Edge caching and computing resource migration control method for novel power system

Abstract

The invention discloses an edge caching and computing resource migration control method for a novel power system. The method comprises the following steps: providing a resource allocation strategy combining service caching and VM migration based on a task unloading model, a service caching model of an edge server and a VM migration model; for the proposed resource allocation problem of joint service caching and VM migration, a continuous variable and an integer variable in an original problem are decoupled into two sub-variables, namely a service caching and computing resource allocation joint optimization sub-problem and a communication resource optimization sub-problem; proving that the service cache and computing resource allocation joint optimization sub-problem has an optimal solution, and simplifying the communication resource optimization sub-problem; and solving the sub-problems through a Monte Carlo algorithm, and carrying out iterative optimization to obtain an optimal solution of the original problem. The method has the beneficial effects that the equipment task processing time delay is reduced, the bearing capacity and the transmission rate of an electric power communication network are improved, and comprehensive optimization of computing resource allocation is realized.

Description

Technical Field

[0001] The present invention relates to the technical field related to a new type of power system with new energy as the main body, and in particular to an edge cache and computing resource migration control method for the new type of power system. Background Art

[0002] With the development of new power systems with new energy as the main body, the distributed characteristics of the system are more obvious and the number of devices has increased dramatically. Intelligent development and improving energy utilization efficiency have become important tasks for the development of the current new energy power system, and are also the key to my country's goal of achieving extremely high penetration of distributed photovoltaic power generation. With the access of a large number of distributed energy sensors and intelligent terminal devices, the daily data collection volume of the power system has reached more than 650GB. However, there is a huge conflict between the task processing requirements of distributed terminals in complex deployment environments and devices with limited computing power; at the same time, delays will cause the supply and demand power of high-proportion distributed devices to be unable to dynamically balance and affect the utilization efficiency of distributed energy. Therefore, it is urgent to propose a communication and computing resource optimization method for new power systems with new energy as the main body, so as to improve the carrying capacity and transmission rate of the power communication network and meet the support capacity of the new power system.

[0003] On the other hand, the communication resource allocation of joint service caching and virtual machine (VM) migration based on edge computing in the new power system scenario faces many challenges: for the task offloading problem of a large number of distributed energy devices, in the process of optimizing resource allocation, the optimal transmission power of each device should be formulated to improve the data transmission efficiency. In order to better deal with this problem, the coordination between device latency requirements and server resources should also be considered comprehensively to ensure the efficiency and stability of data transmission. At the same time, this problem also involves service caching and virtual machine migration problems, and it is necessary to reasonably and effectively coordinate the variable values ​​of the two to achieve optimal resource allocation. Because it involves a mixed optimization problem of integer variables and continuous variables, it is an NP-hard problem. Therefore, it is necessary to propose an optimization algorithm to solve this problem in order to achieve comprehensive optimization of computing resource allocation. Summary of the invention

[0004] The present invention aims to overcome the above-mentioned deficiencies in the prior art and provides an edge caching and computing resource migration control method for a new power system, which reduces the device task processing delay and realizes comprehensive optimization of computing resource allocation.

[0005] In order to achieve the above object, the present invention adopts the following technical solutions:

[0006] A method for controlling edge caching and computing resource migration for a new power system specifically includes the following steps:

[0007] (1) Aiming at the impact of device transmission power, edge server computing and storage resources on device task delay in edge network communication system, a resource allocation strategy combining service caching and VM migration is proposed based on task offloading model, service caching model of edge server and VM migration model, taking into account the factors of device transmission power, edge server computing and storage resources, so as to reduce device task processing delay.

[0008] (2) The proposed resource allocation problem of joint service caching and VM migration is an NP-hard problem. First, the continuous variables and integer variables in the original problem are decoupled into two sub-variables, namely, the joint optimization sub-problem of service caching and computing resource allocation and the communication resource optimization sub-problem. Then, it is proved that the joint optimization sub-problem of service caching and computing resource allocation has an optimal solution, and the communication resource optimization sub-problem is simplified. Finally, the sub-problems are solved by the Monte Carlo algorithm and iterative optimization is performed to obtain the optimal solution of the original problem.

[0009] The present invention proposes a resource allocation strategy for joint service caching and VM migration to reduce the device task processing delay, so as to improve the carrying capacity and transmission rate of the power communication network and meet the support capacity for the new power system; at the same time, it also reasonably and effectively coordinates the resource allocation problems of service caching and VM migration to achieve comprehensive optimization of computing resource allocation.

[0010] Preferably, in step (1), an edge network communication system is composed of W base stations and K devices, which are represented by the sets W = {1, 2, ..., w} and K = {1, 2, ..., k} respectively; in this system, each base station integrates an edge server; assuming that a single task of a device is indivisible and there is a corresponding unique service function, the task attribute of device k in this system can be described as Among them, M k Indicates the data size of the task, O k Indicates the number of CPU cycles required to process unit task data. Indicates the maximum tolerable delay of the device task; consider allocating the limited computing resources of the edge server to a single service hosted by the VM. If the edge server hosts the device's service and VM, it can provide computing services to the device; each VM can perform computing for a specific type of device task, and the edge servers can control the migration of VMs to transfer device service functions; based on this system, the task processing flow is as follows:

[0011] (11) The device generates a computing task and requests the nearest base station to offload the task;

[0012] (12) The edge server implements a service cache strategy based on the task request of the previous device;

[0013] (13) If the service cache required for deploying the device task is not available on the edge server, it will consider converting other types of services or requesting VM migration containing the required services to other edge servers;

[0014] (14) If the computing task offloaded to the edge server is overloaded and the number of VMs required by the device cannot meet the demand, the task will be offloaded to the power cloud center for execution.

[0015] Preferably, in step (11), each device has its own resource configuration but can only process part of the tasks, so task offloading is divided into two cases: if the edge server deploys the service cache of the device task, it is offloaded to the edge server for processing; if the edge server does not deploy the service cache, the device task is offloaded to the power cloud center for execution.

[0016] Preferably, in step (11), an orthogonal frequency division multiplexing communication mode is adopted between the device and the edge server, and the edge server allocates the transmission power to the device, and the maximum value of the transmission power satisfies Therefore, the uplink transmission rate of the device performing task offloading is r k It can be expressed as:

[0017]

[0018] Where B represents the system bandwidth, g k,w represents the channel gain from device k to edge server w. Assuming that the transmission power of the device remains unchanged and is in the same edge network communication system, it can be considered unchanged in a short time interval; σ 2 represents the average noise power in the network, and δ represents the channel crosstalk.

[0019] As a preference, in step (11), when the task is offloaded to the edge server and the power cloud center, the device delays are task upload delay and and

[0020]

[0021] Among them, M k is the data size of the task, r k represents the uplink transmission rate of task offloading, r o,w is the transmission rate from the edge server to the power cloud center.

[0022] Preferably, in step (12), for the service cache strategy, first construct a binary decision variable y wk ∈{0,1} is the service cache decision of edge server w, when y wk=1 means that the edge server w has cached the service of device k. If the edge server does not cache the service required by device k, then Assume that each device task can only be associated with one edge server, so Considering the storage space capacity limit of the edge server, let the cache space size of edge server w be c w , so the following constraints must be satisfied:

[0023]

[0024] Among them, m k Indicates the size of the service buffer required to perform device tasks.

[0025] Preferably, in step (13), for VM migration, if the computing task of the device is overloaded and the number of VMs allocated to the service cannot meet the requirement of task computing delay, the internal function migration of the VM of the edge server w or the request for VM number migration from other servers is considered, specifically:

[0026] (131) The internal VM migration of edge servers is to migrate the service type VM with small computing task requirements to the VM m Activate a VM as a function of another service type k The VM migration delay within the server mainly comes from the VM activation time, so the VM internal migration delay It can be expressed as:

[0027]

[0028] Where: V w,mk is the number of VMs of type m activated as type k in edge server w, T A The activation time required to cache another type of database for the VM;

[0029] (132) Edge server w requests service type VM from other edge servers k The number of migration delays will be generated. In this system, the serial static migration of VMs is considered, and the migration delay mainly comes from the transmission delay.

[0030]

[0031] Where: m k A VM representing a single service type k Memory size, z k Indicates the rate at which VM memory is generated, V iw,k represents the number of VMs of service type k that migrate from edge server i to edge server w, r i,w Indicates the transmission rate between servers;

[0032] (133) After the task is offloaded to the edge server, the edge server deploys the corresponding VM for parallel processing of the task received by the device. Therefore, the processing delay required for the task of device k to reach the edge server w is Expressed as

[0033]

[0034] Where: f k V represents the service rate that the hardware resources equipped by a single VM can provide to device k. w,k represents the number of VMs allocated to device k by edge server w;

[0035] In order to ensure that the device task processing does not exceed the maximum tolerable delay, and because the number of VMs that can be deployed on the edge server is limited, the number of VMs deployed and migrated on the edge server w needs to meet the following conditions:

[0036]

[0037] in It is represented as the maximum delay tolerance of task k of device, H w It represents the maximum number of VMs that can be deployed on edge server w.

[0038] As a preferred choice, considering that the network service requirements of the business in the distributed energy communication scenario are mainly low latency, the goal is to minimize the total latency of the device to complete the task, and construct a joint optimization problem of device transmission power p, service cache decision y and VM number V. The problem is expressed as:

[0039]

[0040] Among them, T tr o，w is the task download delay from the power cloud center to the edge server, which is exactly the same as T tr w，o On the contrary; constraint (11) indicates that the transmission power of the device is limited; constraint (12) indicates the minimum number of VMs required to complete the task; constraint (13) indicates that the number of VMs that can be deployed on the edge server is limited; constraint (14) indicates that the cache capacity of the edge server is limited; constraint (15) indicates that each device task can only be associated with one edge server; constraint (16) indicates that y wk is a binary decision variable;

[0041] To solve the above problems, the original problem is first converted into a joint optimization problem of integer variables y, V and continuous variables p. The decoupling method is used to convert problem P0 into two sub-problems, namely, communication resource optimization problem P2 and service cache and resource allocation joint optimization problem P1. The Monte Carlo algorithm is used to solve the integer variable values. The specific solution idea is:

[0042] (a) In a fixed communication resource p k Solve the optimal service cache decision and computing resource allocation optimization problem under the condition of

[0043] (b) After obtaining the optimal decision {y, V}, the communication resource variables are optimized based on the solution results;

[0044] (c) The solution to the original problem is obtained by setting the number of iterations and the threshold value.

[0045] As a preferred solution, in solution idea (a), specifically:

[0046] After initializing the service cache decision variable y, problem P1 becomes a latency optimization function composed of multiple integer variables, expressed as:

[0047]

[0048] This optimization problem is a nonlinear integer programming problem and can be expressed as:

[0049] P1:minP(y,V|p (0) ) (18)

[0050]

[0051] Determine whether there is a minimum value in problem P1. To simplify the calculation, assign the discrete variable y a value of 1, and then let

[0052]

[0053] Get the Hessian matrix:

[0054]

[0055] It can be judged that H is a positive definite matrix, which proves that the P1 problem is a convex function with a minimum value; therefore, the Monte Carlo method can be used to solve formula (17) to obtain the given communication resource Optimal service caching decision and virtual machine resource allocation decision under The above four overlined parameters represent the average values ​​of the corresponding parameters.

[0056] As a preferred solution, in solution idea (b), specifically: after obtaining the solution to subproblem P1, the communication resource optimization problem should be solved; the obtained optimal task offloading, service cache decision and resource allocation solution y * 、V * Substituting into P0, we get problem P2:

[0057] P2:minP(p|y * , V * ) (25)

[0058]

[0059] To simplify the calculation, Using Taylor's quadratic expansion, The P2 problem is a linear programming problem. Obviously, there is an optimal variable value. The optimal solution can be obtained directly using the CVX solver. The objective function value of the subproblem P1 is UP(1). When the deviation from the objective function value of the subproblem P2 is less than the given threshold value, that is, ||UP(1)-UP(2)||≤τ, the problem is considered to have the optimal solution {p * ,y * ,V *}; Otherwise, the currently acquired communication resources Pass to subproblem P1, solve and update Loop until the threshold condition is met or the maximum number of iterations is reached.

[0060] The beneficial effects of the present invention are: combining the resource allocation strategy of service caching and VM migration to reduce the device task processing delay, so as to improve the carrying capacity and transmission rate of the power communication network and meet the support capacity for the new power system; and reasonably and effectively coordinating the resource allocation problems of service caching and VM migration to achieve comprehensive optimization of computing resource allocation. BRIEF DESCRIPTION OF THE DRAWINGS

[0061] Figure 1 It is a schematic diagram of a system model of the present invention;

[0062] Figure 2 It is a flow chart of the method of the present invention. DETAILED DESCRIPTION

[0063] The present invention is further described below in conjunction with the accompanying drawings and specific embodiments.

[0064] A method for controlling edge caching and computing resource migration for a new power system specifically includes the following steps:

[0065] (1) Aiming at the impact of device transmission power, edge server computing and storage resources on device task delay in edge network communication system, a resource allocation strategy combining service caching and VM migration is proposed based on task offloading model, service caching model of edge server and VM migration model, taking into account the factors of device transmission power, edge server computing and storage resources, so as to reduce device task processing delay.

[0066] (2) The proposed resource allocation problem of joint service caching and VM migration is an NP-hard problem. First, the continuous variables and integer variables in the original problem are decoupled into two sub-variables, namely, the joint optimization sub-problem of service caching and computing resource allocation and the communication resource optimization sub-problem. Then, it is proved that the joint optimization sub-problem of service caching and computing resource allocation has an optimal solution, and the communication resource optimization sub-problem is simplified. Finally, the sub-problems are solved by the Monte Carlo algorithm and iterative optimization is performed to obtain the optimal solution of the original problem.

[0067] like Figure 1 As shown in the figure, the edge network communication system consists of W base stations (BS) and K devices, which are represented by the sets W = {1, 2, ..., w} and K = {1, 2, ..., k} respectively. In this system, each base station integrates an edge server. Assuming that the single task of the device is indivisible and there is a corresponding unique service function, the task attribute of device k in this system can be described as Among them, M k Indicates the data size of the task (bit), O k Indicates the number of CPU cycles required to process unit task data (Hz / bit), It represents the maximum tolerable delay of the device task. Consider allocating the limited computing resources of the edge server to a single service hosted by a VM. If the edge server hosts some services and VMs of the device, it can provide computing services to the device. Each VM can perform computing for a specific type of device task, and the device service function can be transferred by controlling VM migration between edge servers.

[0068] Based on this system, the task processing flow is as follows:

[0069] (11) The device generates a computing task and requests the nearest base station to offload the task;

[0070] Each device has its own resource configuration but can only handle a small number of tasks. Considering that the device tasks are highly complex and cannot be divided, task offloading is divided into two cases: if the edge server deploys the service cache of the device task, it is offloaded to the edge server for processing; if the edge server does not deploy the service cache, the device task is offloaded to the power cloud center for execution.

[0071] The device and the edge server use OFDM to communicate. The edge server allocates the transmission power to the device. The maximum transmission power meets Therefore, the uplink transmission rate of the device performing task offloading is r k It can be expressed as:

[0072]

[0073] Where B represents the system bandwidth, g k,w represents the channel gain from device k to edge server w. Assuming that the transmission power of the device remains unchanged and is in the same edge network communication system, it can be considered unchanged in a short time interval; σ 2 represents the average noise power in the network, and δ represents the channel crosstalk.

[0074] When the task is offloaded to the edge server and the power cloud center, the device delay is task upload delay and

[0075]

[0076] Among them, M k is the data size of the task, r k represents the uplink transmission rate of task offloading, r o,w is the transmission rate from the edge server to the power cloud center. Note: The superscripts u and tr here have no actual meaning. They are used to distinguish two delays. For example, u is the first letter of the English word "upload", and the following internal migration delay Similarly, these three parameters can be removed; k refers to device k, w refers to edge server w, and o refers to power cloud center o.

[0077] (12) The edge server implements a service cache strategy based on the task request of the previous device;

[0078] For the service cache strategy, first construct a binary decision variable y wk ∈{0,1} is the service cache decision of edge server w, when y wk =1 means that the edge server w has cached the service of device k. If the edge server does not cache the service required by device k, then Assume that each device task can only be associated with one edge server, so Considering the storage space capacity limit of the edge server, let the cache space size of edge server w be c w , so the following constraints must be satisfied:

[0079]

[0080] Among them, m k Indicates the size of the service buffer required to perform device tasks.

[0081] (13) If the service cache required for deploying the device task is not available on the edge server, it will consider converting other types of services or requesting VM migration containing the required services to other edge servers;

[0082] For VM migration, if the computing task of the device is overloaded and the number of VMs allocated to the service cannot meet the requirements of task computing latency, consider internal function migration of the VM of edge server w or request VM quantity migration from other servers, specifically:

[0083] (131) The migration of VMs within edge servers is to migrate service-type VMs with smaller computing task requirements to m Activate a VM as a function of another service type k The VM migration delay within the server mainly comes from the VM activation time, so the VM internal migration delay It can be expressed as:

[0084]

[0085] Where: V w,mk is the number of VMs of type m activated as type k in edge server w, T A The activation time required to cache another type of database for the VM;

[0086] (132) Edge server w requests service type VM from other edge servers k The number of migration delays will be generated. In this system, the serial static migration of VMs is considered, and the migration delay mainly comes from the transmission delay.

[0087]

[0088] Where: m k A VM representing a single service type k Memory size, z k Indicates the rate at which VM memory is generated, V iw,k represents the number of VMs of service type k that migrate from edge server i to edge server w, r i,w Indicates the transmission rate between servers;

[0089] (133) After the task is offloaded to the edge server, the edge server deploys the corresponding VM for parallel processing of the task received by the device. Therefore, the processing delay required for the task of device k to reach the edge server w is Expressed as

[0090]

[0091] Where: f k V represents the service rate that the hardware resources equipped by a single VM can provide to device k. w,k represents the number of VMs allocated to device k by edge server w;

[0092] In order to ensure that the device task processing does not exceed the maximum tolerable delay, and because the number of VMs that can be deployed on the edge server is limited, the number of VMs deployed and migrated on the edge server w needs to meet the following conditions:

[0093]

[0094] in It is represented as the maximum delay tolerance of task k of device, H w It represents the maximum number of VMs that can be deployed on edge server w.

[0095] (14) If the computing task offloaded to the edge server is overloaded and the number of VMs required by the device cannot meet the demand, the task will be offloaded to the power cloud center for execution.

[0096] Considering that the network service requirements of the business in the distributed energy communication scenario are mainly low latency, the goal is to minimize the total latency of the device to complete the task. The joint optimization problem of device transmission power p, service cache decision y and VM number V is constructed. The problem is expressed as:

[0097]

[0098] Among them, T tr o，w is the task download delay from the power cloud center to the edge server, which is exactly the same as T tr w，o On the contrary; constraint (11) indicates that the transmission power of the device is limited; constraint (12) indicates the minimum number of VMs required to complete the task; constraint (13) indicates that the number of VMs that can be deployed on the edge server is limited; constraint (14) indicates that the cache capacity of the edge server is limited; constraint (15) indicates that each device task can only be associated with one edge server; constraint (16) indicates that y wk is a binary decision variable.

[0099] To solve the above problems, the original problem is first converted into a joint optimization problem of integer variables y, V and continuous variables p, and the decoupling method is used to convert the problem P0 into two sub-problems, namely the communication resource optimization problem P2 and the service cache and resource allocation joint optimization problem P1. As a computational method based on random sampling and statistical reasoning, the Monte Carlo algorithm has the advantages of high accuracy, strong adaptability, good parallelism, easy implementation, and strong interpretability. The accuracy of the results is gradually improved as the number of sampling increases. Therefore, the Monte Carlo algorithm is used to solve the integer variable value. The specific solution idea is:

[0100] (a) In a fixed communication resource p k Under the condition of , the optimal service cache decision and computing resource allocation optimization problem are solved; the joint optimization of service cache and computing resource allocation is specifically as follows:

[0101] After initializing the service cache decision variable y, problem P1 becomes a latency optimization function composed of multiple integer variables, expressed as:

[0102]

[0103] This optimization problem is a nonlinear integer programming problem, which can be expressed as: P1:minP(y,Vp (0) )(18)

[0104]

[0105]

[0106] Determine whether there is a minimum value in problem P1. To simplify the calculation, assign the discrete variable y a value of 1, and then let

[0107]

[0108] Get the Hessian matrix:

[0109]

[0110] It can be judged that H is a positive definite matrix, which proves that the P1 problem is a convex function with a minimum value; therefore, the Monte Carlo method can be used to solve formula (17) to obtain the given communication resource Optimal service caching decision and virtual machine resource allocation decision under The above four overlined parameters represent the average values ​​of the corresponding parameters.

[0111] (b) After obtaining the optimal decision {y, V}, the communication resource variables are optimized based on the solution. The communication resource optimization is as follows: after obtaining the solution to subproblem P1, the communication resource optimization problem is solved. The optimal task offloading, service cache decision and resource allocation solution y are obtained.* 、V * Substituting into P0, we get problem P2:

[0112] Here, y * 、V * Substituting into P0 means: The parameters are copied to P0, including formula (10) to formula (16),

[0113] P2:minP(p|y * , V * ) (25)

[0114]

[0115] To simplify the calculation, Using Taylor's quadratic expansion, The P2 problem is a linear programming problem. Obviously, there is an optimal variable value. The optimal solution can be obtained directly using the CVX solver. The objective function value of the subproblem P1 is UP(1). When the deviation from the objective function value of the subproblem P2 is less than the given threshold value, that is, ||UP(1)-UP(2)||≤τ, the problem is considered to have the optimal solution {p * ,y * ,V *}; Otherwise, the currently acquired communication resources Pass to subproblem P1, solve and update Loop until the threshold condition is met or the maximum number of iterations is reached.

[0116] (c) The solution to the original problem is obtained by setting the number of iterations and the threshold value.

[0117] Based on the above analysis, the original problem is split into two sub-problems by decoupling variables, and then the solved variables are substituted into the original problem to minimize the time delay for the device to complete the task. The specific steps are as follows: Figure 2 The present invention proposes a resource allocation strategy for joint service caching and VM migration to reduce the device task processing delay, so as to improve the carrying capacity and transmission rate of the power communication network and meet the support capacity for the new power system; at the same time, it also reasonably and effectively coordinates the resource allocation problems of service caching and VM migration to achieve comprehensive optimization of computing resource allocation.

Claims

1. A method for edge caching and computing resource migration control for a new power system, characterized in that: The specific steps include: (1) Aiming at the impact of device transmission power, edge server computing and storage resources on device task delay in edge network communication system, a resource allocation strategy combining service caching and VM migration is proposed based on task offloading model, service caching model of edge server and VM migration model, taking into account the factors of device transmission power, edge server computing and storage resources, so as to reduce device task processing delay. (2) The proposed resource allocation problem of joint service caching and VM migration is an NP-hard problem. First, the continuous variables and integer variables in the original problem are decoupled into two sub-variables, namely, the joint optimization sub-problem of service caching and computing resource allocation and the communication resource optimization sub-problem. Then, it is proved that the joint optimization sub-problem of service caching and computing resource allocation has an optimal solution, and the communication resource optimization sub-problem is simplified. Finally, the sub-problems are solved by the Monte Carlo algorithm and iterative optimization is performed to obtain the optimal solution of the original problem.

2. According to claim 1, a method for controlling edge cache and computing resource migration for a new power system is characterized in that: In step (1), an edge network communication system is composed of W base stations and K devices, which are represented by the sets W = {1, 2, ..., w} and K = {1, 2, ..., k} respectively. In this system, each base station integrates an edge server. Assuming that the single task of the device is indivisible and has a corresponding unique service function, the task attribute of device k in this system can be described as Among them, M k Indicates the data size of the task, O k Indicates the number of CPU cycles required to process unit task data. Indicates the maximum tolerable delay of a device task; consider allocating the limited computing resources of the edge server to a single service hosted by a VM. If the edge server hosts the device's service and VM, it can provide computing services to the device; each VM can perform computing for a specific type of device task, and the edge servers can control the migration of VMs to transfer device service functions; Based on this system, the task processing flow is as follows: (11) The device generates a computing task and requests the nearest base station to offload the task; (12) The edge server implements a service cache strategy based on the task request of the previous device; (13) If the service cache required for deploying the device task is not available on the edge server, it will consider converting other types of services or requesting VM migration containing the required services to other edge servers; (14) If the computing task offloaded to the edge server is overloaded and the number of VMs required by the device cannot meet the demand, the task will be offloaded to the power cloud center for execution.

3. According to claim 2, a method for controlling edge cache and computing resource migration for a new power system is characterized in that: In step (11), each device has its own resource configuration but can only process part of the tasks, so task offloading is divided into two cases: if the edge server deploys the service cache of the device task, it is offloaded to the edge server for processing; If the edge server does not deploy a service cache, the device task is offloaded to the power cloud center for execution.

4. The edge cache and computing resource migration control method for a new power system according to claim 2 or 3, characterized in that: In step (11), the device and the edge server use orthogonal frequency division multiplexing communication, and the edge server allocates transmission power to the device, and the maximum transmission power satisfies Therefore, the uplink transmission rate of the device performing task offloading is r k It can be expressed as: Where B represents the system bandwidth, g k,w represents the channel gain from device k to edge server w. Assuming that the transmission power of the device remains unchanged and is in the same edge network communication system, it can be considered unchanged in a short time interval; σ 2 represents the average noise power in the network, and δ represents the channel crosstalk.

5. The edge cache and computing resource migration control method for a new power system according to claim 4 is characterized in that: In step (11), when the task is offloaded to the edge server and the power cloud center, the device delays are task upload delay and and Among them, M k is the data size of the task, r k represents the uplink transmission rate of task offloading, r o,w is the transmission rate from the edge server to the power cloud center.

6. The edge cache and computing resource migration control method for a new power system according to claim 5 is characterized in that: In step (12), for the service cache strategy, we first construct a binary decision variable y wk ∈{0,1} is the service cache decision of edge server w, when y wk =1 means that the edge server w has cached the service of device k. If the edge server does not cache the service required by device k, then Assume that each device task can only be associated with one edge server, so Considering the storage space capacity limit of the edge server, let the cache space size of edge server w be c w , so the following constraints must be satisfied: Among them, m k Indicates the size of the service buffer required to perform device tasks.

7. The edge cache and computing resource migration control method for a new power system according to claim 6 is characterized in that: In step (13), for VM migration, if the computing task of the device is overloaded and the number of VMs allocated to the service cannot meet the requirements of the task computing delay, the internal function migration of the VM of the edge server w or the request for VM number migration from other servers is considered, specifically: (131) The internal VM migration of edge server w is to activate the service type VMm with small computing task requirements to another service type VMk. The VM migration delay within the server mainly comes from the VM activation time, so the internal VM migration delay It can be expressed as: Where: V w,mk is the number of VMs of type m activated as type k in edge server w, T A The activation time required to cache another type of database for the VM; (132) When edge server w requests the number of service type VMk from other edge servers, migration delay will occur. In this system, serial static migration of VM is considered, and the migration delay mainly comes from transmission delay. Where: m k Indicates the VMk memory size of a single service type, z k Indicates the rate at which VM memory is generated, V iw,k represents the number of VMs of service type k that migrate from edge server i to edge server w, r i,w Indicates the transmission rate between servers; (133) After the task is offloaded to the edge server, the edge server deploys the corresponding VM for parallel processing of the task received by the device. Therefore, the processing delay required for the task of device k to reach the edge server w is Expressed as Where: f k V represents the service rate that the hardware resources equipped by a single VM can provide to device k. w,k represents the number of VMs allocated to device k by edge server w; In order to ensure that the device task processing does not exceed the maximum tolerable delay, and because the number of VMs that can be deployed on the edge server is limited, the number of VMs deployed and migrated on the edge server w needs to meet the following conditions: in It is represented as the maximum delay tolerance of task k of device, H w It represents the maximum number of VMs that can be deployed on edge server w.

8. The edge cache and computing resource migration control method for a new power system according to claim 7 is characterized in that: Considering that the network service requirements of the business in the distributed energy communication scenario are mainly low latency, the goal is to minimize the total latency of the device to complete the task. The joint optimization problem of device transmission power p, service cache decision y and VM number V is constructed. The problem is expressed as: Among them, T tr o，w is the task download delay from the power cloud center to the edge server, which is exactly the same as T tr w，o On the contrary; constraint (11) indicates that the transmission power of the device is limited; constraint (12) indicates the minimum number of VMs required to complete the task; constraint (13) indicates that the number of VMs that can be deployed on the edge server is limited; constraint (14) indicates that the cache capacity of the edge server is limited; constraint (15) indicates that each device task can only be associated with one edge server; constraint (16) indicates that y wk is a binary decision variable; To solve the above problems, the original problem is first converted into a joint optimization problem of integer variables y, V and continuous variables p. The decoupling method is used to convert problem P0 into two sub-problems, namely, communication resource optimization problem P2 and service cache and resource allocation joint optimization problem P1. The Monte Carlo algorithm is used to solve the integer variable values. The specific solution idea is: (a) In a fixed communication resource p k Solve the optimal service cache decision and computing resource allocation optimization problem under the condition of (b) After obtaining the optimal decision {y, V}, the communication resource variables are optimized based on the solution results; (c) The solution to the original problem is obtained by setting the number of iterations and the threshold value.

9. The edge cache and computing resource migration control method for a new power system according to claim 8 is characterized in that: In solution idea (a), the specific solution is: After initializing the service cache decision variable y, problem P1 becomes a latency optimization function composed of multiple integer variables, expressed as: This optimization problem is a nonlinear integer programming problem and can be expressed as: P1:minP(y,V|p (0) ) (18) Determine whether there is a minimum value in problem P1. To simplify the calculation, assign the discrete variable y a value of 1, and then let Get the Hessian matrix: It can be judged that H is a positive definite matrix, which proves that the P1 problem is a convex function with a minimum value; therefore, the Monte Carlo method can be used to solve formula (17) to obtain the given communication resource Optimal service caching decision and virtual machine resource allocation decision under The above four overlined parameters represent the average values ​​of the corresponding parameters.

10. The edge cache and computing resource migration control method for a new power system according to claim 9, characterized in that: In solution idea (b), specifically: after obtaining the solution to subproblem P1, the communication resource optimization problem must be solved; the optimal task offloading, service cache decision and resource allocation solution y * 、V * Substituting into P0, we get problem P2: P2:minP(p|y * ,V * ) (25) To simplify the calculation, Using Taylor's quadratic expansion, The P2 problem is a linear programming problem. Obviously, there is an optimal variable value. The optimal solution can be obtained directly using the CVX solver. The objective function value of the subproblem P1 is UP(1). When the deviation from the objective function value of the subproblem P2 is less than the given threshold value, that is, ||UP(1)-UP(2)||≤τ, the problem is considered to have the optimal solution {p * ,y * ,V * }; Otherwise, the currently acquired communication resources Pass to subproblem P1, solve and update Loop until the threshold condition is met or the maximum number of iterations is reached.

Images