Service unloading and time delay optimization method based on 5G edge computing

Abstract

The invention belongs to the technical field of intelligent power grid power business terminal task calculation, and comprises a business unloading and delay optimization method based on 5G edge calculation, and the method comprises the following steps: S1, constructing a cloud edge-end collaborative architecture; s2, obtaining a power service terminal set and a corresponding mixed type key service; s3, constructing an edge node communication time delay model and a power business calculation time delay model; s4, establishing a joint optimization model containing a wireless network resource allocation and task unloading strategy; obtaining a unique decision solution of wireless transmission network resources; s5, setting a penalty function; and S6, defining a target optimization function as a composite function for minimizing the total time delay of the system, and solving an optimal decision # imgabs0 # for unloading the power terminal and the key service. According to the method, the central cloud is introduced as an auxiliary, and the penalty function is set according to different key levels of the services on the power service terminal, so that the unloading success rate of the services with higher key levels is improved.

Description

Technical Field

[0001] The present invention belongs to the technical field of terminal task computing for smart grid electric power services, and specifically relates to a service offloading and latency optimization method based on 5G edge computing. Background Art

[0002] In the scenario of smart grid, task calculations on power terminal devices will have latency requirements, security requirements, business reliability requirements, etc. At present, power system business unloading mainly relies on three types of traditional technical solutions: allocating resources based on artificially preset fixed QoS levels, which cannot adapt to high dynamic load scenarios such as sudden fault recording, resulting in latency fluctuations of key services such as relay protection up to ±15ms; equally allocating time slices or bandwidth to terminal devices, ignoring differences in business criticality. Actual measurements of a provincial power grid show that this method has caused the bandwidth utilization rate of 5G slices to be lower than 55% for a long time, and meter reading services have frequently occupied relay protection resources; using traditional intelligent algorithms such as genetic algorithms and particle swarm optimization to solve, there are two major defects - the calculation time exceeds 100ms, which cannot meet the μs-level business response requirements; the probability of falling into local optimality is as high as 32%.

[0003] Therefore, whether it is possible to provide a method for service unloading and latency optimization of hybrid key services for power cloud-edge collaborative scenarios based on the latency optimization and resource allocation optimization theory under mobile edge computing applications assisted by the central cloud, combined with the characteristics of hybrid services with different priorities of each power terminal device, is a technical problem that needs to be urgently solved in the present invention. Summary of the invention

[0004] In view of this, the present invention provides a service offloading and latency optimization method based on 5G edge computing.

[0005] In order to solve the above technical problems, the technical solution adopted by the present invention is: A method for service offloading and latency optimization based on 5G edge computing, the method is implemented by the following steps: Step S1: construct a three-level cloud-edge-end collaborative architecture, including a power business terminal device layer, an MEC node layer, and a central cloud layer, wherein the power business terminal device layer is connected to the MEC node layer through a wireless network, and the MEC node layer is connected to the central cloud layer through a high-bandwidth backhaul link; Step S2: Obtain the electric power business terminal set N={1,2,...,n} and the hybrid key business (c n ,λ n , τ n , ε n ), for penalty function and target optimization function to call; where c nThe actual transmission data size required for offloading services of a unit in a wireless network, λ n The average distribution number of single transmission requests for power terminal devices, τ n The maximum tolerable delay for hybrid power services, ε n Represents the priority of the hybrid task; Step S3: Construct an edge node communication delay model and a power service calculation delay model; Step S4: Establish a joint optimization model including wireless network resource allocation and task offloading strategy. The model aims to minimize the total system delay and is a convex optimization problem with equality constraints and inequality constraints. Use the KKT conditions and the Lagrange multiplier method to solve the convex optimization problem, minimize the queuing delay and transmission delay, and obtain the unique decision solution of the wireless transmission network resources by removing some inequality constraints; Step S5: Based on the task computation volume and delay requirements, adopt a differential offloading decision mechanism, set a penalty function, and improve the offloading success rate of high-priority tasks; Step S6: Based on the edge node communication model and the power service calculation delay model, define the objective optimization function as a composite function that minimizes the total system delay, and solve the optimal decisions for power terminals and critical service offloading 。

[0006] Furthermore, in step S3, construct a communication delay model and perform constraints on the total average delay r of all power service terminal offloading calculation tasks: , where r n Is the average delay of the offloading task of the nth power service terminal; the wireless transmission bandwidth b of each power service terminal n Satisfies the relationship with the total transmission bandwidth B 。

[0007] Furthermore, in step S3, construct a power service calculation delay model and perform constraints on the transmission delay, the average calculation delay of the mobile edge service node, the average calculation delay of the central cloud, and the round-trip transmission delay when the MEC executes tasks to the cloud: Transmission delay , where the service rate of the wireless network to power terminal n is , and the system decision-maker needs to satisfy ; Average calculation delay of the mobile edge service node , where Is the total average rate of the mobile edge server side to execute the offloading task, Satisfies ; Average calculation delay of the central cloud , where is the total average rate for the central cloud to execute task offloading, is the maximum computing resource that the central cloud service instance can provide for this system, Satisfy ; The round-trip transmission delay when the MEC executes tasks to the cloud is .

[0008] Furthermore, in step S4, the method for establishing a joint optimization model including wireless network resource allocation and task offloading strategy includes: where and are both Lagrange multipliers, is the inequality constraint in the optimization problem, is the equality constraint condition, defines the total wireless network transmission resource and the total amount of power terminal user task offloading in the cloud-edge collaboration system.

[0009] Furthermore, in step S4, after substituting part of the wireless resource scheduling inequalities into the joint optimization model, the wireless transmission network resource scheduling decision represented by is obtained: : ; The unique decision solution of the wireless transmission network resource: , where represents an infinitesimal positive number. When , is negative infinity. At this time, is negative; when is in the interval , the obtained is in the interval ), meeting the actual requirements.

[0010] Furthermore, in step S5, based on the task computation amount and delay requirements, a differentiated offloading decision mechanism is adopted, and a penalty function is set to improve the offloading success rate of higher-priority services: ; where, represents the penalty coefficients of different priorities, represents the execution delay of the task. When the task is executed at the edge node , when the MEC cannot meet the delay requirements of the smart grid device side, the task is then offloaded to the central cloud, then .

[0011] Furthermore, the steps for constructing the objective optimization function in step S6 include: S61. Construct a weighted response delay model for mobile edge computing assisted by a central cloud: , Among them, the response delay of the mobile edge computing system assisted by the central cloud consists of five parts, namely, the wireless transmission delay of the power terminal device to offload tasks, the task execution delay of the mobile edge computing node, the transmission delay from the mobile edge node to the central cloud, the task execution delay of the central cloud, and the calculation result feedback delay; S62. Construct an optimization objective function for the offloading decision of hybrid critical services for multiple power terminals according to the execution success rate of hybrid critical tasks and the weighted total delay of uploading and computing offloading: , where, is the time weighted sum of each link, is the offloading success rate of the power hybrid service.

[0012] Compared with the prior art, the beneficial effects of the present invention are as follows: The algorithm mentioned in the present invention can dynamically adjust the resource allocation strategy according to the real-time task offloading situation of mobile devices, schedule the computing tasks offloaded by mobile devices, and add the key level setting of tasks, so the effect is better; When the number of mobile devices is large, the average mobile edge response delay of the system is reduced significantly, mainly because the computing tasks offloaded by different mobile devices are different, and when the number of mobile devices is large, the differences in offloaded computing tasks are also obvious; In addition, the above optimization method can also reasonably and effectively allocate resources according to the wireless network resources and mobile edge cloud computing resources in the existing environment and the situation of mobile device offloaded computing tasks, and decide the offloading location of mobile device offloaded tasks;; By introducing a central cloud as an assistant, the present invention constructs a solution for differential offloading decision-making for different resource requirements and delay-sensitive tasks according to the different key levels of services on power service terminals. In the smart grid scenario, each power service terminal has a hybrid service. Therefore, a penalty function is introduced to improve the offloading success rate of higher key level services, and a delay optimization offloading algorithm of an improved Lagrange multiplier algorithm is proposed to improve the overall efficiency of the system. BRIEF DESCRIPTION OF THE DRAWINGS

[0013] The present invention will be further described in detail below with reference to the accompanying drawings.

[0014] Figure 1 : Schematic diagram of the cooperation framework in the present invention.

[0015] Figure 2 : Schematic diagram of the simulation program through the Matlab simulation platform in the present invention.

[0016] Figure 3 : Schematic diagram of simulation parameter setting in the embodiment of the present invention.

[0017] Figure 4 : Schematic diagram of average moving edge response delay in the embodiment of the present invention.

[0018] Figure 5 : Schematic diagram of the influence of different critical levels on service offloading in the present invention. Detailed implementation manners

[0019] To better understand the present invention, the content of the present invention will be further clearly elaborated below in conjunction with embodiments and drawings. However, the protection scope of the present invention is not limited to the following embodiments only. In the following description, a large number of specific details are given to provide a more thorough understanding of the present invention. However, it is obvious to those skilled in the art that the present invention can be implemented without one or more of these details. In the following description, a large number of specific details are given to provide a more thorough understanding of the present invention. However, it is obvious to those skilled in the art that the present invention can be implemented without one or more of these details. In the following description, a large number of specific details are given to provide a more thorough understanding of the present invention. However, it is obvious to those skilled in the art that the present invention can be implemented without one or more of these details. For those skilled in the art, it is obvious that the present invention can be implemented without one or more of these details.

[0020] Embodiment 1: Refer to Figure 1 , a service offloading and delay optimization method based on 5G edge computing in this embodiment includes the following steps: Step S1, construct a three-level cloud-edge-end collaborative architecture, including a power service terminal device layer, an MEC node layer, and a central cloud layer, where the power service terminal device layer is connected to the MEC node layer through a wireless network, and the MEC node layer is connected to the central cloud layer through a high-bandwidth backhaul link; Step S2, obtain a power service terminal set N = {1, 2,..., n} and the hybrid critical services (c n , λ n , τ n , ε n ) of each power service terminal user for the penalty function and the target optimization function to call. In this step, N represents that the system consists of N power service terminal devices, and each power service terminal has a hybrid critical service; Where c n is the actual transmission data size required for the wireless network to transmit a unit of offloaded service; when the number of CPU cycles required for each transmission request of the power terminal device follows a negative exponential distribution, λ n is the distribution average of a single transmission request of the power terminal device; τ n is the maximum tolerable delay for the hybrid power service, and ε n represents the priority of the hybrid task; Step S3, construct an edge node communication delay model and a power service calculation delay model; ​Step S4: Establish a joint optimization model that includes wireless network resource allocation and task offloading strategies. The model aims to minimize the total system delay and is a convex optimization problem with equality constraints and inequality constraints. Use the KKT conditions and the Lagrange multiplier method to solve the convex optimization problem, minimize the queuing delay and transmission delay, and obtain a unique decision solution for the wireless transmission network resources by removing some inequality constraints. Step S5: Based on the task computation amount and delay requirements, adopt a differentiated offloading decision mechanism, set a penalty function, and improve the offloading success rate of high-priority tasks. Step S6: Based on the edge node communication model and the power service computation delay model, define the objective optimization function as a composite function that minimizes the total system delay, and solve the optimal decisions for power terminals and critical service offloading. 。

[0021] Refer to Figure 1 In step S1, the network architecture is an integrated architecture composed of a cloud, MEC, and power service devices. Since the network scenario consists of many power service devices, MEC servers, and cloud servers. The cloud server, also known as the data center, has powerful computing capabilities and a large storage capacity. The cloud server conducts in-depth data analysis and processing based on users' service requests. The MEC server is a mobile edge computing server located at the edge layer. It can support data refinement and processing by deploying some hardware devices at the network edge, thereby realizing local data computing. Numerous power service devices, such as distribution automation terminals, monitoring terminals, and acquisition terminals, are deployed at the bottom of the network. Some data collection devices can also act as power application terminals to send requests to the cloud server.

[0022] In the present invention, the system decision maker located at the mobile edge service node will, according to the pre-estimated task offloading strategy, decide the resource allocation of the wireless transmission network and the task offloading location of the power terminal. The power terminal device will first offload the computing task to the edge service node for computing. When the mobile edge service node cannot meet the requirements of the terminal device, the computing task will be offloaded to the central cloud for execution. The scenario discussed in the present invention is a situation where a mobile edge computing node relies on a central cloud.

[0023] In step S3, construct a communication delay model and impose a constraint on the total average delay r of all power service terminals offloading computing tasks: , where r n is the average delay of the nth power service terminal offloading the task; the wireless transmission bandwidth b n of each power service terminal and the total transmission bandwidth B satisfy .

[0024] Specifically, for each edge node, the communication module and the computing module are usually separate. Therefore, the calculation of part of the data in mobile edge computing can be carried out simultaneously with the data transmission from the edge to the cloud. In addition, edge servers are all connected to cloud servers through backhaul links, and these backhaul links all have high bandwidths. In practical applications, the backhaul links are shared by users, and due to factors such as the randomness of packet arrival, resource scheduling for multiple users, complex routing algorithms, etc., it is very difficult to model the overall delay of the system. The present invention formulates an optimal cooperation strategy for cloud computing and edge computing based on the above problems, assuming that the resource scheduling strategy and the routing algorithm are fixed.

[0025] Assume that the arrival rate of computing requests on each power terminal device follows a Poisson distribution. The average delay for the nth power terminal user to offload tasks can be expressed as , where r represents the total average delay for all user terminals to offload computing tasks, then it is: .

[0026] The total wireless network bandwidth available to power devices in a wireless network system is fixed. The total transmission bandwidth can be set as B. The system decision-maker will allocate the wireless transmission bandwidth of each device to each power terminal, satisfying: . In step S3, a power service computing delay model is constructed, and constraints on transmission delay, average computing delay of the mobile edge service node, average computing delay of the central cloud, and round-trip transmission delay when the MEC executes tasks to the cloud are carried out.

[0027] According to the application theory of queuing theory, for a power terminal device n ( ), the average transmission delay in the network is: , where the service rate of the wireless network for power terminal n ( ) is , and the system decision-maker needs to satisfy .

[0028] In the system studied in this application, a mobile edge computing service node scenario is considered, and the resources of each mobile edge node are limited. When the total transmission bandwidth is fixed, assume that the maximum computing resources that the mobile edge node can provide is D. The total average rate of executing offloading tasks at the mobile edge server side is represented by , according to the scenario application of the actual smart grid, satisfies . Therefore, the average computing delay of the mobile edge service node is: .

[0029] In the smart grid scenario, the resources of mobile edge nodes are often limited. This invention uses a central cloud to assist in the computing offloading of power terminal services to expand the computing resources of mobile edge nodes. Power terminal users can offload tasks to the central cloud through mobile edge nodes. Then, according to the relevant theory of queuing theory, the average computing delay of the central cloud can be obtained: , where is the total average rate of the central cloud to execute task offloading, is the maximum computing resource that the central cloud service instance can provide for this system, satisfies .

[0030] In the scenario of central cloud-assisted smart grid task computing offloading, power terminals often give priority to offloading to the MEC server for execution. When the computing resources of mobile edge nodes cannot meet the delay requirements of power terminals, the computing tasks of the load are then offloaded to the cloud server for execution. Then, the round-trip transmission delay when the MEC executes tasks to the cloud needs to be taken into account. In this invention, the round-trip transmission delay when the MEC executes tasks to the cloud is set as .

[0031] The above analyzes the optimization problem of the power hybrid service computing offloading proposed by this invention, and models this problem based on the theory of queuing theory. It can be found that the optimization problem of this model is a convex optimization problem. To solve this problem, this application describes a cloud-edge collaborative power hybrid service offloading and delay optimization algorithm. This algorithm is divided into two aspects: (1) First, decide the optimal wireless network allocation resources and power terminal service offloading strategy; (2) After introducing a penalty function to improve the offloading success rate of higher-priority services, finally obtain the optimal decision for power terminal and critical service offloading.

[0032] In step S4, first, the system decision-maker decides on the scheduling of wireless network resources and the MEC or central cloud task offloading strategy of power service terminals. Without setting the service critical level, the wireless network resource scheduling and the edge server and central cloud offloading strategy of power service terminals is still a convex optimization problem, and includes equality constraints and inequality constraint conditions. The optimization objective is . This algorithm uses the KKT condition and the Lagrange multiplier algorithm to decide the optimal wireless network allocation resources and power terminal service offloading strategy. If F(x) is a convex optimization problem, then the KKT condition and the Lagrange multiplier algorithm are used to help solve this optimal decision problem. First, establish a joint optimization model method that includes wireless network resource allocation and task offloading strategy, and express it in the form of a Lagrange equation as follows: where and are both Lagrange multipliers, is the inequality constraint in the optimization problem, is the equality constraint condition, defines the total wireless network transmission resources and the total amount of power terminal user task offloading in the cloud-edge collaboration system.

[0033] According to the convex optimization theory, to solve this convex optimization problem, we only need to find , which satisfies the KKT constraint conditions, then this solution is the optimal solution of this optimization problem. Thus, the above formula can be transformed into: To solve the above problem, first remove some inequality constraints. Since the decision result of the wireless transmission network resource scheduling through the inequality is unique, so first remove other inequality constraints and retain this inequality condition. After substituting the partial inequality of the wireless resource scheduling, the wireless transmission network resource scheduling decision represented by can be obtained : .

[0034] Substitute the above formula into the equality constraint , and the solution of the Lagrange multiplier can be obtained: .

[0035] Thus, the wireless resource scheduling strategy of cloud-edge collaborative mobile edge computing can be obtained. Next, remove the remaining inequalities, and the solution equation becomes the following form: .

[0036] Thus, we get ; .

[0037] From the above formula, it can be seen that and both increase monotonically with as the variable, then decreases monotonically with as the variable. The algorithm design linearly restricts ([[]] , ) in the interval, and then uses to obtain , that is, the unique decision solution of the wireless transmission network resources: , where represents an infinitesimal positive number. When , is negative infinity, and at this time is negative; when In the interval when, the obtained In the interval ), it meets the actual requirements.

[0038] In step S5, in the actual smart grid scenario, there are hybrid services with different priorities on each different device. Based on the task computing volume and latency requirements, a differential offloading decision mechanism is adopted, and a penalty function is set to improve the offloading success rate of higher-priority services: ; Among them, represents the penalty coefficients of different priorities, represents the execution latency of the task. When the task is executed at the edge node , when the MEC cannot meet the latency requirements of the smart grid device side, the task is then offloaded to the central cloud, then .

[0039] According to the above, a weighted response latency model of mobile edge computing assisted by the central cloud is constructed.

[0040] The steps of constructing the objective optimization function in step S6 include: S61. Construct a weighted response latency model of mobile edge computing assisted by the central cloud: .

[0041] Among them, the response latency of the mobile edge computing system assisted by the central cloud consists of five parts, namely, the wireless transmission latency of the power terminal device to offload the task, the task execution latency of the mobile edge computing node, the transmission latency from the mobile edge node to the central cloud, the task execution latency of the central cloud, and the latency of the calculation result backhaul. Latency, as an important measurement index in QoS, is worthy of great attention. Therefore, the mobile edge computing framework system assisted by the central cloud constructed in this application not only linearly weights the latency generated by the modeling, but also considers the response latency of the weighted mobile edge computing system on this basis.

[0042] S62. Construct an optimization objective function for the offloading decision of hybrid critical services for multiple power terminals according to the execution success rate of the hybrid critical tasks and the weighted total latency of uploading and computing offloading: , where is the time weighted sum of each link, It is the offloading success rate of the power hybrid service. The above assumes that each power end-user has a hybrid service with different priorities. The power terminal device sends a request for a specific service to the MEC server. The MEC server pre-estimates the response delay for offloading to the cloud or edge node. The system decision-maker makes a decision based on the pre-estimated situation and determines the task offloading method according to the decision result. Based on the above process, the optimization goal of the present invention is to improve the offloading success rate of higher-priority services and optimize the total system delay on this basis.

[0043] After introducing a penalty function to improve the offloading success rate of higher-priority services, solve the optimal decision for power terminal and critical service offloading 。

[0044] In a cloud-assisted mobile edge computing system, the resources of mobile edge nodes are often limited, which makes the mobile edge server unable to provide specific services for the specific needs of all tasks. Setting a penalty function can improve the offloading success rate of specific needs, thereby improving the user's QoE. When solving the optimal infinite resource scheduling scheme, the penalty function varies depending on whether it is offloaded to the cloud or the edge. Therefore, in this application, it is required to solve the offloading strategy of hybrid services under different critical levels.

[0045] In this application, the priority of the task is set as , where LO < HI, and the priority services are randomly allocated proportionally. The maximum tolerable delay takes values of , and the maximum tolerable delay is cut into different intervals to divide it into delay-sensitive services, delay-tolerant services, and other types of services.

[0046] Compared with most current solutions that introduce a fixed weight factor to represent user preferences, the present invention introduces a penalty function to dynamically improve the offloading success rate of higher-priority services in combination with delay parameters, thereby improving the quality of service of users. After introducing the penalty function , according to the solution provided above, solve the target optimization function of this application to obtain the optimal strategy 。

[0047] Embodiment 2: Refer to Figures 2 - 5 , in this embodiment, the algorithm proposed by the present invention is experimentally verified through the Matlab simulation platform. The delay simulation results include the influence of the number of power service terminals, the mobile edge computing response delay, the size of the input data of power equipment, and the division of service critical levels, and are compared with the following three methods: 1. Resource allocation optimization algorithm; 2. Resource equal allocation algorithm; 3. Resource random allocation calculation algorithm. The specific parameter settings are as Figure 3 shown.

[0048] Figure 4 It shows the average mobile edge response delay under different numbers of mobile devices. The number of mobile devices in this simulation experiment ranges from [10, 60], and the changing step size is 5. When the number of mobile devices is small, the experimental results of the method proposed in the present invention and the three comparison algorithms are not significantly different, that is, the impact on the system average mobile edge response delay is relatively small. As the number of mobile devices increases, the overall average mobile edge response delay of the system shows an upward trend. When the number of mobile devices is higher than 40, the system average mobile edge response delay under the algorithm described in this application is significantly lower than the other three comparison algorithms. When the number of mobile devices is around 60, the algorithm described in this application can reduce the system average mobile edge response delay by nearly 50%.

[0049] Compared with the resource allocation optimization calculation algorithm, the algorithm proposed in this application dynamically adjusts the resource allocation strategy according to the real-time task offloading situation of mobile devices, schedules the computing tasks offloaded by mobile devices, and adds the setting of the critical level of tasks, so the effect is better.

[0050] Compared with the equal resource allocation calculation algorithm, when the number of mobile devices is large, the system average mobile edge response delay is reduced more, mainly because the computing tasks offloaded by different mobile devices are different, and when the number of mobile devices is large, the difference in offloaded computing tasks is also more obvious.

[0051] Compared with the resource random allocation calculation algorithm, the algorithm proposed in this application can reasonably and effectively allocate resources according to the wireless network resources, mobile edge cloud computing resources in the existing environment and the situation of mobile device offloaded computing tasks, and decide the offloading location of mobile device offloaded tasks, so the effect is better than the resource random allocation calculation algorithm.

[0052] Refer to Figure 5 which shows the influence of different critical levels on service offloading. Among them , , . When the system resources are limited and often cannot meet the delay requirements of all tasks, setting a penalty function can improve the success rate of tasks. It can be seen from the above figure that as the penalty coefficient increases continuously, the success rate of tasks at the HI critical level also increases. When , the higher critical tasks cannot all be completed before the maximum tolerance delay, but after adjusting the penalty coefficient, it can be seen that this scheme can significantly improve the success rate of high critical tasks. When , as the penalty coefficient increases continuously, the success rate of high critical tasks can reach 100%. And when the maximum tolerance delay , it indicates that the tasks are not delay-sensitive, so all tasks can be completed before the maximum tolerance delay.

[0053] This application protects a service offloading and latency optimization method based on 5G edge computing. By introducing the central cloud as an auxiliary, according to the different critical levels of services on the power service terminals, a solution for making differential offloading decisions for tasks with different resource requirements and latency sensitivity is constructed. In the smart grid scenario, each power service terminal has a mixed service. Therefore, a penalty function is introduced to improve the offloading success rate of services with higher critical levels, and a latency optimization offloading algorithm based on an improved Lagrange multiplier algorithm is proposed to improve the overall efficiency of the system. The simulation results show that the proposed solution can preferentially ensure the processing rate of high-critical-level tasks and at the same time reduce the overall average latency of the system, effectively improving the total system revenue.

[0054] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit them. Any other modifications or equivalent replacements made by those of ordinary skill in the art to the technical solutions of the present invention should be covered within the scope of the claims of the present invention as long as they do not depart from the spirit and scope of the technical solutions of the present invention.

Claims

1. A service offloading and latency optimization method based on 5G edge computing, characterized in that, This method is implemented by the following steps: Step S1: Construct a three-level cloud-edge-end collaborative architecture, including a power service terminal device layer, an MEC node layer, and a central cloud layer. The power service terminal device layer is connected to the MEC node layer through a wireless network, and the MEC node layer is connected to the central cloud layer through a high-bandwidth backhaul link; Step S2: Obtain the set N = {1, 2,..., n} of power service terminals and the hybrid critical services (c n , λ n , τ n , ε n ) of each power service terminal user for the penalty function and the objective optimization function to call; where c n is the actual transmission data size required for offloading services of a unit in wireless network transmission, λ n is the average distribution number of single transmission requests of power terminal devices, τ n is the maximum tolerable delay for hybrid power services, ε n represents the priority of hybrid tasks; Step S3: Construct an edge node communication delay model and a power service computing delay model; Step S4: Establish a joint optimization model including wireless network resource allocation and task offloading strategies. The model aims to minimize the total system delay and is a convex optimization problem with equality constraints and inequality constraints. Use the KKT conditions and the Lagrange multiplier method to solve the convex optimization problem, minimize the queuing delay and transmission delay, and obtain a unique decision solution for the wireless transmission network resources by removing some inequality constraints; Step S5: Based on the task computation amount and delay requirements, adopt a differential offloading decision mechanism, set a penalty function, and improve the offloading success rate of high-priority tasks; Step S6: Based on the edge node communication model and the power service computing delay model, define the target optimization function as a composite function that minimizes the total system delay, and solve the optimal decisions for power terminals and critical service offloading .

2. The method for service offloading and latency optimization based on 5G edge computing according to claim 1, characterized in that In step S3, a communication delay model is constructed to impose a constraint on the total average delay r of all power service terminal offloading computing tasks: , where r n is the average delay of the offloading task of the nth power service terminal; the wireless transmission bandwidth b n of each power service terminal and the total transmission bandwidth B satisfy .

3. A method for service offloading and latency optimization based on 5G edge computing according to claim 2, characterized in that, In step S3, when constructing the power service computing delay model, constraints on the transmission delay, the average computing delay of the mobile edge service node, the average computing delay of the central cloud, and the round-trip transmission delay when the MEC executes tasks to the cloud are carried out: Transmission delay , where the service rate of the wireless network for power terminal n is , and the system decision maker needs to satisfy when scheduling resources ; Average computing delay of mobile edge service nodes , where is the total average rate at which the mobile edge server executes offloading tasks, satisfies ; Average computing delay of the central cloud , where is the total average rate at which the central cloud executes task offloading, is the maximum computing resource that the central cloud service instance can provide for this system, satisfies ; The round-trip transmission delay when the MEC executes tasks to the cloud is .

4. The method for service offloading and latency optimization based on 5G edge computing according to claim 3, wherein In step S4, the method for establishing a joint optimization model including wireless network resource allocation and task offloading strategies includes: where and are both Lagrange multipliers, is the inequality constraint in the optimization problem, is the equality constraint condition, defines the total wireless network transmission resources and the total amount of power terminal user task offloading in the cloud-edge collaborative system.

5. A service offloading and latency optimization method based on 5G edge computing according to claim 4, characterized in that, In step S4, after substituting the wireless resource scheduling partial inequality into the joint optimization model, the wireless transmission network resource scheduling decision represented by is obtained : ; Unique decision solution for wireless transmission network resources: , where represents an infinitesimal positive number. When , is negative infinity, and at this time is negative; when is in the interval , the obtained is in the interval ) and meets the actual requirements.

6. The service offloading and latency optimization method based on 5G edge computing according to claim 5, characterized in that In step S5, based on the task computation amount and latency requirements, a differential offloading decision mechanism is adopted, and a penalty function is set to improve the offloading success rate of higher-priority services: ; Among them, represents the penalty coefficients for different priorities, represents the execution delay of the task. When the task is executed at the edge node , when the MEC cannot meet the delay requirements of the smart grid device side, the task is then offloaded to the central cloud, then .

7. A service offloading and latency optimization method based on 5G edge computing according to claim 6, characterized in that, In step S6, the steps for constructing the target optimization function include: S61: Construct a weighted response delay model for mobile edge computing assisted by the central cloud; , Among them, the response delay of the mobile edge computing system assisted by the central cloud consists of five parts, namely, the wireless transmission delay of the power terminal device to offload tasks, the task execution delay of the mobile edge computing node, the transmission delay from the mobile edge node to the central cloud, the task execution delay of the central cloud, and the transmission delay of the calculation result back; S62. Construct an optimization objective function for the offloading decision of hybrid critical services for various power terminals based on the execution success rate of hybrid critical tasks and the weighted total delay of uploading and computing offloading: , where is the weighted sum of time for each link, is the offloading success rate of the power hybrid service.

Images