Multi-agent based collaborative scheduling method and system for distribution network unmanned aerial vehicle inspection

Abstract

The application discloses a multi-agent-based distribution network unmanned aerial vehicle inspection cooperative scheduling method and system, relates to the technical field of unmanned aerial vehicle scheduling, and comprises the following steps: acquiring task demand parameters of each inspection task and model capability parameters of each unmanned aerial vehicle, and calculating and acquiring a task matching degree set between each inspection task and each unmanned aerial vehicle; acquiring current state parameters of a nest, and calculating a nest task bearing pressure degree; constructing each to-be-allocated inspection task into a task agent, constructing each nest into a resource agent, performing multi-round negotiation, and obtaining an initial nest allocation scheme; iteratively optimizing the initial nest allocation scheme, acquiring an optimized nest allocation scheme; and issuing the inspection task to a corresponding nest for execution. The application solves the technical problems of low task and capability matching degree, unbalanced nest resource allocation and low task allocation coverage in the existing distribution network unmanned aerial vehicle inspection scheduling method.

Description

Technical Field

[0001] This application relates to the field of drone scheduling technology, specifically to a collaborative scheduling method and system for distribution network drone inspection based on multi-agent technology. Background Technology

[0002] With the continuous advancement of power distribution network construction, the distribution range of overhead lines is becoming increasingly wide. Manual inspection is not only inefficient but also poses high operational safety risks. Unmanned aerial vehicle (UAV) inspection, with its advantages of high mobility and wide coverage, is gradually becoming an important method for power distribution network inspection. As the application scale of UAV inspection continues to expand, multiple UAVs and multiple drone nests are often deployed in the same area, with multiple inspection tasks pending. How to coordinate and schedule inspection tasks, UAVs, and drone nest resources directly affects the overall efficiency of the inspection operation.

[0003] However, most existing methods for dispatching drones for power distribution network inspections allocate drone resources based on the proximity of the task area. This fails to match the actual needs of the inspection task with the capabilities of different drone models, which can easily lead to mismatches in capabilities, resulting in substandard inspection quality, or underutilization of drones and waste of resources.

[0004] Meanwhile, existing scheduling methods fail to fully consider the real-time task load of different nests, which can easily lead to uneven problems such as some nests being overloaded and causing overall task delays, while some nests are idle. Existing solutions that introduce multi-agent scheduling mostly have fixed negotiation strategies and do not consider dynamically adjusting the request strategy after the negotiation is rejected. They also lack iterative optimization after the initial allocation, which can easily result in many tasks not being allocated. The rationality of the scheduling results and the task allocation coverage need to be improved. Summary of the Invention

[0005] This application provides a collaborative scheduling method and system for distribution network drone inspection based on multi-agent technology, which solves the technical problems of low task-capability matching, uneven allocation of drone nest resources, and low task allocation coverage in existing distribution network drone inspection scheduling methods.

[0006] The technical solution to the above-mentioned technical problems in this application is as follows: Firstly, this application provides a collaborative scheduling method for unmanned aerial vehicle (UAV) inspection of power distribution networks based on multi-agent systems, the method comprising: Obtain the task requirement parameters of each inspection task in the set of inspection tasks to be assigned, and the model capability parameters of each drone in the set of available drones; Based on the task requirement parameters and the aircraft capability parameters, calculate and obtain the task matching degree set between each inspection task and each UAV. Obtain the current status parameters of the hive, and calculate the hive task load pressure based on the current status parameters; Each inspection task to be assigned is constructed as a task agent, and each nest is constructed as a resource agent. Based on the task matching degree and the nest task carrying pressure, multiple rounds of negotiation are conducted between the task agent and the resource agent to obtain an initial nest allocation scheme. When the number of task agents whose allocation requests are not accepted in the initial nest allocation scheme is greater than a preset unallocated ratio threshold, the initial nest allocation scheme is iteratively optimized to obtain an optimized nest allocation scheme. The optimized nest allocation scheme is used to distribute inspection tasks to the corresponding nests for execution.

[0007] Secondly, this application provides a multi-agent-based unmanned aerial vehicle (UAV) inspection and collaborative scheduling system for power distribution networks, including: The initial parameter acquisition module is used to obtain the task requirement parameters of each inspection task in the set of inspection tasks to be assigned, as well as the model capability parameters of each drone in the set of available drones. The matching degree calculation module is used to calculate and obtain the task matching degree set between each inspection task and each UAV based on the task requirement parameters and the aircraft capability parameters; The nest parameter acquisition module is used to acquire the current status parameters of the nest and calculate the nest task load pressure based on the current status parameters. The initial scheme acquisition module is used to construct each inspection task to be assigned as a task agent and each nest as a resource agent. Based on the task matching degree and the nest task carrying pressure, the module performs multiple rounds of negotiation between the task agent and the resource agent to obtain the initial nest allocation scheme. The allocation scheme optimization module is used to iteratively optimize the initial nest allocation scheme to obtain an optimized nest allocation scheme when the number of task agents whose allocation requests have not been accepted in the initial nest allocation scheme is greater than a preset unallocated ratio threshold. The inspection task execution module is used to distribute inspection tasks to the corresponding nests for execution using the optimized nest allocation scheme.

[0008] This application provides one or more technical solutions, which have at least the following technical effects or advantages: This application provides a multi-agent-based collaborative scheduling method and system for power distribution network drone inspections. First, by combining task requirement parameters and drone capability parameters, the matching degree between the task and the drone is calculated, ensuring precise adaptation between the inspection task and the drone's capabilities. This avoids both insufficient capability leading to substandard inspection quality and resource waste caused by high-performance drones performing low-requirement tasks. Second, by incorporating the current task load pressure of the computer nests into the decision-making basis for multi-agent negotiation, the task allocation of different nests can be balanced, avoiding resource imbalances where some nests are overloaded while others are idle. Finally, during the multi-agent negotiation process, rejected task agents are allowed to dynamically adjust their request thresholds and task urgency weights to re-initiate negotiations. Furthermore, after the negotiation concludes, the initial scheme with an excessive proportion of unassigned tasks is iteratively optimized, effectively improving task allocation coverage.

[0009] Through the above technical solution, this application solves the problems of low task-capability matching, uneven allocation of nest resources, and low task allocation coverage in traditional scheduling methods by using a dynamic negotiation and iterative optimization mechanism of multiple agents. It can effectively improve the overall scheduling efficiency of multi-UAV inspection of power distribution network and ensure that inspection operations are completed on time and with high quality. Attached Figure Description

[0010] To more clearly illustrate the technical solutions in the embodiments of this application, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0011] Figure 1 This is a flowchart illustrating the collaborative scheduling method for distribution network drone inspection based on multi-agent technology provided in this application embodiment; Figure 2 This is a schematic diagram of the structure of the multi-agent-based distribution network drone inspection and collaborative scheduling system provided in this application embodiment.

[0012] The components represented by each number in the attached diagram are explained below: Initial parameter acquisition module 11, matching degree calculation module 12, nest parameter acquisition module 13, initial scheme acquisition module 14, allocation scheme optimization module 15, and inspection task execution module 16. Detailed Implementation

[0013] This application provides a collaborative scheduling method and system for distribution network drone inspection based on multi-agent technology, which addresses the technical problems of low task-capability matching, uneven allocation of drone nest resources, and low task allocation coverage in existing distribution network drone inspection scheduling methods.

[0014] Example 1, as Figure 1 As shown in the embodiments of this application, a collaborative scheduling method for distribution network drone inspection based on multi-agent technology is provided, including: S10: Obtain the task requirement parameters of each inspection task in the set of inspection tasks to be assigned, and the model capability parameters of each drone in the set of available drones; In this embodiment of the application, the task requirement parameter is used to characterize the core capability requirements of the UAV for the inspection task to be assigned, while the model capability parameter corresponds to the actual capability parameters of each available UAV, which can be adjusted according to the actual task and UAV configuration.

[0015] Specifically, step S10 in the method includes: The task type, work area, estimated work duration and minimum shooting accuracy requirements of each inspection task are obtained from the inspection task management system. The endurance requirement weight, speed requirement weight, wind resistance requirement weight and hovering requirement weight corresponding to the task type are obtained from the task type and capability requirement weight mapping table and used as the task requirement parameters. The maximum flight time, cruising speed, wind resistance level, hovering accuracy level, and compatible payload type of each drone are obtained from the drone resource database and used as the capability parameters of the drone model.

[0016] In this embodiment, firstly, based on the inspection task management system, the task type, operation area, estimated operation time, and minimum shooting accuracy requirements of each inspection task are extracted. Then, according to a preset mapping relationship, the required weights of each capability are matched as task requirement parameters to quantify the capability requirements of different inspection tasks on the UAV. Meanwhile, the system extracts the model capability parameters of each drone from the drone resource database, including the maximum flight time, cruising speed, wind resistance level, hovering accuracy level, and compatible payload type. This can reflect the capability differences of different drones and ensure the accuracy of subsequent matching.

[0017] S20: Based on the task requirement parameters and the aircraft capability parameters, calculate and obtain the task matching degree set between each inspection task and each UAV; In this embodiment of the application, for each inspection task, the various capability parameters of each available UAV are normalized, and the weighted sum of the various capability requirements corresponding to the inspection task is performed to obtain the task matching degree between the inspection task and the current UAV. Then, after traversing all inspection tasks to be assigned and all available UAVs, a complete set of task matching degrees is obtained.

[0018] By combining the capability weights of the mission requirements with the weighted calculation of the UAV's capabilities, a quantitative matching degree can be obtained, making the matching results more in line with the actual core needs of different inspection missions and avoiding capability mismatch.

[0019] Specifically, step S20 in the method includes: Divide the maximum flight time of the drone by the estimated operation time of the inspection task to obtain the flight time matching score; Divide the drone's cruising speed by the preset benchmark cruising speed corresponding to the inspection task to obtain a speed matching score; Divide the wind resistance level of the drone by the preset benchmark wind resistance level corresponding to the inspection task to obtain the wind resistance matching score; Divide the hovering accuracy level of the UAV by the accuracy level value corresponding to the minimum shooting accuracy requirement of the inspection task to obtain the hovering matching score. The endurance matching score, speed matching score, wind resistance matching score, and hovering matching score are weighted and calculated to obtain the task matching degree between the inspection task and the UAV; The task matching degree between each inspection task and each UAV is traversed and calculated to obtain a task matching degree set, wherein the task matching degree set includes multiple task matching degrees.

[0020] In this embodiment of the application, firstly, the upper limit of the drone's endurance is divided by the expected operation time of the inspection task to obtain the endurance matching score, that is, the endurance matching score = upper limit of endurance ÷ expected operation time. The endurance matching score reflects the degree to which the drone's endurance capability meets the requirements of this inspection task.

[0021] Secondly, the speed matching score is obtained by dividing the drone's cruising speed by the preset benchmark cruising speed corresponding to the inspection task. That is, the speed matching score = cruising speed ÷ preset benchmark cruising speed. The preset benchmark cruising speed is preset according to the range of the inspection task's operation area. For example, it can be set to 100km / h. The larger the cruising range, the higher the preset benchmark cruising speed. This score can reflect the degree to which the drone's cruising speed meets the efficiency requirements of large-area inspection operations.

[0022] Next, divide the wind resistance level of the drone by the preset benchmark wind resistance level corresponding to the inspection task to obtain the wind resistance matching score, that is, wind resistance matching score = wind resistance level ÷ preset benchmark wind resistance level. The preset benchmark wind resistance level is determined based on the historical meteorological data of the operation area where the inspection task is located. The higher the wind speed in the operation area, the higher the preset benchmark wind resistance level. This score reflects the degree to which the drone's wind resistance capability meets the environmental requirements of the current inspection area.

[0023] Furthermore, the hovering accuracy level of the UAV is divided by the accuracy level value corresponding to the minimum shooting accuracy requirement of the inspection task to obtain the hovering matching score, that is, hovering matching score = hovering accuracy level ÷ accuracy level value corresponding to the minimum shooting accuracy requirement. The higher the shooting accuracy requirement, the higher the corresponding accuracy level value. This score reflects the degree to which the UAV's hovering shooting capability meets the inspection imaging accuracy requirements.

[0024] Finally, if any of the above matching scores are greater than 1, they are corrected to 1 to ensure that the score range is [0,1]. The four matching scores are then multiplied by the corresponding capability requirement weights and summed. The weight coefficients are obtained from a preset mapping table based on the task type. Different task types have different emphases on various capabilities. After traversing all combinations of inspection tasks to be assigned and available drones, a complete set of task matching scores can be obtained.

[0025] For example, if a certain inspection task to be assigned is a detailed inspection of a line in a mountainous area with strong winds, its endurance requirement weight is 0.3, speed requirement weight is 0.2, wind resistance requirement weight is 0.3, and hovering capability requirement weight is 0.2. The calculated endurance matching score for a certain UAV corresponding to this task is 0.9, speed matching score is 0.85, wind resistance matching score is 1.0, and hovering matching score is 1.0. The final task matching degree for this combination is: It can be determined that the drone is highly compatible with this mission.

[0026] S30: Obtain the current status parameters of the hive, and calculate the hive task load pressure based on the current status parameters; In this embodiment, the task load pressure is calculated based on the current state parameters of the hive. The higher the task load pressure, the smaller the remaining space of the hive that can accept new tasks. In the subsequent negotiation process, the priority of the selected allocation will be reduced, thereby balancing the task load of different hives and avoiding overload or idleness.

[0027] Specifically, step S30 in the method includes: Obtain the total number of pending tasks currently assigned to the drone nest but not yet executed, the total number of available drones, and the historical average execution time per task; Multiply the total number of tasks to be executed by the historical average execution time of a single task to obtain the estimated total execution time of the tasks to be executed. Obtain the preset maximum daily operating time of a single drone, multiply the number of available drones by the maximum daily operating time, and obtain the maximum daily service capacity of the drone nest. Divide the estimated total duration of the tasks to be executed by the maximum daily service capacity of the data center to obtain the task load pressure of the data center.

[0028] In this embodiment of the application, firstly, the total number of currently allocated but unexecuted tasks in the drone nest, the current number of available drones, and the average execution time of a single task are obtained. The total number of unexecuted tasks includes inspection tasks that have been previously scheduled and allocated but have not yet started execution. Completed tasks are not included in the statistics to avoid overestimating the load capacity of the drone nest.

[0029] Then, by multiplying the total number of tasks to be executed by the historical average execution time of a single task, the estimated total time of tasks currently accumulated in the nest is obtained, which intuitively reflects the total scale of tasks that the nest currently needs to process.

[0030] Secondly, by multiplying the number of available drones in the hive by the preset maximum daily operating time of a single drone, the maximum service capacity that the hive can provide in a single day is obtained. Finally, by dividing the estimated total duration of tasks to be executed by the maximum daily service capacity of the hive, the task load pressure of the hive in the range of 0 to 1 is obtained. The higher the value, the smaller the remaining capacity of the hive. Through the above calculation method, the current task load of different hives can be quantitatively reflected, providing an accurate decision-making basis for subsequent multi-agent negotiation.

[0031] For example, a certain drone nest currently has 5 assigned but unexecuted inspection tasks. The historical average execution time of a single task is 1.2 hours. The number of drones currently available in the nest is 3. The preset maximum daily working time of a single drone is 8 hours. Then the task load pressure of the nest is calculated as: (5×1.2)÷(3×8)=0.25, indicating that the current load pressure of the nest is low and there is sufficient space remaining to accommodate new tasks.

[0032] S40: Construct each inspection task to be assigned as a task agent and each nest as a resource agent. Based on the task matching degree and the nest task carrying pressure, conduct multiple rounds of negotiation between the task agent and the resource agent to obtain an initial nest allocation scheme. In this embodiment, each task agent has its own task requirements, urgency, and matching information, while each resource agent has its own task load and available drone information. The two agents complete the allocation through distributed autonomous negotiation, eliminating the need for a centralized scheduling center to perform a one-time forced allocation, making the allocation process more flexible.

[0033] Specifically, step S40 in the method includes: Each of the inspection tasks to be assigned is constructed as a task intelligent agent, and each of the machine nests is constructed as a resource intelligent agent; Obtain the request intention value of the task agent; When the request intention value is greater than the preset request threshold, the task agent initiates an allocation request to the resource agent, which includes the request intention value and the task urgency weight corresponding to the task agent. The resource agent collects the allocation requests received in the current round, and multiplies the request willingness value of the allocation request by the task urgency weight to obtain the request priority score; The allocation requests are sorted in descending order of priority score. Based on the number of available drones remaining in the resource agent, allocation requests are accepted sequentially from the top of the sorting queue, and the corresponding task agent is notified of the acceptance of allocation confirmation until the number of available drones is zero or the sorting queue has been traversed. For any unaccepted allocation request, feedback is sent to the task agent corresponding to the unaccepted allocation request, indicating a rejection of allocation and the current nest task load pressure.

[0034] In this embodiment of the application, each inspection task to be assigned is first constructed as a task agent, and each nest is constructed as a resource agent. Specifically, each task agent corresponds to an inspection task to be assigned, and each resource agent corresponds to a nest that is put into use. Each agent stores its own relevant decision parameters, without relying on a centralized scheduling node to store all information.

[0035] Secondly, the request willingness value of the task agent is obtained. The request willingness value combines the matching degree of the drone's capabilities and the current load pressure of the nest. The higher the willingness value, the stronger the rationality of the task being assigned to the nest.

[0036] Furthermore, the task agent only initiates requests to resource agents whose request willingness value exceeds a preset request threshold, reducing unnecessary information interaction and improving negotiation efficiency. After receiving all allocation requests in this round, the resource agent multiplies the request willingness value of each request by the urgency weight of the corresponding task to obtain the final request priority score. The higher the urgency of the task, the greater the urgency weight, which can ensure that high-urgency tasks are given priority in resource allocation.

[0037] Afterwards, resource agents accept allocation requests in descending order of priority until their remaining available drones are exhausted or all requests are processed. Task agents that have not received allocations can re-initiate requests to other resource agents in the next round of negotiation.

[0038] Specifically, after receiving the rejection information, the task agent adjusts its preset request threshold for the next round of negotiation based on the current task load pressure of each feedback resource agent, updates the task urgency weight according to the remaining waiting time for scheduling, recalculates the request willingness value to each resource agent, and initiates the next round of negotiation.

[0039] The above multi-round negotiation process is repeated until no task agent initiates a new allocation request. After the negotiation ends, all accepted allocation requests are aggregated to obtain the initial hive allocation scheme. Through distributed multi-round autonomous negotiation, tasks and resources can autonomously complete the matching. At the same time, the negotiation priority is dynamically adjusted according to the urgency of the tasks, which can ensure that urgent tasks are allocated first, and can also balance the task allocation of different hives to avoid excessive pressure on any one hive.

[0040] Further, obtaining the request intention value of the task agent includes: Each task agent calculates the maximum value of the task matching degree between the task agent and each UAV under the jurisdiction of the resource agent for each resource agent, and obtains the highest task matching degree. The initial request intention value is obtained by multiplying the highest task matching degree by 1 and the difference between the nest task carrying pressure degree. When the current negotiation round is the first round of negotiation, the initial request intention value will be used as the request intention value; If the current negotiation round is not the first round of negotiation, obtain the historical request intention value corresponding to when the task agent initiated the allocation request to the resource agent in the previous round of negotiation; Multiply the initial request willingness value by the rejection decay coefficient to obtain the current willingness component; Calculate the first difference between 1 and the historical intention retention coefficient, and multiply the current intention component by the first difference to obtain the current intention adjustment component; Multiply the historical request intention value by the historical intention retention coefficient to obtain the historical intention retention component; The historical intention retention component is added to the current intention adjustment component to obtain the corrected request intention value, which is then used as the request intention value.

[0041] In this embodiment of the application, firstly, for each resource agent, each task agent finds the highest task matching degree among all the drones under the resource agent's jurisdiction. Specifically, this is obtained by iterating through the task matching degrees of all available drones under the resource agent and comparing them. This highest matching degree directly reflects the capability adaptation level of the drones most suitable for the task within the resource agent's jurisdiction.

[0042] Secondly, the initial request willingness value is obtained by multiplying the highest task matching degree by 1 and subtracting the task load pressure of the nest. If the initial request willingness value is negative, that is, when the pressure is greater than 1, the initial request willingness value is truncated to 0 to prevent the willingness value from being negative. Specifically, the initial request willingness value = highest task matching degree × (1 - nest task load pressure). The resulting initial request willingness value retains the capability matching advantage between the drone and the task, and is also adjusted according to the current load pressure of the nest. The lower the load pressure, the higher the initial request willingness value, which is more in line with the rational logic of task allocation.

[0043] Furthermore, if it is the first round of negotiation, the initial request intention value can be directly used as the final request intention value; if it is not the first round of negotiation, the historical request intention value when the task agent made a request to the resource agent in the previous round of negotiation is taken out, the current intention component is calculated first, that is, the initial request intention value is multiplied by the preset rejection attenuation coefficient, the historical request intention value is multiplied by the historical intention retention coefficient, and the current intention component is multiplied by 1 and the historical intention retention coefficient is subtracted to finally obtain the corrected request intention value, which is used as the request intention value for the current round.

[0044] Specifically, the request intention value of the current round = historical request intention value × historical intention retention coefficient + initial request intention value × rejection decay coefficient × (1 - historical intention retention coefficient). Through the above correction method, reasonable intention information in the historical negotiation process can be retained. At the same time, the intention of resource agents that are continuously rejected is decayed, guiding task agents to prioritize resource agents that have not been tried and have lower load pressure, further improving negotiation convergence efficiency, avoiding task agents from continuously launching invalid requests to the same high-load resource agent, and reducing unnecessary communication losses between multiple agents.

[0045] For example, when a task agent initiates a request to a resource agent in the first round, the resource agent's highest task matching degree is 0.94, and the nested task load pressure is 0.25. The calculated initial request willingness value is... In the first round, 0.705 is used as the willingness value for the request. If the request is not accepted, in the second round of negotiation, the preset rejection attenuation coefficient is 0.8, and the historical willingness retention coefficient is 0.3. Therefore, the revised willingness value for the request in the second round is... This is a decrease from 0.705 in the first round, which is consistent with the adjustment logic of diminishing willingness after rejection. It can guide the task agent to reduce the priority of initiating against the resource agent in the next round of negotiation and turn to other resource agents with more lenient requirements.

[0046] Specifically, the multi-round negotiation also includes: In the second and subsequent rounds of negotiation, the task agent that was rejected in the previous round multiplies the preset request threshold by a preset threshold decay coefficient to obtain a modified request threshold, and multiplies the task urgency weight of the task agent by a preset waiting accumulation coefficient to obtain a modified task urgency weight. A new allocation request is initiated based on the aforementioned correction request threshold and the urgency weight of the correction task; The negotiation terminates when all the task agents receive confirmation of acceptance of the allocation, or when the task allocation status remains unchanged in two consecutive rounds of negotiation, or when the number of negotiation rounds reaches the preset maximum number of negotiation rounds. The current accepted allocation results are then output as the initial nest allocation scheme.

[0047] In this embodiment, firstly, in the second and subsequent rounds of negotiation, the task agent that has been rejected in one round will lower its request threshold. The original preset request threshold is modified by a threshold decay coefficient. For example, if the preset decay coefficient is less than 1, the modified request threshold is lower, which can allow resource agents with slightly lower willingness values ​​but still have the possibility of allocation to enter the candidate range, expand the range of optional resources for the task agent, and increase the probability of successful allocation.

[0048] Meanwhile, the more rounds a task waits without being assigned, the higher its urgency level becomes. Therefore, by multiplying the original task urgency weight by a waiting accumulation coefficient greater than 1, the urgency weight is adjusted to be larger, allowing tasks that have been waiting for a longer period to receive higher priority in subsequent negotiations, thus preventing urgent tasks from remaining unassigned for extended periods.

[0049] Secondly, the allocation request is re-initiated based on the revised request threshold and the revised task urgency weight. That is, each round of negotiation adapts the current allocation status by adjusting the request threshold and urgency weight. When the termination condition is met, the negotiation stops immediately and the currently determined allocation result is output as the initial nest allocation scheme.

[0050] S50: When the number of task agents whose allocation requests have not been accepted in the initial nest allocation scheme is greater than the preset unallocated ratio threshold, the initial nest allocation scheme is iteratively optimized to obtain an optimized nest allocation scheme. In this embodiment of the application, if a large number of tasks are still not allocated after multiple rounds of negotiation, it indicates that the overall resources are currently tight. It is necessary to iteratively optimize and adjust the already allocated schemes to allocate resources to more urgent and more reasonably matched tasks, so as to avoid a large backlog of tasks that cannot be allocated.

[0051] Specifically, step S50 in the method includes: The number of task agents whose allocation requests were not accepted in the initial nest allocation scheme is counted as the number of unallocated tasks. Divide the number of unassigned tasks by the total number of inspection tasks to be assigned to obtain the proportion of unassigned tasks. When the proportion of unassigned tasks exceeds a preset unassigned task proportion threshold, iterative optimization is triggered to obtain an optimized nest allocation scheme.

[0052] In this embodiment, firstly, the proportion of unassigned tasks to the total number of inspection tasks to be assigned is calculated. When the unassigned proportion exceeds a preset threshold, it indicates that the existing initial allocation scheme has not made full use of existing resources, or that there is a situation where the allocation of some tasks is not reasonable enough. It is necessary to release the unreasonable allocated resources to the unassigned tasks through iterative adjustment to improve the overall allocation completion rate.

[0053] If the proportion of unassigned tasks exceeds the preset unassigned task proportion threshold, an iterative optimization mechanism is triggered. In this mechanism, each iteration allocates only one priority object until the proportion of unassigned tasks is no greater than the preset unassigned task proportion threshold or there are no tasks to be assigned, at which point an optimized nest allocation scheme is obtained.

[0054] Specifically, when the proportion of unassigned tasks exceeds a preset unassigned task proportion threshold, iterative optimization is triggered to obtain an optimized nest allocation scheme, including: Among the task agents that have never had their assignment requests accepted, the task agent with the highest task urgency weight is selected as the priority assignment target. Traverse each resource agent, calculate the updated nest task load pressure of the resource agent after the priority allocation object is assigned to the resource agent, and select the resource agent with the lowest updated nest task load pressure as the target resource agent. The priority allocation object is assigned to the target resource agent, the allocation result is obtained, and the allocation result is added to the initial nest allocation scheme. At the same time, the priority allocation object is removed from the task agents whose allocation requests have not been accepted. The proportion of unassigned tasks is recalculated. If the proportion of unassigned tasks is still greater than the preset unassigned task proportion threshold, the process is iterated until the proportion of unassigned tasks is no greater than the preset unassigned task proportion threshold or all task agents that have not accepted assignment requests have completed their processing, thus obtaining the optimized nest allocation scheme.

[0055] In this embodiment of the application, firstly, for tasks that have not yet been assigned, the task with the highest urgency is given priority to obtain resources. The task with the highest urgency weight is selected as the priority allocation target. Then, the resource agents for all assigned tasks are calculated one by one. If the priority allocation target is added to the nest, the nest task load pressure is updated. Then, the resource agent corresponding to the nest with the lowest overall pressure after the load pressure is increased is selected and determined as the target resource agent.

[0056] Furthermore, after the assignment is completed, the task is removed from the list of unassigned tasks, the current initial hive assignment scheme is updated, and then the current unassigned task ratio is recalculated. If it still exceeds the preset unassigned task ratio threshold, the above steps are repeated to continue processing the next unassigned task with the highest urgency, until the unassigned task ratio meets the requirements, or all unassigned tasks have been processed, and finally the optimized hive assignment scheme is obtained.

[0057] Through the above iterative optimization, under the condition of overall resource shortage, priority can be given to ensuring that high-urgency tasks are allocated, thereby maximizing the overall task allocation completion rate and avoiding a backlog of a large number of high-urgency tasks that cannot be carried out during inspections.

[0058] S60: The inspection task is assigned to the corresponding nest for execution using the optimized nest allocation scheme.

[0059] In this embodiment, based on the determined optimized nest allocation scheme, each inspection task to be allocated is directly sent to the corresponding nest according to the allocation result. The nest then schedules the drones under its jurisdiction to go to the designated area to carry out power distribution network inspection work according to the task requirements. The entire allocation and scheduling process is completed through multi-agent distributed autonomous negotiation combined with iterative optimization.

[0060] In summary, compared to existing technologies, this application achieves task and drone resource matching through distributed multi-agent autonomous negotiation. This eliminates the need for centralized scheduling nodes to store all parameters, reducing the computational and communication burden on centralized nodes. Furthermore, by dynamically adjusting request willingness, request thresholds, and allocation priorities during the negotiation process based on task urgency and drone load, it ensures that high-urgency tasks receive resources first while balancing the load on different drones, preventing excessive pressure on a single node. Additionally, addressing the issue of numerous unallocated tasks in resource-constrained scenarios, iterative optimization further adjusts the allocation scheme, prioritizing high-urgency tasks and effectively improving the overall task allocation completion rate. Ultimately, this enhances the rationality and efficiency of collaborative scheduling for power distribution network drone inspections, meeting the scheduling needs of large-scale inspection tasks in multi-drone, multi-drone scenarios.

[0061] In summary, the embodiments of this application have at least the following technical effects: This application provides a multi-agent-based collaborative scheduling method for distribution network UAV inspections. First, by combining task requirement parameters and UAV capability parameters, the matching degree between tasks and UAVs is calculated, ensuring precise adaptation between inspection tasks and UAV capabilities. This avoids both insufficient capabilities leading to substandard inspection quality and resource waste caused by high-performance UAVs performing low-requirement tasks. Second, by incorporating the current task load pressure of computer nests into the decision-making basis for multi-agent negotiation, the task allocation of different nests can be balanced, avoiding resource imbalances where some nests are overloaded while others are idle. Finally, during multi-agent negotiation, rejected task agents are allowed to dynamically adjust request thresholds and task urgency weights to re-initiate negotiations. Furthermore, after negotiation, the initial scheme with an excessive proportion of unassigned tasks is iteratively optimized, effectively improving task allocation coverage.

[0062] Through the above technical solution, this application solves the problems of low task-capability matching, uneven allocation of nest resources, and low task allocation coverage in traditional scheduling methods by using a dynamic negotiation and iterative optimization mechanism of multiple agents. It can effectively improve the overall scheduling efficiency of multi-UAV inspection of power distribution network and ensure that inspection operations are completed on time and with high quality.

[0063] Example 2, as Figure 2 As shown, based on the same inventive concept as the multi-agent-based distribution network drone inspection collaborative scheduling method provided in Embodiment 1, this application also provides a multi-agent-based distribution network drone inspection collaborative scheduling system, including: The initial parameter acquisition module 11 is used to acquire the task requirement parameters of each inspection task in the set of inspection tasks to be assigned, as well as the model capability parameters of each drone in the set of available drones. Matching degree calculation module 12 is used to calculate and obtain the task matching degree set between each inspection task and each UAV based on the task requirement parameters and the aircraft capability parameters; Nest parameter acquisition module 13 is used to acquire the current status parameters of the nest and calculate the nest task load pressure based on the current status parameters; The initial scheme acquisition module 14 is used to construct each inspection task to be assigned as a task agent and each nest as a resource agent, and to conduct multiple rounds of negotiation between the task agent and the resource agent based on the task matching degree and the nest task carrying pressure degree to obtain an initial nest allocation scheme. The allocation scheme optimization module 15 is used to iteratively optimize the initial nest allocation scheme to obtain an optimized nest allocation scheme when the number of task agents whose allocation requests have not been accepted in the initial nest allocation scheme is greater than a preset unallocated ratio threshold. The inspection task execution module 16 is used to distribute the inspection task to the corresponding nest for execution using the optimized nest allocation scheme.

[0064] In one embodiment, the initial parameter acquisition module 11 is specifically used for: The task type, work area, estimated work duration and minimum shooting accuracy requirements of each inspection task are obtained from the inspection task management system. The endurance requirement weight, speed requirement weight, wind resistance requirement weight and hovering requirement weight corresponding to the task type are obtained from the task type and capability requirement weight mapping table and used as the task requirement parameters. The maximum flight time, cruising speed, wind resistance level, hovering accuracy level, and compatible payload type of each UAV are obtained from the UAV resource database and used as the capability parameters of the aforementioned UAV models.

[0065] In one embodiment, the matching degree calculation module 12 is specifically used for: Divide the maximum flight time of the drone by the estimated operation time of the inspection task to obtain the flight time matching score; Divide the drone's cruising speed by the preset benchmark cruising speed corresponding to the inspection task to obtain a speed matching score; Divide the wind resistance level of the drone by the preset benchmark wind resistance level corresponding to the inspection task to obtain the wind resistance matching score; Divide the hovering accuracy level of the UAV by the accuracy level value corresponding to the minimum shooting accuracy requirement of the inspection task to obtain the hovering matching score. The endurance matching score, speed matching score, wind resistance matching score, and hovering matching score are weighted and calculated to obtain the task matching degree between the inspection task and the UAV; The task matching degree between each inspection task and each UAV is traversed and calculated to obtain a task matching degree set, wherein the task matching degree set includes multiple task matching degrees.

[0066] In one embodiment of the application, the nest parameter acquisition module 13 is specifically used for: Obtain the total number of pending tasks currently assigned to the drone nest but not yet executed, the total number of available drones, and the historical average execution time per task; Multiply the total number of tasks to be executed by the historical average execution time of a single task to obtain the estimated total execution time of the tasks to be executed. Obtain the preset maximum daily operating time of a single drone, multiply the number of available drones by the maximum daily operating time, and obtain the maximum daily service capacity of the drone nest. Divide the estimated total duration of the tasks to be executed by the maximum daily service capacity of the data center to obtain the task load pressure of the data center.

[0067] In one application embodiment, the initial scheme acquisition module 14 is specifically used for: Each of the inspection tasks to be assigned is constructed as a task intelligent agent, and each of the machine nests is constructed as a resource intelligent agent; Obtain the request intention value of the task agent; When the request intention value is greater than the preset request threshold, the task agent initiates an allocation request to the resource agent, which includes the request intention value and the task urgency weight corresponding to the task agent. The resource agent collects the allocation requests received in the current round, and multiplies the request willingness value of the allocation request by the task urgency weight to obtain the request priority score; The allocation requests are sorted in descending order of priority score. Based on the number of available drones remaining in the resource agent, allocation requests are accepted sequentially from the top of the sorting queue, and the corresponding task agent is notified of the acceptance of allocation confirmation until the number of available drones is zero or the sorting queue has been traversed. For any unaccepted allocation request, feedback is sent to the task agent corresponding to the unaccepted allocation request, indicating a rejection of allocation and the current nest task load pressure.

[0068] Further, obtaining the request intention value of the task agent includes: Each task agent calculates the maximum value of the task matching degree between the task agent and each UAV under the jurisdiction of the resource agent for each resource agent, and obtains the highest task matching degree. The initial request intention value is obtained by multiplying the highest task matching degree by 1 and the difference between the nest task carrying pressure degree. When the current negotiation round is the first round of negotiation, the initial request intention value will be used as the request intention value; If the current negotiation round is not the first round of negotiation, obtain the historical request intention value corresponding to when the task agent initiated the allocation request to the resource agent in the previous round of negotiation; Multiply the initial request willingness value by the rejection decay coefficient to obtain the current willingness component; Calculate the first difference between 1 and the historical intention retention coefficient, and multiply the current intention component by the first difference to obtain the current intention adjustment component; Multiply the historical request intention value by the historical intention retention coefficient to obtain the historical intention retention component; The historical intention retention component is added to the current intention adjustment component to obtain the corrected request intention value, which is then used as the request intention value.

[0069] Furthermore, the multi-round negotiation also includes: In the second and subsequent rounds of negotiation, the task agent that was rejected in the previous round multiplies the preset request threshold by a preset threshold decay coefficient to obtain a modified request threshold, and multiplies the task urgency weight of the task agent by a preset waiting accumulation coefficient to obtain a modified task urgency weight. A new allocation request is initiated based on the aforementioned correction request threshold and the urgency weight of the correction task; The negotiation terminates when all the task agents receive confirmation of acceptance of the allocation, or when the task allocation status remains unchanged in two consecutive rounds of negotiation, or when the number of negotiation rounds reaches the preset maximum number of negotiation rounds. The current accepted allocation results are then output as the initial nest allocation scheme.

[0070] In one embodiment, the allocation scheme optimization module 15 is specifically used for: The number of task agents whose allocation requests were not accepted in the initial nest allocation scheme is counted as the number of unallocated tasks. Divide the number of unassigned tasks by the total number of inspection tasks to be assigned to obtain the proportion of unassigned tasks. When the proportion of unassigned tasks exceeds a preset unassigned task proportion threshold, iterative optimization is triggered to obtain an optimized nest allocation scheme.

[0071] Furthermore, when the proportion of unassigned tasks exceeds a preset unassigned task proportion threshold, iterative optimization is triggered to obtain an optimized nest allocation scheme, including: Among the task agents that have never had their assignment requests accepted, the task agent with the highest task urgency weight is selected as the priority assignment target. Traverse each resource agent, calculate the updated nest task load pressure of the resource agent after the priority allocation object is assigned to the resource agent, and select the resource agent with the lowest updated nest task load pressure as the target resource agent. The priority allocation object is assigned to the target resource agent, the allocation result is obtained, and the allocation result is added to the initial nest allocation scheme. At the same time, the priority allocation object is removed from the task agents whose allocation requests have not been accepted. The proportion of unassigned tasks is recalculated. If the proportion of unassigned tasks is still greater than the preset unassigned task proportion threshold, the process is iterated until the proportion of unassigned tasks is no greater than the preset unassigned task proportion threshold or all task agents that have not accepted assignment requests have completed their processing, thus obtaining the optimized nest allocation scheme.

Claims

1. A collaborative scheduling method for UAV inspection of power distribution networks based on multi-agent systems, characterized in that, include: Obtain the task requirement parameters of each inspection task in the set of inspection tasks to be assigned, and the model capability parameters of each drone in the set of available drones; Based on the task requirement parameters and the aircraft capability parameters, calculate and obtain the task matching degree set between each inspection task and each UAV. Obtain the current status parameters of the hive, and calculate the hive task load pressure based on the current status parameters; Each inspection task to be assigned is constructed as a task agent, and each hive is constructed as a resource agent. Based on the task matching degree and the hive task load pressure, multiple rounds of negotiation are conducted between the task agents and the resource agents to obtain an initial hive allocation scheme, including: Each of the inspection tasks to be assigned is constructed as a task intelligent agent, and each of the machine nests is constructed as a resource intelligent agent; Obtain the request intention value of the task agent; When the request intention value is greater than the preset request threshold, the task agent initiates an allocation request to the resource agent, which includes the request intention value and the task urgency weight corresponding to the task agent. The resource agent collects the allocation requests received in the current round, and multiplies the request willingness value of the allocation request by the task urgency weight to obtain the request priority score; The allocation requests are sorted in descending order of priority score. Based on the number of available drones remaining in the resource agent, allocation requests are accepted sequentially from the top of the sorting queue, and the corresponding task agent is notified of the acceptance of allocation confirmation until the number of available drones is zero or the sorting queue has been traversed. For any unaccepted allocation request, feedback of allocation rejection information and current nest task load pressure is sent to the task agent corresponding to the unaccepted allocation request. When the number of task agents whose allocation requests are not accepted in the initial nest allocation scheme is greater than a preset unallocated ratio threshold, the initial nest allocation scheme is iteratively optimized to obtain an optimized nest allocation scheme. The optimized nest allocation scheme is used to distribute inspection tasks to the corresponding nests for execution.

2. The method for collaborative scheduling of UAV inspection of power distribution networks based on multi-agent systems according to claim 1, characterized in that, The process of obtaining the task requirement parameters of each inspection task in the set of inspection tasks to be assigned, and the model capability parameters of each drone in the set of available drones, includes: The task type, work area, estimated work duration and minimum shooting accuracy requirements of each inspection task are obtained from the inspection task management system. The endurance requirement weight, speed requirement weight, wind resistance requirement weight and hovering requirement weight corresponding to the task type are obtained from the task type and capability requirement weight mapping table and used as the task requirement parameters. The maximum flight time, cruising speed, wind resistance level, hovering accuracy level, and compatible payload type of each UAV are obtained from the UAV resource database and used as the capability parameters of the aforementioned UAV models.

3. The method for collaborative scheduling of UAV inspection of power distribution networks based on multi-agent systems according to claim 1, characterized in that, The step of calculating and obtaining a set of task matching degrees between each inspection task and each UAV based on the task requirement parameters and the aircraft capability parameters includes: Divide the maximum flight time of the drone by the estimated operation time of the inspection task to obtain the flight time matching score; Divide the drone's cruising speed by the preset benchmark cruising speed corresponding to the inspection task to obtain a speed matching score; Divide the wind resistance level of the drone by the preset benchmark wind resistance level corresponding to the inspection task to obtain the wind resistance matching score; Divide the hovering accuracy level of the UAV by the accuracy level value corresponding to the minimum shooting accuracy requirement of the inspection task to obtain the hovering matching score. The endurance matching score, speed matching score, wind resistance matching score, and hovering matching score are weighted and calculated to obtain the task matching degree between the inspection task and the UAV; The task matching degree between each inspection task and each UAV is traversed and calculated to obtain a task matching degree set, wherein the task matching degree set includes multiple task matching degrees.

4. The method for collaborative scheduling of UAV inspection of power distribution networks based on multi-agent systems according to claim 1, characterized in that, Obtain the current state parameters of the hive, and calculate the hive task load pressure based on the current state parameters, including: Obtain the total number of pending tasks currently assigned to the drone nest but not yet executed, the total number of available drones, and the historical average execution time per task; Multiply the total number of tasks to be executed by the historical average execution time of a single task to obtain the estimated total execution time of the tasks to be executed. Obtain the preset maximum daily operating time of a single drone, multiply the number of available drones by the maximum daily operating time, and obtain the maximum daily service capacity of the drone nest. Divide the estimated total duration of the tasks to be executed by the maximum daily service capacity of the data center to obtain the task load pressure of the data center.

5. The method for collaborative scheduling of UAV inspection of power distribution networks based on multi-agent systems according to claim 1, characterized in that, Obtaining the request intention value of the task agent includes: Each task agent calculates the maximum value of the task matching degree between the task agent and each UAV under the jurisdiction of the resource agent for each resource agent, and obtains the highest task matching degree. The initial request intention value is obtained by multiplying the highest task matching degree by 1 and the difference between the nest task carrying pressure degree. When the current negotiation round is the first round of negotiation, the initial request intention value will be used as the request intention value; If the current negotiation round is not the first round of negotiation, obtain the historical request intention value corresponding to when the task agent initiated the allocation request to the resource agent in the previous round of negotiation; Multiply the initial request willingness value by the rejection decay coefficient to obtain the current willingness component; Calculate the first difference between 1 and the historical intention retention coefficient, and multiply the current intention component by the first difference to obtain the current intention adjustment component; Multiply the historical request intention value by the historical intention retention coefficient to obtain the historical intention retention component; The historical intention retention component is added to the current intention adjustment component to obtain the corrected request intention value, which is then used as the request intention value.

6. The method for collaborative scheduling of UAV inspection of power distribution networks based on multi-agent systems according to claim 1, characterized in that, The multi-round negotiation also includes: In the second and subsequent rounds of negotiation, the task agent that was rejected in the previous round multiplies the preset request threshold by a preset threshold decay coefficient to obtain a modified request threshold, and multiplies the task urgency weight of the task agent by a preset waiting accumulation coefficient to obtain a modified task urgency weight. A new allocation request is initiated based on the aforementioned correction request threshold and the urgency weight of the correction task; The negotiation terminates when all the task agents receive confirmation of acceptance of the allocation, or when the task allocation status remains unchanged in two consecutive rounds of negotiation, or when the number of negotiation rounds reaches the preset maximum number of negotiation rounds. The current accepted allocation results are then output as the initial nest allocation scheme.

7. The method for collaborative scheduling of UAV inspection of power distribution networks based on multi-agent systems according to claim 5, characterized in that, When the number of task agents whose allocation requests have not been accepted in the initial nest allocation scheme exceeds a preset unallocated proportion threshold, the initial nest allocation scheme is iteratively optimized to obtain an optimized nest allocation scheme, including: The number of task agents whose allocation requests were not accepted in the initial nest allocation scheme is counted as the number of unallocated tasks. Divide the number of unassigned tasks by the total number of inspection tasks to be assigned to obtain the proportion of unassigned tasks. When the proportion of unassigned tasks exceeds a preset unassigned task proportion threshold, iterative optimization is triggered to obtain an optimized nest allocation scheme.

8. The method for collaborative scheduling of UAV inspection of power distribution networks based on multi-agent technology according to claim 7, characterized in that, When the proportion of unassigned tasks exceeds a preset unassigned task proportion threshold, iterative optimization is triggered to obtain an optimized nest allocation scheme, including: Among the task agents that have never had their assignment requests accepted, the task agent with the highest task urgency weight is selected as the priority assignment target. Traverse each resource agent, calculate the updated nest task load pressure of the resource agent after the priority allocation object is assigned to the resource agent, and select the resource agent with the lowest updated nest task load pressure as the target resource agent. The priority allocation object is assigned to the target resource agent, the allocation result is obtained, and the allocation result is added to the initial nest allocation scheme. At the same time, the priority allocation object is removed from the task agents whose allocation requests have not been accepted. The proportion of unassigned tasks is recalculated. If the proportion of unassigned tasks is still greater than the preset unassigned task proportion threshold, the process is iterated until the proportion of unassigned tasks is no greater than the preset unassigned task proportion threshold or all task agents that have not accepted assignment requests have completed their processing, thus obtaining the optimized nest allocation scheme.

9. A multi-agent-based unmanned aerial vehicle (UAV) inspection and collaborative scheduling system for power distribution networks, characterized in that: The method for implementing the multi-agent-based UAV inspection and collaborative scheduling method for power distribution networks as described in any one of claims 1-8 includes: The initial parameter acquisition module is used to obtain the task requirement parameters of each inspection task in the set of inspection tasks to be assigned, as well as the model capability parameters of each drone in the set of available drones. The matching degree calculation module is used to calculate and obtain the task matching degree set between each inspection task and each UAV based on the task requirement parameters and the aircraft capability parameters; The nest parameter acquisition module is used to acquire the current status parameters of the nest and calculate the nest task load pressure based on the current status parameters. The initial scheme acquisition module is used to construct each inspection task to be assigned as a task agent and each nest as a resource agent. Based on the task matching degree and the nest task carrying pressure, the module performs multiple rounds of negotiation between the task agent and the resource agent to obtain the initial nest allocation scheme. The allocation scheme optimization module is used to iteratively optimize the initial nest allocation scheme to obtain an optimized nest allocation scheme when the number of task agents whose allocation requests have not been accepted in the initial nest allocation scheme is greater than a preset unallocated ratio threshold. The inspection task execution module is used to distribute inspection tasks to the corresponding nests for execution using the optimized nest allocation scheme.

Images