A resource allocation optimization method for unmanned aerial vehicle power inspection

Abstract

A resource allocation optimization method for unmanned aerial vehicle power inspection, comprising: forming a federated learning system by an unmanned aerial vehicle performing an inspection task and nearby edge servers and ground base stations; the unmanned aerial vehicle uploads edge server parameters to the ground base station; the ground base station calculates the optimal computing load of each edge server according to the edge server parameters; the ground base station calculates the edge server most suitable for transmission by the unmanned aerial vehicle as the optimal edge server transmission node and broadcasts it to the unmanned aerial vehicle, and the unmanned aerial vehicle sends data to the edge server according to the optimal edge server transmission node and the optimal computing load; the ground base station broadcasts the initial model parameters to the edge server to start a new round of federated learning model training; and the ground base station judges whether the model training reaches the optimal model coordination round, and if so, stops the global model training. The present application reasonably schedules the task allocation of the unmanned aerial vehicle, reduces the total communication cost of the system, and improves the operation efficiency and inspection quality of the power system.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of unmanned aerial vehicle (UAV) power inspection, and specifically to a resource allocation optimization method for UAV power inspection. BACKGROUND

[0002] With the increasing demand for equipment inspection and maintenance in power systems, unmanned aerial vehicles (UAVs) have been widely used in the monitoring and maintenance of power facilities as an efficient inspection tool. UAVs have high flexibility, wide coverage, and low cost, and can collect data in real time in complex environments. However, with the increasing number of UAVs and the increasing complexity of tasks, how to efficiently allocate the computing tasks of UAVs has become an urgent problem to be solved.

[0003] In power systems, UAVs will pass through multiple ground relay nodes, including base stations (BSs) and ground nodes (GNs), during flight. UAVs can transmit computing tasks to these ground relay nodes through licensed band transmission (LBT) or unlicensed bandwidth transmission (UBT). After receiving the computing tasks, the relay nodes will send them to the cloud or edge server for processing.

[0004] Traditional computing task allocation methods mostly rely on centralized management systems, which may be effective when dealing with simple tasks, but often appear to be inadequate when facing dynamic changes in power systems and multiple UAVs working together. In addition, existing technologies do not fully consider the optimization of communication costs during task allocation, resulting in low system efficiency and unnecessary communication overhead.

[0005] Therefore, the present application proposes a method for optimizing the allocation of existing computing tasks based on UAVs, aiming to reduce the total communication cost of the system and improve the operation efficiency and inspection quality of the power system by reasonably scheduling the task allocation of UAVs. This method not only improves the work efficiency of UAVs, but also effectively reduces the waste of resources, and has important theoretical and practical significance. In addition, this method should also consider the energy constraints and information transmission time of UAVs, as well as the signal coverage of different regions, to ensure the fairness and real-time nature of task allocation. By introducing advanced algorithms such as machine learning or artificial intelligence technology, the intelligent level of task allocation can be further improved, achieving more accurate and dynamic task scheduling. This will greatly help to improve the intelligent level of the entire power system and ensure the continuity and reliability of power supply. SUMMARY

[0006] Therefore, the application provides an unmanned aerial vehicle existing computing task allocation optimization method, aiming to reduce the total communication cost of the system and improve the operation efficiency and inspection quality of the power system by reasonably scheduling the task allocation of the unmanned aerial vehicle.

[0007] In order to achieve the above purpose, the application provides a resource allocation optimization method for unmanned aerial vehicle power inspection, comprising the following steps:

[0008] S1: the unmanned aerial vehicle performing the inspection task and the nearby edge server and ground base station jointly form a federated learning system;

[0009] S2: each unmanned aerial vehicle uploads the edge server parameters to the ground base station;

[0010] S3: the ground base station calculates the optimal computing load of each edge server according to the edge server parameters uploaded by the unmanned aerial vehicle;

[0011] S4: the ground base station calculates the edge server most suitable for transmission of each unmanned aerial vehicle as the optimal edge server transmission node, and broadcasts it to each unmanned aerial vehicle, and the unmanned aerial vehicle sends data to the edge server according to the obtained optimal edge server transmission node and optimal computing load;

[0012] S5: the ground base station broadcasts the initial model parameters to each edge server to start a new round of federated learning model training, and the unmanned aerial vehicle loads the data samples required for this round of training from the memory after receiving the broadcast global model, updates the local model using the gradient descent algorithm, and uploads the updated local model to the ground base station for weighted aggregation processing;

[0013] S6: the ground base station judges whether the model training reaches the optimal model coordination round number, if yes, stops the global model training, if not, continues to broadcast the aggregated model parameters to the edge servers participating in the federated learning training to start a new round of model training.

[0014] Further, the unmanned aerial vehicle serves as a participating client of the federated learning, for transmitting relevant data information obtained in the process of inspecting the power grid to the edge server for local model training processing; the ground base station serves as a central server, for broadcasting initial model parameters to each edge server and aggregating the model parameters uploaded by each edge server after completing local model training; the edge server is used for local model training based on relevant data information provided by the unmanned aerial vehicle inspection, and the result of the local model training is uploaded to the ground base station for model updating processing, so as to obtain new model parameters, and the new model parameters are sent to each edge server.

[0015] Further, the edge server parameters include data type f of edge server k, data set size n k , local model training round number e k , data volume size b required for each round of training k and position coordinates.

[0016] Further, step S3 specifically includes:

[0017] An empirical mapping function is set to represent the relationship between edge server parameters and computing load:

[0018]

[0019] wherein represents the computing load of edge server k, r represents the convergence requirement, and the above formula indicates that when performing federated learning model training, given data type, convergence requirement and related parameters n k , e k , b k and computing load, wherein n k represents the data set size of edge server k, e k represents the iteration number of edge server k in each round of model training, b k represents the data set size used by edge server k in each model training;

[0020] The problem of optimal computing load resource allocation scheme is expressed in the following form:

[0021]

[0022] wherein u(n) represents the average resource utilization of each edge server in the edge computing system, n represents a vector related to the number of data samples, n = <n k > k∈K , represents the maximum computing load supported by each edge server; K represents the set of all edge servers participating in federated learning, and it is assumed that there are always and wherein represents the maximum number of sample data that edge server k can receive and process;

[0023] The optimal result of each n k is obtained by calculating , denoted as Then, it is checked whether each satisfies the two constraint conditions, if not, adjust to meet the constraint condition;

[0024] Then update n using gradient descent algorithm k :

[0025]

[0026] where i0 represents the number of update iterations, and a represents the step size;

[0027] Until one of the following conditions is met, the iteration is stopped:

[0028] (1) the change of the objective function is less than a preset threshold;

[0029] (2) the maximum number of iterations is reached;

[0030] Thus, we get represent the data sample size that makes the computational resource utilization of the edge server k reach the optimal trade-off under the given model convergence requirement and the computational resource limit of the edge server k.

[0031] Further, step S4 specifically comprises:

[0032] S41: Based on the edge server parameters uploaded by the unmanned aerial vehicle in step S2, the ground base station calculates and represents the distance between the unmanned aerial vehicle with serial number i and each edge server with a fictitious sequence d ik ={d i1 ,d i2 ,d i3 ,...,d iK}, where d iK represents the distance from the unmanned aerial vehicle i to the edge server k;

[0033] S42: The ground base station obtains the computing power of each edge server and represents the computing power of each edge server with a fictitious sequence c k ={c1,c2,c3...,c K};

[0034] S43: Calculate the time required for the unmanned aerial vehicle i to transmit and process data to the corresponding edge server, which is specifically modeled as follows:

[0035]

[0036] where p k is a weight coefficient set by human, which aims to weigh the importance of the two parts of time, data transmission of the unmanned aerial vehicle and data processing analysis of the edge server, and the value range of p k is 0≤p k≤ 1; N0 represents the power size of the Gaussian noise, S(d ik ) represents the signal power of the signal changing with the distance d ik during the transmission process, B is the transmission bandwidth size possessed by the UAV, T ik represents the total time required for the UAV i to transmit data to the edge server k and for the edge server k to perform processing;

[0037] The following restrictive factors are considered when calculating T ik :

[0038] (a) If the UAV i selects the edge server k, then for other UAVs j, and j≠i, the same edge server k cannot be selected; and i, j∈I, I represents the set of all UAVs; specifically, if k=arg(min(T ik ) for the UAV i, then k≠arg(min(T jk )) for the UAV j;

[0039] (b) The computing load of each edge server needs to be within their processing capacity range to avoid overload and resource waste, which can be specifically represented as follows.

[0040]

[0041] Where S k,max represents the maximum data set size that can be processed by the edge server k;

[0042] (c) The data set size uploaded by the UAV may be limited by its own storage and processing capacity, so it is necessary to reasonably control the data set size transmitted by the UAV under the premise of ensuring the model training effect, which can be specifically represented as:

[0043]

[0044] Where n i,max represents the maximum data set size that can be collected by the UAV i;

[0045] (d) The battery endurance of the UAV limits its computing and communication capabilities when performing tasks, so it is necessary to find a balance between energy consumption and task execution, which can be specifically represented as follows:

[0046] E i ≥ E i,min

[0047] Where E i represents the total energy size possessed by the UAV i, and E i,min represents the minimum energy size required for the UAV to perform a task once;

[0048] The ground base station calculates the minimum T ik in k = arg(min(T ik )), and the minimum value of k is solved by using the interior point method, and the specific steps are as follows:

[0049] The target is to minimize the total time required for the unmanned aerial vehicle to transmit data to the edge server and perform processing, which is represented as:

[0050] f(k) = min(max∑ k∈K T ik )

[0051] Next, considering the above four restrictive factors (a)-(d), the inequality constraint is processed using the barrier function, which is in the following form:

[0052]

[0053] Where x ik is a binary decision variable of whether the unmanned aerial vehicle i selects the edge server k for transmission, and μ is a barrier parameter that controls the weight of the logarithmic term in the barrier function;

[0054] The above formula is solved by multiple iterations, and the value of μ is updated at the same time. When one of the following conditions is met, the algorithm can be terminated:

[0055] (1) The change of the objective function is less than the preset threshold;

[0056] (2) The violation degree of all constraints is less than the preset threshold;

[0057] (3) The maximum number of iterations is reached;

[0058] Once the algorithm is terminated, the optimal k * is output as the optimal solution, that is, the edge server k * is the best edge server transmission node;

[0059] S44: The ground base station broadcasts the determined best edge server transmission node to each unmanned aerial vehicle;

[0060] S45: According to the best edge server transmission node provided by the ground base station, each unmanned aerial vehicle starts to transmit data to the corresponding best edge server transmission node, and the size of the transmitted data is the best computing load of each edge server.

[0061] Compared with the prior art, the present application has the following advantages:

[0062] 1. Optimize resource utilization and improve efficiency: By assigning appropriate computing tasks to each edge server in federated learning, we can ensure that the computing resources of each device are fully utilized, avoiding problems such as resource waste. This method is particularly suitable for resource-constrained environments and can significantly improve the overall efficiency of the federated learning system. The computing power, communication speed, and storage capacity of client devices in federated learning vary greatly, which poses a great challenge to the overall performance of the model. Therefore, dynamically adjusting computing tasks to adapt to the computing power of different devices can improve the flexibility and efficiency of the system.

[0063] 2. Reduce communication costs and enhance data privacy protection: Since only model updates, not raw data, are uploaded to the central server, this method reduces the amount of data transmitted over the network, thereby reducing communication costs. At the same time, it also enhances data privacy protection because raw data does not leave the local device. This is particularly important when dealing with sensitive data and can improve users' trust in the federated learning system. As the search results show, in federated learning, data remains on the local client, and only model update information is exchanged with the central server, which does not directly leak private data, providing great help to data privacy protection.

[0064] 3. Improve the robustness and adaptability of the system: By reasonably distributing computing loads, we can reduce the problem of training interruption caused by device failure or network problems, thereby improving the robustness of the system. In addition, this method allows the system to adapt to different network conditions and device performance, enabling federated learning to be deployed in more diverse environments. This is also reflected in the search results, which show that federated learning needs to adapt to the situation where client devices may exit existing training due to network failures, computing power limitations, etc.

[0065] 4. Improve the intelligent level of the power system: By intelligently distributing computing tasks, we not only improve resource utilization efficiency and system robustness, but also reduce communication costs and enhance data privacy protection. This is of great significance to the application and development of federated learning systems. In particular, in the field of power inspection, the application of this technology can greatly improve the intelligent level of the power system, ensuring the continuity and reliability of power supply.

[0066] 5. Achieve precise and dynamic task scheduling: By introducing advanced algorithms such as machine learning or artificial intelligence technology, we can further improve the intelligent level of task allocation and achieve more precise and dynamic task scheduling. This will be very helpful in improving the intelligent level of the entire power system and ensuring the continuity and reliability of power supply. In this way, drones can flexibly adjust their computing loads according to real-time changes in the environment and task requirements, thereby achieving more efficient power inspection and maintenance work.

[0067] 6. Improve the real-time performance of task execution: By evaluating the computing power of edge servers and the task requirements of drones in real time, this technology can quickly make decisions and assign the most suitable amount of tasks to drones. This real-time performance is crucial for tasks that require quick response, such as emergency maintenance or accident response. Drones can quickly receive task assignments and immediately begin execution, greatly improving the power system's response speed and processing capacity in the face of emergencies.

[0068] In summary, the present application improves resource utilization efficiency and system robustness by intelligently allocating computing tasks, reduces communication costs and enhances data privacy protection, and has important significance for the application and development of federated learning systems. BRIEF DESCRIPTION OF DRAWINGS

[0069] Figure 1 A flowchart of a resource allocation optimization method for unmanned aerial vehicle power inspection according to an embodiment of the present application;

[0070] Figure 2 A flowchart of selecting the best edge server by the unmanned aerial vehicle according to an embodiment of the present application. DETAILED DESCRIPTION

[0071] To make the objectives, technical solutions, and advantages of the embodiments of the present application clearer, the technical solutions in the embodiments of the present application will be described below in conjunction with the accompanying drawings of the embodiments of the present application. Obviously, the described embodiments are only a part of the embodiments of the present application, rather than all the embodiments of the present application. Based on the embodiments in the present application, all other embodiments obtained by those of ordinary skill in the art without creative work fall within the scope of protection of the present application.

[0072] Please refer to Figure 1 The embodiment of the present application provides a resource allocation optimization method for unmanned aerial vehicle power inspection, which includes the following steps:

[0073] S1: The unmanned aerial vehicle performing the inspection task, the nearby edge server, and the ground base station jointly form a federated learning system. The unmanned aerial vehicle serves as a participating client of the federated learning and transmits relevant data information obtained during the power grid inspection process to the edge server for local model training. The ground base station serves as a central server of the federated learning and broadcasts initial model parameters to each edge server and aggregates the model parameters uploaded by each edge server after completing local model training, so as to complete model training; the edge server uses the relevant data information provided by the unmanned aerial vehicle inspection to perform local model training and uploads the results of the local model training to the central server for model updating, so as to obtain new model parameters and send the new model parameters to each edge server.

[0074] S2: Each UAV uploads edge server parameters to the ground base station before the start of federated learning training, including the data type f of edge server k, the data set size n k , the number of local model training rounds e k , the size of the data required for each round of training b k and position coordinates and other related parameters.

[0075] S3: The ground base station will evaluate and determine the optimal computing load that each edge server should bear according to the edge server parameters uploaded by each UAV. This process involves comprehensive consideration of the data processing capacity of the edge server, the size of the data set, the number of training rounds and other parameters to ensure that each server can achieve optimal allocation and utilization of resources without exceeding its performance limit.

[0076] Specifically, an empirical mapping function is set to represent the relationship between edge server parameters and computing load:

[0077]

[0078] wherein represents the computing load of edge server k, and r represents the convergence requirement. The above formula shows the relationship between the computing load and the given data type, convergence requirement and related parameters n k , e k , b k during federated learning model training, wherein n k represents the data set size of edge server k, e k represents the number of iterations of edge server k in each round of model training, and b k represents the data set size used by edge server k in each model training.

[0079] The problem of optimal computing load resource allocation scheme can be expressed as follows:

[0080]

[0081] wherein u(n) represents the average resource utilization of each edge server in the edge computing system, and n represents a vector related to the number of data samples, specifically n = <n k >. k∈K . represents the maximum computing load supported by each edge server; K represents the set of all edge servers participating in federated learning, and it is assumed that there is always and wherein represents the maximum number of sample data that edge server k can receive and process.

[0082] So, by calculating This will get the optimal result of each n k , denoted as Then check whether each satisfies Two constraint conditions. If not, you need to adjust to meet the constraint conditions.

[0083] Then use the gradient descent algorithm to update n k :

[0084]

[0085] Where i0 represents the number of update iterations, and a represents the step size.

[0086] Until one of the following conditions is met, stop iteration:

[0087] (3) The change of the objective function is less than the preset threshold.

[0088] (4) Reach the maximum number of iterations.

[0089] Thus, we can get This value represents the data sample size that makes the computational resource utilization of edge server k the optimal trade-off under the given model convergence requirements and the computational resource limitations of edge server k.

[0090] S4: The ground base station calculates the edge server most suitable for transmission for each UAV as the best edge server transmission node and broadcasts it to each UAV. The UAV sends data to the edge server according to the obtained best edge server transmission node and best computing load. The specific steps are as shown in Figure 2 :

[0091] S41: Based on the edge server parameters uploaded by the UAV in step S2, the ground base station calculates and represents the distance between UAV i and each edge server with a fictitious sequence d ik ={d i1 ,d i2 ,d i3 ,...,d iK}, where d iK represents the distance from UAV i to edge server k.

[0092] S42: The ground base station obtains the computing power of each edge server and represents the computing power of each edge server with a fictitious sequence c k ={c1,c2,c3...,c K};

[0093] S43: The ground base station calculates the edge node most suitable for data transmission for each unmanned aerial vehicle according to the uploaded relevant information data.

[0094] Specifically, the time required for the unmanned aerial vehicle i to transmit and process data to the corresponding edge server is calculated, and the modeling is as follows:

[0095]

[0096] where p k is a weight coefficient set by man, aiming to balance the importance of the two parts of time for data transmission by the unmanned aerial vehicle and data processing and analysis by the edge server, and the value range of p k is 0≤p k ≤1. N0 represents the power size of Gaussian noise, S(d ik ) represents the signal power of the signal changing with the distance d ik during transmission, B is the transmission bandwidth size of the unmanned aerial vehicle, and T ik represents the total time required for the unmanned aerial vehicle i to transmit data to the edge server k and for the edge server k to process the data.

[0097] In addition, many restrictive factors are considered when calculating T ik , which are as follows:

[0098] (1) The data type collected by the unmanned aerial vehicle needs to meet the requirements of the corresponding edge server processing, and the data quality needs to reach a certain standard to ensure the effectiveness of model training.

[0099] Suppose D ik represents the data type and quality of the data set collected by the unmanned aerial vehicle i at the edge server k. It needs to be ensured that D ik meets the processing requirements of the edge server k:

[0100] D ik ∈D k

[0101] where D k represents the set of data types and qualities that can be processed by the edge server k.

[0102] (2) If the unmanned aerial vehicle i selects the edge server k, then for other unmanned aerial vehicles j, and j≠i, the same edge server k cannot be selected. And i,j∈I, I represents the set of all unmanned aerial vehicles. Specifically, if for the unmanned aerial vehicle i, k=arg(min(T ik )), then for the unmanned aerial vehicle j, k≠arg(min(T jk )).

[0103] (3) The computational load of each edge server needs to be within their processing capacity to avoid overload and resource waste, which can be represented as follows.

[0104]

[0105] where S k,max represents the maximum data set size that edge server k can handle.

[0106] (4) The data set size uploaded by the UAV may be limited by its own storage and processing capacity, so it is necessary to reasonably control the data set size transmitted by the UAV while ensuring the effectiveness of model training, which can be represented as:

[0107]

[0108] where n i,max represents the maximum data set size that UAV i can collect.

[0109] (5) The battery endurance of the UAV limits its computing and communication capabilities when performing tasks, so it is necessary to find a balance between energy consumption and task execution, which can be represented as follows:

[0110] E i ≥ E i,min

[0111] where E i represents the total energy size of UAV i, E i,min represents the minimum energy size required for the UAV to perform a task.

[0112] (6) During data transmission and model updating, the security and privacy protection of data need to be ensured to avoid the leakage of sensitive information.

[0113] Use an indicator function to express whether the data is leaked or not, which can be represented as follows:

[0114] M(data leakage) = 0

[0115] where M is an indicator function, taking the value of 1 when there is data leakage and 0 when there is no data leakage.

[0116] The ground base station calculates the minimum T ik for UAV i, where k = arg(min(T ik )), and then uses the interior point method to solve the minimum value of k, with the following specific steps:

[0117] The goal is to minimize the total time required for the UAV to transmit data to the edge server and perform processing, which can be represented as:

[0118] f(k) = min(max∑ k∈K T ik )

[0119] Next, consider the restrictive factors in the above six points, because the inner product method mainly considers inequality constraints, so it focuses on the remaining inequality factors except the first and sixth factors. The inequality constraints will be handled using a barrier function, which is in the following form:

[0120]

[0121] Where x ik is a binary decision variable of whether the UAV i chooses the edge server k for transmission. μ is a barrier parameter that controls the weight of the logarithmic term in the barrier function, which gradually decreases as the algorithm progresses, making it closer and closer to the boundary of the feasible region.

[0122] The above formula is solved by multiple iterations, and the value of μ is updated at the same time. When one of the following conditions is met, the algorithm can be terminated:

[0123] (1) The change of the objective function is less than the preset threshold.

[0124] (2) The violation degree of all constraints is less than the preset threshold.

[0125] (3) The maximum number of iterations is reached.

[0126] Once the algorithm is terminated, the optimal k * is output as the optimal solution, that is, the edge server k * is the best edge server transmission node.

[0127] S44: The ground base station can select the most suitable edge server for each UAV to transmit data by the above method, and broadcast the result to each UAV.

[0128] S45: According to the best edge server transmission node provided by the ground base station, each UAV starts to transmit data to the corresponding best edge server transmission node, and the size of the transmitted data is the best computing load of each edge server.

[0129] S5: The ground base station determines the initial parameter size w0 of the model, and broadcasts the result to each edge server. Each edge server updates w0 using the gradient descent algorithm, and then transmits the updated result to the ground base station for weighted aggregation processing.

[0130] Specifically, for the edge server with serial number k, the local model update formula is:

[0131]

[0132] wherein, is the local model parameter of the e+1th iteration, is the local model parameter of the e th iteration, η k is the learning rate of the model, is the gradient size at the e th iteration.

[0133] S6: The ground base station determines whether the model training coordination round has been reached. If the ground base station has reached the initial set total model coordination training period, it will send a broadcast to the unmanned aerial vehicle and each edge server, instructing them to stop training and save the current global model parameters. If not, the weighted processed results will be broadcast again to each edge server for the next round of model training.

[0134] The present application not only improves the work efficiency of the unmanned aerial vehicle, but also effectively reduces the waste of resources, and has important theoretical and practical significance. In addition, this method should also consider the energy limitation and flight time of the unmanned aerial vehicle, as well as the signal coverage situation of different areas, to ensure the fairness and real-time performance of task allocation. By introducing advanced algorithms such as machine learning or artificial intelligence technology, the intelligent level of task allocation can be further improved, and more accurate and dynamic task scheduling can be achieved. This will help to improve the intelligent level of the entire power system, and ensure the continuity and reliability of power supply.

[0135] The above is only a specific embodiment of the present application, but the protection scope of the present application is not limited thereto. Any changes or replacements within the technical scope disclosed by the present application can be easily thought of by any person skilled in the art, and should be covered within the protection scope of the present application. Therefore, the protection scope of the present application should be subject to the protection scope of the claims.

Claims

1. A resource allocation optimization method for unmanned aerial vehicle (UAV) power line inspection, characterized in that: Includes the following steps: S1: The drones performing inspection tasks will be combined with nearby edge servers and ground base stations to form a federated learning system; S2: Each drone uploads edge server parameters to the ground base station; S3: The ground base station calculates the optimal computing load for each edge server based on the edge server parameters uploaded by the drone; S4: The ground base station calculates the most suitable edge server for each drone to transmit data and selects the best edge server transmission node. The base station then broadcasts this information to each drone. The drones then send data to the edge server based on the best edge server transmission node and the best computing load. S5: The ground base station broadcasts the initial model parameters to each edge server to start a new round of federated learning model training. After receiving the broadcast global model, the UAV loads the data samples required for this round of training from the memory, updates the local model using the gradient descent algorithm, and uploads the updated local model to the ground base station for weighted aggregation processing. S6: The ground base station determines whether the model training has reached the optimal number of model coordination rounds. If it has, the global model training is stopped. If it has not, the aggregated model parameters are broadcast to the edge servers participating in federated learning training to start a new round of model training. Step S3 specifically includes: Define an empirical mapping function to represent the relationship between edge server parameters and computational load: ； in Let represent the computational load of edge server k, and r represent the convergence requirement. The above formula illustrates that when training a federated learning model, given the data type, convergence requirement, and related parameters... , , The relationship with computational load, where This represents the size of the dataset for edge server k. This represents the number of iterations for edge server k during each round of model training. This indicates the size of the dataset used by edge server k during each model training session; The problem of optimal resource allocation for computational loads can be formulated as follows: ； in, This represents the average resource utilization of each edge server in the edge computing system, where n represents a vector related to the number of data samples. This indicates the maximum computing load supported by each edge server; Let represent the set of all edge servers participating in federated learning, and assume here that there always exists . ,and ,in Represents edge server The maximum number of sample data that can be received and processed; Through calculation Get each The optimal result is denoted as Next, check each one. Does it meet the requirements? , If these two constraints are not met, then adjust... To satisfy the constraints; Then update using gradient descent algorithm. : ； in Indicates the number of update iterations. Indicates the step size; The iteration stops when one of the following conditions is met: (1) The change in the objective function is less than the preset threshold; (2) Reach the maximum number of iterations; Therefore, we obtain , This represents the data sample size that achieves the optimal trade-off between the given model convergence requirements and the computational resource constraints of edge server k. The drone, acting as a client in federated learning, transmits relevant data obtained during power grid inspections to edge servers for local model training. The ground base station, acting as a central server, broadcasts initial model parameters to each edge server and aggregates the model parameters uploaded by each edge server after completing local model training. The edge servers use the relevant data provided by the drone inspections for local model training, upload the training results to the ground base station for model update processing to obtain new model parameters, and then send the new model parameters to each edge server. The edge server parameters include the data type of edge server k. Dataset size Number of training rounds for local models The amount of data required for each training round and position coordinates.

2. The resource allocation optimization method for UAV power line inspection as described in claim 1, characterized in that: Step S4 specifically includes: S41: Based on the edge server parameters uploaded by the drone in step S2, the ground base station calculates and uses a fictitious sequence. To indicate the serial number is The distance between the drone and each edge server, among which Indicates drone Distance to edge server k; S42: The ground base station obtains the computing power of each edge server and uses a fictitious sequence. The computing power of each edge server is represented; S43: Computational Drone The time required to transmit and process data to the corresponding edge server is modeled as follows: ； in The manually set weighting coefficients are designed to balance the importance of the time spent transmitting data from the drone and the time spent processing and analyzing data on the edge server. The range of values ​​is ; This indicates the power of the Gaussian noise. This indicates that the signal changes with distance during transmission. The changed signal power, The amount of transmission bandwidth possessed by the drone. Indicates drone To edge servers Data transmission and edge servers The total time required for processing; In calculation The following limiting factors should be considered: (a) If the drone Edge server selected For other drones ,and You cannot select the same edge server. ;and , This represents the collection of all drones; specifically, if we consider drones... To be honest So, for drones In other words, ; (b) The computational load of each edge server needs to be within its processing capacity to avoid overload and resource waste, as specifically stated below: ； in Represents edge server The maximum dataset size that can be processed; (c) The size of the dataset uploaded by the drone may be limited by its own storage and processing capabilities. Therefore, it is necessary to reasonably control the size of the dataset transmitted by the drone while ensuring the model training effect. Specifically, it can be expressed as follows: ； in Indicates drone The maximum dataset size that can be collected; (d) The battery life of a drone limits its computing and communication capabilities during mission execution. Therefore, a balance needs to be struck between energy consumption and mission execution, as detailed below: ； in Indicates drone The total amount of energy possessed This indicates the minimum amount of energy required for the drone to perform a single mission; Ground base station calculates drone The corresponding minimum In At this point, the interior point method is used to... The minimum value is found by the following steps: The goal is to minimize the total time required for the drone to transmit and process data to the edge server, expressed as: Next, considering the four restrictive factors (a)-(d) above, we use barrier functions to handle the inequality constraints, in the following form: ； in It is a drone Should we select an edge server? The binary decision variables to be transmitted These are the barrier parameters, which control the weights of the logarithmic terms in the barrier function; The above formula is solved through multiple iterative operations, and the results are updated simultaneously. The algorithm can terminate when one of the following conditions is met: (1) The change in the objective function is less than the preset threshold; (2) The degree of violation of all constraints is less than the preset threshold; (3) Reach the maximum number of iterations; Once the algorithm terminates, the optimal output is reached. As the optimal solution, that is, the edge server at this time For optimal edge server transmission nodes; S44: The ground base station broadcasts the determined optimal edge server transmission node to each drone; S45: Based on the optimal edge server transmission node provided by the ground base station, each UAV begins to transmit data to the corresponding optimal edge server transmission node. The amount of data transmitted is the optimal computing load of each edge server.

Images