow to child nodes i reactive power, c parent node j child node index, Let j be the set of all child nodes of parent node j. parent node j Flow to child nodes c active power, parent node j Flow to child nodes c The reactive power, where c is the index of the child node of parent node j. , For the line impedance, For the line Reactive power, For the line Active power The voltage of parent node j, (13)； Repeat the calculation in the order of priority. and Until the voltage change at all nodes satisfies: (14)； in The number of iterations is the output after convergence. and To calculate the voltage deviation correction factor and power margin correction factor The final electrical quantity, This is the convergence threshold; The correction factor is obtained by simulating 100 typhoon attacks in Monte Carlo and combining them with the BATTS typhoon model. Power is calculated back from the end of the network to the root node, and voltage is calculated forward from the root node. The process is iterated until convergence, thereby obtaining the electrical state of the system under typhoon disturbance. To quantify the degree of disturbance that typhoons cause to the electrical operating status of the system, a voltage deviation correction factor is defined. With power margin correction factor : (15)； (16)； in, Rated voltage, The rated capacity of the line. and As voltage deviation correction factor and power margin correction factor, respectively characterize the degree of node voltage exceeding the limit and line load rate, the larger the value, the worse the electrical operating condition. To incorporate both electrical operating conditions and load fluctuations into priority dynamic adjustment, a load dynamic correction coefficient is constructed. : (17)； In the formula, where This is a typhoon intensity adjustment factor, with values ​​determined based on the typhoon category and geographical location. Node type weights are assigned to differentiate between emergency, industrial, and residential loads, and the impact of distributed power source integration is considered, using a geometric mean. The correction factor comprehensively reflects the dual risks of voltage and power exceeding limits; the larger the correction factor, the greater the increase in node priority. The nodal load under typhoon conditions is determined by the baseline load. The correction factor was obtained after correction using the typhoon attenuation model. and Then by and Seek; To characterize the real-time load impact of typhoon dynamic disturbances on load nodes, a dynamic correction coefficient is introduced. and typhoon load nodes i Comprehensive risks The overall risk is multiplied by the correction factor to obtain the overall priority of the load nodes. : (18)； Equation (18) embodies the combined effect of static vulnerability and real-time operational risk in a product form, thereby achieving dynamic correction of node importance; For non-load nodes, i.e., baseline load Load correction is not applicable. Its priority is primarily based on three factors: physical risk, information risk, and topological importance, supplemented by social risk for comprehensive assessment. Physical risk reflects the likelihood of physical damage to equipment; information risk reflects the degree of communication and control failure; topological importance reflects the criticality of a node in the power grid structure; and social risk reflects the indirect impact of interconnection node failure on downstream user power supply. Since non-load nodes do not directly supply power to users, non-load nodes under typhoon conditions... i Overall Priority Physical risks under typhoon Information risks during typhoons Social risks under typhoons Superimposed topological importance constitute: (19)； in, The weighting coefficients for the three-dimensional risk of non-load nodes and satisfying ,and Indicate the topological importance of non-loaded node i: (20)； in For node degree weight, The weights are the distances from the root node, and they satisfy... , For nodes The degree, that is, the number of lines directly connected to it. This represents the maximum degree of all nodes in the network. Characterizes the relative density of the connecting lines within a node itself. Let be the electrical distance from node i to the root node. For its network maximum value, It reflects the relative distance from the node to the root node of the distribution network. Adjustment coefficient for node function type; Ultimately, through The function will Limited to interval, This is the lower bound threshold for topological importance.

4. The multi-UAV collaborative inspection method for power distribution networks under typhoon influence based on improved epsilon-constraint optimization as described in claim 1, characterized in that, In step S2 Based on improvements The objective function and constraints of the path optimization model using the constraint method are expressed as follows: Objective function setting for the path optimization model: Parameter: Let For the set of available drone IDs, For the set of all nodes to be inspected, The set of candidate take-off and landing points is generated and updated in real time by the dynamic take-off and landing point joint optimization stage in step S3 based on the typhoon location, node priority and safety constraints. k is the UAV index, representing the kth UAV, and d is the index of the candidate take-off and landing point, representing the dth candidate point. Decision variables: Indicates drone Is it enabled? Indicates drone Does the flight originate from node i and proceed to node j? Indicates whether node i is controlled by a drone. access; Indicates drone From candidate take-off and landing points take off; Indicates drone From candidate take-off and landing points landing; Objective: Minimize the number of drones. (21)； Application of the constraint method: The objective of minimizing the maximum inspection time is transformed into a constraint condition, setting a maximum allowable single-machine inspection task time. As a parameter By By scanning with variable parameters and repeatedly solving the above single-objective path optimization model, the desired result can be obtained. The frontier solution set is used to quantify the trade-off between the number of drones and inspection time; Constraints of the path optimization model: Constraint 1: The time constraint requires that the total mission time of each drone must not exceed the maximum allowable single-drone inspection time. ,in, Let be the distance from node i to node j. For drones from nodes i To the node j Inspection time, To ensure the drone moves at a constant speed, : (22)； Constraint 2: The node full coverage constraint requires that each node must be visited by one and only one drone once. (23)； The takeoff and landing point selection constraint ensures that each activated drone must select one departure point and exactly one landing point: (24)； (25)； Constraint 3: Segment flow balance constraint requirements: Starting point: If drone From candidate points Takeoff requires executing a procedure from... Flight segment departing for the node to be inspected: (26)； Landing point inflow: If drone From candidate points For landing, a process of arriving at the node to be inspected must be executed. Flight segment: (27)； Outflow from intermediate nodes: For each node to be inspected If it is attacked by drones The visit, then, happens to be from The outbound segment; otherwise, there is no outflow. (28)； Inflow from intermediate nodes: For each node to be inspected If it is attacked by drones The visit, then, happens to be from Inflow segments, otherwise no inflow: (29)； Constraint 4: Formula for UAV area boundary constraints, assuming the node to be inspected. The geographic coordinates are ( ): (30)； (31)； in , and , These are the horizontal and vertical coordinate boundaries of the distribution network inspection area, which are preset by the geographical scope of the inspection task.

5. The multi-UAV collaborative inspection method for power distribution networks under typhoon influence based on improved epsilon-constraint optimization as described in claim 4, characterized in that, In step S2 The two-stage hybrid solution strategy is as follows: Phase 1: Candidate Path Generation Clustering is performed based on node geographic location and priority, with high-priority nodes as seeds; within each cluster, an initial path is constructed using a priority-based nearest neighbor insertion method, and the 2-opt algorithm is applied for local optimization; timed-out paths are split and a candidate path pool is output. Phase Two: Path Selection and Optimization Construct an integer linear programming model for the set covering problem, incorporating priority weights into the objective function. Prioritize covering key nodes; if the integer linear programming solution fails, backtrack to a greedy algorithm that prioritizes covering unassigned high-priority nodes to ensure a feasible solution is obtained.

6. The multi-UAV collaborative inspection method for power distribution networks under typhoon influence based on improved epsilon-constraint optimization as described in claim 1, characterized in that, In step S3 The scoring function for dynamic selection of starting point is: (32)； in, For the set of candidate take-off and landing points, and For the weighting coefficients, satisfying , This represents the normalized minimum distance from the candidate point to the typhoon's trajectory. Let represent the specific coordinates of the d-th candidate starting point, traj be the discrete point set of the typhoon trajectory, and let p be any point on the typhoon trajectory traj. Indicates the d-th candidate starting point The shortest Euclidean distance to all points on the typhoon's trajectory is calculated. The denominator uses `maxmin` for dynamic normalization, ensuring the safety metric is between [0,1]. For each candidate point... , Let T be the normalized weighted average distance from the candidate starting point to all high-priority nodes, where T is the set of nodes to be inspected. This indicates that all nodes to be inspected will be traversed, with i as the node index. Represents the coordinates of the node. The distance is the Euclidean distance between the candidate starting point and each node to be inspected. The priority weight of the node is set according to the node level. The level node is 3. The level node is 2. The level node is set to 1; Candidate sites must be located outside the typhoon-affected area as a safety constraint. (33)； in This is the current radius of the typhoon's influence. For safety margin.

7. The multi-UAV collaborative inspection method for power distribution networks under typhoon influence based on improved epsilon-constraint optimization as described in claim 1, characterized in that, In step S3 Landing points are intelligently merged as follows: First, collect all planned drone endpoints and set the endpoint set. ,in For the total number of drones, each endpoint Including coordinates and the planned arrival time of the drone ; Then based on the spatial proximity threshold Clustering of endpoints: If the Euclidean distance between two points is less than 1 / 2... Then it is considered mergeable, and thus Divide into several non-overlapping groups Each group End points within a group can be considered for merging. The size is ; Finally, define the binary decision variables. Indicate whether to group All endpoints within the range are merged to its geometric center. : (34)； If merge ( If so, then the group contributes 1 landing point. If not merged ( ), then contribute Given 1 initial endpoint, the optimization objective is to minimize the total number of final landing points used: (35)； The time it takes for the typhoon to reach the merge point must be later than the longest time it takes for all drones to fly from their original terminal points to the merge point, i.e., for each merge group. ,like Then, the spatiotemporal safety constraint must be met: the time required for the latest arriving drone to reach the merging point from its endpoint should be less than or equal to the time when the typhoon is expected to reach the merging point. (36)； in The point of arrival of the typhoon center The estimated time, For the drone's cruising speed, For the drone to fly from endpoint e to the merging point The time required.

8. A multi-UAV collaborative inspection power distribution network system based on improved epsilon-constraint optimization under typhoon influence, as described in any one of claims 1-7, characterized in that, include: Data acquisition and input module: used to acquire typhoon forecast data, power distribution network model data, equipment and social attribute data; Dynamic priority evaluation module: used to realize multi-dimensional risk fusion calculation and dynamic correction, and output node priority; Collaborative trajectory planning module: used to implement improved - Path optimization and solution using constraint methods; Dynamic take-off and landing point optimization module: used to realize the dynamic selection of the departure point and the intelligent merging of the landing point; Solution Output and Scheduling Module: This module integrates path and take-off / landing point information, generates executable collaborative inspection and scheduling instructions, and supports visual display.

Images