Multi-task edge computing resource scheduling method and system

Abstract

The invention relates to the field of mobile edge computing, and discloses a multi-task edge computing resource scheduling method and system, and the method comprises the steps: constructing a new energy electric field mobile edge computing scheduling model, and achieving the efficient operation under a multi-task condition; and efficient operation is realized under the condition of multiple tasks. In order to improve the timeliness of data processing, the equipment migrates a task to an edge node with low delay and sufficient resources so as to maximize the experience quality of the equipment under the constraint of service migration cost and time delay; a random optimization problem is converted into a deterministic optimization problem by adopting a Lyapunov method, and combined optimization of task scheduling and resource allocation is performed for a time-varying dynamic environment in combination with deep reinforcement learning, so that QoE (Quality of Experience) is improved to the greatest extent and the overall performance of the system is optimized. According to the invention, by combining edge calculation and deep reinforcement learning technologies, intelligent management and optimal scheduling of the new energy electric field are realized, and medium and long term planning and stability of a new energy electric field system are guaranteed.

Description

Technical Field

[0001] The present invention relates to the field of mobile edge computing technology, and in particular to a multi-task edge computing resource scheduling method and system. Background Art

[0002] As the global energy structure transforms towards a more environmentally friendly and sustainable direction, new energy power plants (such as wind power, solar power, etc.), as an important part of renewable energy, account for an increasing proportion of global electricity supply. However, this growth also brings new challenges, especially in task scheduling. New energy power plants usually contain multiple types of energy equipment, and the operation of these equipment involves multiple complex and interdependent task links. In order to ensure the efficient operation and stability of the entire system, these tasks must be managed in an orderly manner.

[0003] In the traditional power plant operation mode, task scheduling mainly relies on static planning and preset execution, which is difficult to adapt to the uncertainty caused by factors such as weather changes and equipment status changes in new energy power plants. In addition, with the development of edge computing (MEC) technology, the mobility of sensor devices and unpredictable computing environments have further increased the complexity of task scheduling. In this context, the service migration cost during user migration has become a long-term constraint, affecting system design and resource allocation strategies.

[0004] To solve the above problems, intelligent scheduling and dynamic resource management are key. By analyzing and predicting the status of different energy devices and their working environment in real time, the task execution order and resource allocation are dynamically adjusted to cope with the complexity of the system and the uncertainty of the external environment. Multi-task edge computing resource scheduling methods have emerged, which can not only optimize the efficiency of computing resources, but also improve the overall performance and reliability of new energy power plants. Summary of the invention

[0005] In view of the above existing problems, the present invention is proposed.

[0006] Therefore, the present invention provides a multi-task edge computing resource scheduling method and system to solve the problem that the existing technology is difficult to cope with the system complexity and uncertainty caused by the complex dependencies of various energy devices, the mobility of sensor equipment, the unpredictable edge computing environment and the long-term constraints brought by user migration costs when dealing with multi-task scheduling of new energy power fields.

[0007] In order to solve the above technical problems, the present invention provides the following technical solutions: In the first aspect, the present invention provides a multi-task edge computing resource scheduling method, including: combining various parameters of new energy electric field equipment and mobile edge computing coverage, constructing a new energy electric field mobile edge computing scheduling model; constructing an ordered set of auxiliary task scheduling based on the new energy electric field mobile edge computing scheduling model and the dependency relationship between new energy tasks; maximizing the experience quality of the new energy electric field data acquisition system by migrating tasks to edge nodes, and constructing a maximum experience quality optimization problem under the condition of satisfying the data task processing delay and energy constraints of the new energy electric field system; using Lyapunov optimization to decouple the constraints of long-term migration costs in the maximum experience quality optimization problem, and converting the problem into an optimization problem without long-term constraints by constructing a virtual migration cost optimization queue; using a deep reinforcement learning algorithm to jointly optimize task scheduling and resource allocation for the optimization problem without long-term constraints, so as to realize resource scheduling of multi-task edge computing.

[0008] As a preferred solution of a multi-task edge computing resource scheduling method described in the present invention, wherein: the construction of a new energy electric field mobile edge computing scheduling model includes: set up Represents the set of new energy electric field devices. Each device generates M computing tasks in each time slot. The task set is ,set up represents a collection of edge servers, each of which has computing resources and bandwidth resources , when n is within the communication coverage of the edge server, the service scheduling of the new energy electric field equipment n can be performed; Assume that the user schedules the task in time slot t as ,in Indicates the size of the scheduled task, Indicates the total number of CPU cycles required to complete the scheduled task. Indicates the task Maximum latency requirement; The system has N*M tasks, and the set of all tasks is Part of the task is executed on the new energy electric field equipment, and part is dispatched to the nearby edge server. For limited local computing resources, define is the ratio coefficient of task scheduling to edge servers, where .

[0009] As a preferred solution of a multi-task edge computing resource scheduling method described in the present invention, wherein: the construction of an ordered set of auxiliary task scheduling includes: Set the task set received by the edge server to , the set of tasks performed locally by the new energy electric field equipment is ,in and It is an ordered set, and all tasks are executed in the order they are entered; Tasks are scheduled sequentially according to data type dependencies. When the new energy electric field device n schedules part of the tasks to the nearby edge server, the communication model is: The signal-to-noise ratio from the new energy electric field device n to the edge server k in time slot t is ; Transmission rate from new energy electric field device n to edge server k , where B k is the communication bandwidth resource from the new energy electric field equipment to the edge server, then the transmission delay caused by task scheduling is: , The corresponding transmission energy consumption is ; The computing tasks of the new energy electric field equipment can be processed simultaneously on the local device and the edge server. The system computing model is: when the mth data of the new energy electric field equipment n is calculated locally, the local computing delay in time slot t is expressed as: , The corresponding local energy consumption is ; When the mth data of the new energy electric field device n is dispatched to the edge server k, the computation delay generated by the edge server at time slot t is expressed as: , The corresponding energy consumption is .

[0010] As a preferred solution of a multi-task edge computing resource scheduling method described in the present invention, wherein: maximizing the experience quality of the new energy electric field data acquisition system by migrating tasks to edge nodes includes: The new energy electric field device n dispatches its task to an edge server in time slot t through a direct cellular link. In the next time slot, the new energy electric field device n roams to the communication coverage of another edge node. represents the cost of migrating a task from one edge server to another edge server. The migration cost of the mth task service of the new energy electric field device n is expressed as ,in, is the decision result of service migration. When the edge server of time slot t is different from the edge server of time slot t−1, ,otherwise ; The local computation of data and task scheduling are performed in parallel, and the execution time of device n is: , The total energy consumption of new energy electric field equipment n in time slot t is: , The task of each new energy electric field equipment is C nm There is a certain execution time , let task C nm Completion time for ,in, Represents task C nm The start time of For Task C n and Task C n+1 The dependency relationship between tasks C n+1 Need to be in Task C n After completion, execute Task C n+1 Start time Should meet , i.e. task C n+1 The start time of task C is no earlier than n completion time.

[0011] As a preferred solution of the multi-task edge computing resource scheduling method described in the present invention, the construction of the maximum experience quality optimization problem includes: The quality of experience (QoE) of a new energy electric field device n is determined by the task delay and energy consumption of the device. The following optimization objectives are defined: , in, represents the QoE penalty factor of the new energy electric field device n at time slot t, involving task delay and local energy consumption, weighted parameter represents the QoE preference factor of the user in time slot t; The optimization variables of the system in time slot t are defined as , , , the optimization problem is formally expressed as: , in, represents the maximum computing resources of device n, represents the delay requirement of the task, represents the used migration cost, represents the available migration cost, constraint C1 represents the equipment task scheduling ratio coefficient constraint, constraint C2 represents the equipment resource constraint, constraint C3 represents the equipment delay requirement, constraint C4 represents the service migration cost limit, and constraint C5 represents the execution dependency between the new energy electric field equipment tasks.

[0012] As a preferred solution of the multi-task edge computing resource scheduling method described in the present invention, wherein: the problem is converted into an optimization problem without long-term constraints by constructing a virtual migration cost optimization queue, including: Design a virtual migration cost optimization queue and define the evolution representation of the virtual migration cost optimization queue of the n slot of the new energy electric field equipment at time slot t; The Lyapunov function is defined to describe the sum of squares of all virtual migration cost optimization queue backlogs during the tth time slot. To ensure the stability of the system, As conditional Lyapunov drift; The drift penalty algorithm is used to obtain the Lyapunov drift penalty function. , the Lyapunov queue optimization technique is used to transform the original optimization problem into an optimization problem without long-term constraints, and the objective function is expressed as: , Among them, the positive control parameters V and Used to dynamically balance the trade-off between processing latency performance and migration cost consumption. represents the delay of the kth type of data of the nth node at time t.

[0013] As a preferred solution of the multi-task edge computing resource scheduling method of the present invention, the joint optimization of task scheduling and resource allocation for the optimization problem without long-term constraints using a deep reinforcement learning algorithm includes: The dynamic edge server scenario is described using a discrete-time Markov decision process (MDP). The state of the MDP is the communication state of the available edge servers around the new energy electric field equipment, and the energy consumption when the SE schedules tasks to the edge servers. According to the current system state, the mobile scheduling module takes action, and the agent performs actions based on the observed state and obtains rewards from the environment, where the rewards include the energy consumption of the system and the penalty for violating the delay constraint; The PPO algorithm consists of an action network and an evaluation network. The action network is divided into two parts, corresponding to the parameters and , the action network is based on the state Output Action and interact with the environment; the evaluation network calculates the state value based on the state information; To evaluate actions The performance of the advantage function is introduced , and calculate the objective function of the evaluation network, using the objective function of the evaluation network Back propagation updates the Critic network; The objective function of PPO constrains the update range by comparing the changes between the current strategy and the old strategy; Update the Actor network using backpropagation of the objective function.

[0014] In a second aspect, the present invention provides a multi-task edge computing resource scheduling system, comprising: The scheduling model building module is used to build a new energy electric field mobile edge computing scheduling model by combining various parameters of the new energy electric field equipment and the mobile edge computing coverage; An ordered set construction module, used to construct an ordered set auxiliary task scheduling based on the new energy electric field mobile edge computing scheduling model and the dependency relationship between new energy tasks; An optimization problem construction module is used to maximize the experience quality of the new energy electric field data acquisition system by migrating tasks to edge nodes, and to construct a maximum experience quality optimization problem under the conditions of satisfying the data task processing delay and energy constraints of the new energy electric field system; An optimization problem conversion module, used for decoupling the constraints of long-term migration cost in the optimization problem of maximizing the quality of experience by using Lyapunov optimization, and converting the problem into an optimization problem without long-term constraints by constructing a virtual migration cost optimization queue; The resource optimization scheduling module is used to use a deep reinforcement learning algorithm to jointly optimize task scheduling and resource allocation for the optimization problem without long-term constraints, so as to realize resource scheduling of multi-task edge computing.

[0015] In a third aspect, the present invention provides an electronic device, comprising: Memory and processor; The memory is used to store computer-executable instructions, and the processor is used to execute the computer-executable instructions. When the computer-executable instructions are executed by the processor, the steps of the multi-task edge computing resource scheduling method are implemented.

[0016] In a fourth aspect, the present invention provides a computer-readable storage medium storing computer-executable instructions, which, when executed by a processor, implement the steps of the multi-task edge computing resource scheduling method.

[0017] Compared with the prior art, the beneficial effects of the present invention are as follows: the present invention provides a multi-task edge computing resource scheduling method and system, which combines various parameters of new energy electric field equipment and base stations and system communication bandwidth to construct a new energy electric field mobile edge computing scheduling model; by considering the dependencies between tasks, the system can cope with more complex task scheduling requirements, construct an ordered set of auxiliary task scheduling, and perform real-time data transmission in sequence to maximize the execution efficiency of the overall task and the utilization rate of the system; to ensure the real-time and timeliness of the data of the new energy electric field equipment, the new energy electric field equipment can migrate tasks to edge nodes with low latency and sufficient resources to achieve system reliability, optimize resource utilization and improve QoE; under the conditions of meeting the data task processing delay and energy constraints of the new energy electric field system, a maximum QoE optimization problem P1 is constructed to ensure that each node maximizes the overall efficiency of the system while maintaining a good response time; Lyapunov optimization is used to establish a virtual migration cost optimization queue to decouple the long-term migration cost constraints of NE in the optimization problem P1; the PPO algorithm is used to jointly optimize task scheduling and resource allocation for optimization problems without long-term constraints to maximize QoE and optimize the overall performance of the system. The present invention combines edge computing and deep reinforcement learning technology to achieve intelligent management and optimized scheduling of new energy power fields, thereby ensuring the medium- and long-term planning and stability of the new energy power field system. BRIEF DESCRIPTION OF THE DRAWINGS

[0018] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings required for use in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other accompanying drawings can be obtained based on these accompanying drawings without paying creative work.

[0019] Figure 1 A schematic diagram of the overall process logic of a multi-task edge computing resource scheduling method according to an embodiment of the present invention; Figure 2 An edge computing model diagram of multi-device multi-task scheduling in a new energy electric field of a multi-task edge computing resource scheduling method according to an embodiment of the present invention; Figure 3 A schematic diagram of the architecture of a PPO algorithm of a multi-task edge computing resource scheduling method according to an embodiment of the present invention; Figure 4 A schematic diagram of the convergence performance of PPO at different learning rates of a multi-task edge computing resource scheduling method according to an embodiment of the present invention; Figure 5 A schematic diagram of system QoE under different numbers of NEs and MECs in a multi-task edge computing resource scheduling method according to an embodiment of the present invention; Figure 6 This is a schematic diagram comparing system performance of different algorithms of the multi-task edge computing resource scheduling method described in an embodiment of the present invention at different task scales. DETAILED DESCRIPTION

[0020] In order to make the above-mentioned purposes, features and advantages of the present invention more obvious and easy to understand, the specific implementation methods of the present invention are described in detail below in conjunction with the drawings of the specification. Obviously, the described embodiments are part of the embodiments of the present invention, but not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary persons in the art without creative work should fall within the scope of protection of the present invention.

[0021] Example 1 Reference Figure 1-Figure 3 As an embodiment of the present invention, a multi-task edge computing resource scheduling method is provided, such as Figure 2 The applications shown include Sensor devices and edge server MEC; to optimize multi-task data processing, the present invention aims to efficiently allocate and schedule various tasks, and ultimately achieve the goal of maximizing the QoE of each sensor device; by comprehensively considering the performance and task requirements of multiple sensor devices in the electric field, coordinating the resource allocation of each device, and ensuring the priority and timeliness of real-time data collection, analysis and transmission tasks; in practical applications, such as Figure 1 The edge computing method for scheduling new energy electric field tasks specifically performs the following steps: S100: Combine various parameters of new energy electric field equipment and mobile edge computing coverage to build a new energy electric field mobile edge computing scheduling model; S200: Based on the new energy electric field mobile edge computing scheduling model and the dependency relationship between new energy tasks, an ordered set of auxiliary task scheduling is constructed; S300: Maximize the experience quality of the new energy electric field data acquisition system by migrating tasks to edge nodes; S400: Construct the optimization problem of maximizing the experience quality under the condition of meeting the data task processing delay and energy constraints of the new energy electric field system; S500: Lyapunov optimization is used to decouple the constraints of long-term migration costs in the optimization problem of maximizing the quality of experience. By building a virtual migration cost optimization queue, the problem is converted into an optimization problem without long-term constraints. S600: Use deep reinforcement learning algorithms to jointly optimize task scheduling and resource allocation for optimization problems without long-term constraints to achieve resource scheduling for multi-task edge computing.

[0022] It should be noted that the present invention realizes efficient operation in multi-task conditions by constructing a multi-device edge computing scheduling network model; to improve the timeliness of data processing, the device can migrate tasks to edge nodes with low latency and sufficient resources to maximize the device experience quality under the constraints of service migration cost and delay; in order to further improve the long-term utility of the system, the present invention uses the Lyapunov method to convert random optimization problems into deterministic optimization problems, and combines deep reinforcement learning to perform joint optimization of task scheduling and resource allocation for time-varying dynamic environments. The present invention realizes intelligent management and optimized scheduling of new energy electric fields by combining edge computing and deep reinforcement learning technology, and ensures the medium- and long-term planning and stability of new energy electric field systems.

[0023] In the embodiment of the present application, the above step S100 includes the following sub-steps A1 to A3: In A1: Represents the set of new energy electric field devices. Each sensor device generates M computing tasks in each time slot. The task set is ,set up Represents a collection of edge servers, which have computing resources and bandwidth resources , when n is within the communication coverage of the edge server, the service scheduling of the new energy electric field equipment n can be performed; In A2: Assume that the user schedules the task in time slot t as ,in Indicates the size of the scheduled task, Indicates the total number of CPU cycles required to complete the scheduled task. Indicates the task Maximum latency requirement; In A3: The new energy power plant dispatching system has N*M tasks, and the set of all tasks is expressed as Part of the task is executed on the new energy electric field equipment, and part is dispatched to the nearby edge server. For limited local computing resources, define is the ratio coefficient of task scheduling to edge servers, where .

[0024] It should be noted that the above step S100 provides an optimization basis for subsequent task scheduling and resource allocation by constructing a targeted mobile edge computing scheduling model, effectively improves data processing efficiency and system response speed, and supports maximizing experience quality.

[0025] In the embodiment of the present application, the above step S200 includes the following sub-steps B1 to B3: In B1: All tasks of the system can be scheduled, and the task set received by the edge server is set to , the set of tasks performed locally by the new energy electric field equipment is ,in and It is an ordered set, and all tasks are executed in the order they are entered; In B2: tasks are scheduled sequentially according to data type dependencies. When the new energy electric field device n schedules part of the tasks to the nearby edge server, the communication model is as follows: The signal-to-noise ratio from the new energy electric field device n to the edge server k in time slot t is: , Among them, p n,m is the transmission power of the mth task of the new energy electric field equipment n, is the channel gain from the new energy electric field device n to the edge server k, is the noise density; Transmission rate from new energy electric field device n to edge server k , where B k is the communication bandwidth resource from the new energy electric field equipment to the edge server, then the transmission delay caused by task scheduling is: , The corresponding transmission energy consumption is ; In B3: The computing tasks of the new energy electric field equipment can be processed simultaneously on the local device and the edge server. The system computing model is: When the mth data of the new energy electric field device n is calculated locally, the local calculation delay in time slot t is expressed as: , The corresponding local energy consumption is ,in is the effective capacitance coefficient of the new energy electric field device n; when the mth data of the new energy electric field device n is dispatched to the edge server k, the calculation delay generated by the edge server at time slot t is expressed as: , The corresponding energy consumption is .

[0026] It should be noted that the above step S200 forms an ordered set of auxiliary task scheduling based on the constructed new energy electric field mobile edge computing scheduling model and the dependencies between tasks, ensures the sequentiality and efficiency of task execution, optimizes resource allocation, and improves overall task processing efficiency and system stability.

[0027] In the embodiment of the present application, the above step S300 includes the following sub-steps C1 to C3: In C1: the new energy electric field device n dispatches its task to an edge server in time slot t through a direct cellular link. In the next time slot, the new energy electric field device n roams to the communication coverage of another edge node. At this time, NE n will establish a connection with the newly added MEC node through a new cellular link. This roaming process is based on the mobility of NE n and the change of network topology. It usually requires real-time evaluation of network conditions and allocation of computing resources to ensure smooth task processing and reduce communication delays. In this process, the process of migrating tasks from one edge server to another needs to consider a certain migration cost, which is used represents the cost of migrating a task from one edge server to another. The migration cost of the mth task service of the new energy electric field device n is expressed as: , in, is the decision result of service migration. When the edge server of time slot t is different from the edge server of time slot t−1, ,otherwise ; The total service migration cost of new energy electric field equipment n is expressed as: , In C2: local data computation and task scheduling are performed in parallel, and the execution time of device n is: , The total energy consumption of new energy electric field equipment n in time slot t is: , In C3: Each NE's task C nm There is a certain execution time , let task C nm The completion time is: , in, Represents task C nm The start time of For Task C n and Task C n+1 The dependency relationship between tasks C n+1 Need to be in Task C n After completion, execute Task C n+1 Start time Should meet , i.e. task C n+1 The start time of task C is no earlier than n completion time.

[0028] It should be noted that the above step S300 reduces data transmission delay and network load by migrating tasks to edge nodes, effectively improves the experience quality of the new energy electric field data acquisition system, achieves more efficient data processing and real-time response, and enhances the overall performance and reliability of the system.

[0029] In the embodiment of the present application, the above step S400 includes the following sub-steps D1-D2: In D1, the quality of experience (QoE) of the new energy electric field device n is determined by the task delay and energy consumption of the device, and the following optimization objectives are defined: , in, represents the QoE penalty factor of the new energy electric field device n at time slot t, involving task delay and local energy consumption. Will cause the end user's QoE to decrease, weighted parameters represents the QoE preference factor of the user in time slot t; the larger This means that the task delay ratio The local energy consumption is more important in It can be specified by the user or derived from the task processing history; In D2: The optimization variables of the system in time slot t are defined as , , , the optimization problem can be formally expressed as: , Among them, constraint C1 represents the device task scheduling ratio coefficient constraint, constraint C2 represents the device resource constraint, constraint C3 represents the device delay requirement, and constraint C4 represents the service migration cost limit. It represents the upper limit of the service mobility rate in the whole journey, and constraint C5 represents the execution dependency between the tasks of the new energy electric field equipment.

[0030] It should be noted that the above step S400 ensures the efficient operation of the system under resource-constrained conditions, while improving the quality of task execution and user satisfaction, and provides a solid foundation for the optimal scheduling of new energy power plants.

[0031] In the embodiment of the present application, the above step S500 includes the following sub-steps E1 to E4: In E1: The equipment in the optimization problem P1 is affected by the migration cost and cannot be solved directly. A virtual migration cost optimization queue is designed, and the evolution of the virtual migration cost optimization queue of the n slot of the new energy electric field equipment at time slot t is defined as: , in, represents the deviation between the used migration cost and the available migration cost; In E2: Define the Lyapunov function to describe the sum of squares of all virtual migration cost optimization queue backlogs during the tth time slot, and the formula is expressed as: , in In order to ensure the stability of the system, we introduce As a conditional Lyapunov drift, it is expressed as: , At E3: Control Reduce queue backlog, stabilize the system, and use the drift penalty algorithm to obtain the Lyapunov drift penalty function , expressed as: , Among them, the non-negative weight coefficient V is used to measure the weight of drift and objective function; the upper bound of Lyapunov drift penalty function is expressed as , in ; In E4: Lyapunov queue optimization technology is used to transform the original optimization problem into an optimization problem without long-term constraints, and the objective function is expressed as: , Among them, the positive control parameters V and Used to dynamically balance the trade-off between processing latency performance and migration cost consumption.

[0032] It should be noted that the above step S500 simplifies the calculation complexity, making resource scheduling more flexible and efficient, effectively reducing the migration cost and improving the scalability and response speed of the system.

[0033] In the embodiments of the present application, Figure 3 The above step S600 includes the following sub-steps F1-F2: In F1: The dispatching of renewable energy power plants needs to deal with the operation and management of multiple devices (such as wind turbines, solar panels, energy storage devices, etc.); the status and tasks of each device may be different, and the execution requirements of the tasks often change dynamically; PPO has strong adaptability when dealing with such multi-task problems, and can automatically learn and adjust the scheduling strategy of each device to ensure the overall performance of the system is maximized. Discrete-time MDP is used to describe dynamic MEC scenarios, and the PPO-based DRL algorithm is used to deal with long-term formal optimization problems P2. The changes in available computing resources and the changes in the described wireless environment follow the Markov property. The three key elements of the learning environment system are the state set s, the action set a, and the reward function set r. The state of MDP is the communication state of the available edge servers around the new energy electric field equipment, and the energy consumption when SE dispatches tasks to the edge servers. The state of the edge server system at each decision time t is expressed as: , According to the current system status , the mobile scheduling module takes action, expressed as: , The agent performs actions based on the observed state and obtains rewards from the environment. The reward obtained by the new energy electric field device in time slot t is expressed as: , The reward includes the energy consumption of the system and the penalty for violating the delay constraint; In F2: PPO algorithm includes action network and evaluation network , the action network is divided into two parts, corresponding to the parameters and , the action network is based on the state Output Action , and interact with the environment; set the PPO learning parameters, where the learning rate is 0.001, the experience pool size is fixed to 64, and the discount factor is set to 0.9. The hidden layers of the action network and the evaluation network are both fully connected neural networks, the activation function is ReLU, the number of network layers is 2, and the hidden layer sizes of the action network and the evaluation network are 128 and 64 respectively.

[0034] The evaluation network calculates the state value based on the state information as follows: , in, is the state of the new energy electric field equipment at time t+1, is the action of the new energy electric field equipment at time t+1; To evaluate actions The performance of the advantage function is introduced ,Right now: , in, is the residual of time difference method; Calculate the objective function of the evaluation network, the formula is expressed as: , Using the objective function of the evaluation network Back propagation updates the Critic network; The objective function of PPO constrains the update range by comparing the changes between the current strategy and the old strategy. The formula is expressed as: , Among them, clip is the cutting operation, is constrained to [1−ε,1+ε] to prevent the algorithm strategy from updating too quickly, using Back propagation updates the Actor network; The loss of the PPO algorithm is: , It should be noted that the above step S600 uses a deep reinforcement learning algorithm to jointly optimize task scheduling and resource allocation for optimization problems without long-term constraints, thereby achieving efficient scheduling of multi-task edge computing resources, which not only improves the system's resource utilization and task processing efficiency, but also enhances the ability to cope with dynamic environmental changes and ensures the optimization of service quality.

[0035] Example 2 Reference Figure 4~Figure 6 Based on the previous embodiment, this embodiment provides an application example of a multi-task edge computing resource scheduling method and system to verify and illustrate the technical effects used in this method.

[0036] Based on the simulation scenario, this embodiment obtained the following experimental data, such as Figure 4 The figure shows the convergence effect of the PPO algorithm applied to the edge computing method for scheduling tasks in new energy power plants, where M = 10, each round contains 200 unloading decision explorations, and the ordinate represents the average of the rewards obtained from the 200 explorations. The reward is a penalty for task processing delay and energy consumption cost. When the reward is relatively small, it means that scheduling services can be provided to NE in a more efficient way. Figure 4 It can be seen that when the algorithm learning rate is 0.0001 and 0.001, as the training continues, the reward function tends to converge, which indicates that the trained model is suitable for the scenario of this embodiment and has good convergence performance.

[0037] Figure 5The average delay of the proposed scheme under different NE numbers N and edge servers K is compared. The average delay refers to the ratio of the total delay of the offloaded task to the number of tasks in the entire journey, which measures the average satisfaction of the user with the task offloading service throughout the journey. From a vertical perspective, when the number of users remains unchanged, the average delay of the proposed scheme decreases with the increase of the number of MECs. When there are fewer edge servers, the time and resource constraints required for NE migration will increase, and the load of each server may be high, resulting in longer processing delays. As the number of MECs increases, more edge servers are available, load balancing is better, and task migration and computing resource allocation are more efficient, thereby reducing latency. From a horizontal perspective, as the number of users increases, the available resources for a single user will be limited, resulting in a higher task failure rate and making the average delay close to the task request delay. In addition, with the redundant computing resources of more edge servers, the average delay will intuitively decrease.

[0038] Figure 6 The system performance comparison under different task scales is given. Scheme 1 is a non-migration scheme, that is, after the NE's task is scheduled to an MEC, no task migration service is performed, and the task must be completed on the MEC; Scheme 2 is a scheme without migration cost constraints, that is, after the NE's task is scheduled to an MEC, the task migration service can be performed, but the long-term migration cost of the NE is not considered; Scheme 3 is a task offloading-only scheme, in which the edge server provides offloading services to the MD within its communication coverage, without considering the resource allocation problem of computing resources; Scheme 4 is a pure local computing scheme, in which the NE's task is only calculated locally, and no task scheduling is performed. Figure 6 As shown, when the task size remains unchanged, the system QoE performance of the solution proposed in this embodiment is optimal. This shows that reasonable resource allocation and task migration mechanisms are crucial to improving system performance, enabling the system to flexibly adjust resources and perform task migration. Solution 1 is limited by the computing resources of a single MEC. Especially when the task volume is large, the solution cannot efficiently process all tasks of the NE and dynamically adjust computing resources, resulting in a decrease in system performance. Solution 2 reduces the load of a single MEC node, but ignores the cost of long-term migration, which may lead to unnecessary frequent migration, thereby increasing system overhead when the task volume is large. For Solution 3, task offloading can reduce the burden on local devices, but there is a problem of unreasonable resource allocation, which affects the overall QoE of the system. In Solution 4, the ability of NE to handle all tasks is limited, and the system delay is the largest.

[0039] Therefore, the present invention realizes efficient operation in multi-task conditions by constructing a multi-device edge computing scheduling network model; to improve the timeliness of data processing, the device can migrate tasks to edge nodes with low latency and sufficient resources to maximize the device experience quality under the constraints of service migration cost and delay; in order to further improve the long-term utility of the system, the present invention uses the Lyapunov method to convert the random optimization problem into a deterministic optimization problem, and combines deep reinforcement learning to perform joint optimization of task scheduling and resource allocation for time-varying dynamic environments. The present invention realizes intelligent management and optimal scheduling of new energy electric fields by combining edge computing and deep reinforcement learning technology, and ensures the medium- and long-term planning and stability of new energy electric field systems.

[0040] Example 3 This embodiment provides a multi-task edge computing resource scheduling system, including: The scheduling model building module is used to build a new energy electric field mobile edge computing scheduling model by combining various parameters of the new energy electric field equipment and the mobile edge computing coverage; An ordered set construction module is used to construct an ordered set auxiliary task scheduling based on the new energy electric field mobile edge computing scheduling model and the dependency relationship between new energy tasks; An optimization problem construction module is used to maximize the experience quality of the new energy electric field data acquisition system by migrating tasks to edge nodes, and to construct a maximum experience quality optimization problem under the conditions of satisfying the data task processing delay and energy constraints of the new energy electric field system; The optimization problem conversion module is used to use Lyapunov optimization to decouple the constraints of long-term migration costs in the optimization problem of maximizing the quality of experience, and to convert the problem into an optimization problem without long-term constraints by building a virtual migration cost optimization queue; The resource optimization scheduling module is used to jointly optimize task scheduling and resource allocation for optimization problems without long-term constraints using a deep reinforcement learning algorithm to achieve resource scheduling for multi-task edge computing.

[0041] It should be noted that the technical solution of the multi-task edge computing resource scheduling system and the technical solution of the multi-task edge computing resource scheduling method mentioned above belong to the same concept. For details not described in detail in the technical solution of the multi-task edge computing resource scheduling system in this embodiment, please refer to the description of the technical solution of the multi-task edge computing resource scheduling method mentioned above.

[0042] The above-mentioned unit modules may be embedded in or independent of a processor in a computer device in the form of hardware, or may be stored in a memory in a computer device in the form of software, so that the processor can call and execute operations corresponding to the above-mentioned modules.

[0043] This embodiment also provides an electronic device, which includes a processor, a memory, a communication interface, a display screen and an input device connected through a system bus. Among them, the processor of the computer device is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system and a computer program. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The communication interface of the computer device is used to communicate with an external terminal in a wired or wireless manner, and the wireless manner can be implemented through WIFI, an operator network, NFC (near field communication) or other technologies. When the computer program is executed by the processor, a multi-task edge computing resource scheduling method is implemented. The display screen of the computer device can be a liquid crystal display screen or an electronic ink display screen, and the input device of the computer device can be a touch layer covered on the display screen, or a button, trackball or touchpad set on the housing of the computer device, or an external keyboard, touchpad or mouse.

[0044] This embodiment also provides a computer-readable storage medium on which a computer program is stored. When the program is executed by a processor, the method proposed in the above embodiment is implemented.

[0045] The storage medium proposed in this embodiment and the method proposed in the above embodiment belong to the same inventive concept. The technical details not fully described in this embodiment can be referred to the above embodiment, and this embodiment has the same beneficial effects as the above embodiment.

[0046] Through the above description of the implementation mode, the technicians in the relevant field can clearly understand that the present invention can be implemented by means of software and necessary general hardware, and of course can also be implemented by hardware, but in many cases the former is a better implementation mode. Based on such an understanding, the technical solution of the present invention is essentially or the part that contributes to the prior art can be embodied in the form of a software product, and the computer software product can be stored in a computer-readable storage medium, such as a computer's floppy disk, read-only memory (ReadOnly, Memory, ROM), random access memory (RandomAccess Memory, RAM), flash memory (FLASH), hard disk or optical disk, etc., including a number of instructions for a computer device (which can be a personal computer, a server, or a network device, etc.) to perform the method of the embodiment of the present invention.

[0047] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention rather than to limit it. Although the present invention has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that the technical solutions of the present invention may be modified or replaced by equivalents without departing from the spirit and scope of the technical solutions of the present invention, which should all be included in the scope of the claims of the present invention.

[0048] Those skilled in the art will appreciate that the embodiments of the present application may be provided as methods, systems, or computer program products. Therefore, the present application may adopt the form of a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware. Moreover, the present application may adopt the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program codes. The scheme in the embodiments of the present application may be implemented in various computer languages.

[0049] The present application is described with reference to the flowcharts and / or block diagrams of the methods, devices (systems), and computer program products according to the embodiments of the present application. It should be understood that each process and / or box in the flowchart and / or block diagram, as well as the combination of the processes and / or boxes in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to generate a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 A process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0050] These computer program instructions may also be stored in a computer-readable memory capable of directing a computer or other programmable data processing device to operate in a specific manner, so that the instructions stored in the computer-readable memory produce an article of manufacture comprising an instruction device, which implements the process Figure 1 A process or multiple processes and / or boxes Figure 1 A function specified in one or more boxes.

[0051] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operating steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing instructions for implementing the process. Figure 1 A process or multiple processes and / or boxes Figure 1 The steps for the functions specified in one or more boxes.

[0052] Although the preferred embodiments of the present application have been described, those skilled in the art may make other changes and modifications to these embodiments once they have learned the basic creative concept. Therefore, the appended claims are intended to be interpreted as including the preferred embodiments and all changes and modifications falling within the scope of the present application.

[0053] Obviously, those skilled in the art can make various changes and modifications to the present application without departing from the spirit and scope of the present application. Thus, if these modifications and variations of the present application fall within the scope of the claims of the present application and their equivalents, the present application is also intended to include these modifications and variations.

Claims

1. A multi-task edge computing resource scheduling method, characterized in that: include: Combining various parameters of new energy electric field equipment and the coverage of mobile edge computing, a new energy electric field mobile edge computing scheduling model is constructed; Based on the new energy electric field mobile edge computing scheduling model and the dependency relationship between new energy tasks, an ordered set of auxiliary task scheduling is constructed; The experience quality of the new energy electric field data collection system is maximized by migrating tasks to edge nodes, and the optimization problem of maximizing the experience quality is constructed under the conditions of satisfying the data task processing delay and energy constraints of the new energy electric field system; Lyapunov optimization is used to decouple the constraints of long-term migration costs in the optimization problem of maximizing the quality of experience. By constructing a virtual migration cost optimization queue, the problem is converted into an optimization problem without long-term constraints. A deep reinforcement learning algorithm is used to jointly optimize task scheduling and resource allocation for the optimization problem without long-term constraints to achieve resource scheduling for multi-task edge computing.

2. A multi-task edge computing resource scheduling method as claimed in claim 1, characterized in that: The construction of the new energy electric field mobile edge computing scheduling model includes: set up Represents the set of new energy electric field devices. Each device generates M computing tasks in each time slot. The task set is ,set up represents a collection of edge servers, each of which has computing resources and bandwidth resources , when n is within the communication coverage of the edge server, the service scheduling of the new energy electric field equipment n can be performed; Assume that the user schedules the task in time slot t as ,in Indicates the size of the scheduled task, Indicates the total number of CPU cycles required to complete the scheduled task. Indicates the task Maximum latency requirement; The system has N*M tasks, and the set of all tasks is Part of the task is executed on the new energy electric field equipment, and part is dispatched to the nearby edge server. For limited local computing resources, define is the ratio coefficient of task scheduling to edge servers, where .

3. A multi-task edge computing resource scheduling method as claimed in claim 2, characterized in that: The construction of ordered set auxiliary task scheduling includes: Set the task set received by the edge server to , the set of tasks performed locally by the new energy electric field equipment is ,in and It is an ordered set, and all tasks are executed in the order they are entered; Tasks are scheduled sequentially according to data type dependencies. When the new energy electric field device n schedules part of the tasks to the nearby edge server, the communication model is: The signal-to-noise ratio from the new energy electric field device n to the edge server k in time slot t is ; Transmission rate from new energy electric field device n to edge server k , where B k is the communication bandwidth resource from the new energy electric field equipment to the edge server, then the transmission delay caused by task scheduling is: ， The corresponding transmission energy consumption is ; The computing tasks of the new energy electric field equipment can be processed simultaneously on the local device and the edge server. The system computing model is: when the mth data of the new energy electric field equipment n is calculated locally, the local computing delay in time slot t is expressed as: ， The corresponding local energy consumption is ; When the mth data of the new energy electric field device n is dispatched to the edge server k, the computation delay generated by the edge server at time slot t is expressed as: ， The corresponding energy consumption is .

4. A multi-task edge computing resource scheduling method as claimed in claim 3, characterized in that: The method of maximizing the experience quality of the new energy electric field data acquisition system by migrating tasks to edge nodes includes: The new energy electric field device n dispatches its task to an edge server in time slot t through a direct cellular link. In the next time slot, the new energy electric field device n roams to the communication coverage of another edge node. represents the cost of migrating a task from one edge server to another edge server. The migration cost of the mth task service of the new energy electric field device n is expressed as ,in, is the decision result of service migration. When the edge server of time slot t is different from the edge server of time slot t−1, ,otherwise ; The local computation of data and task scheduling are performed in parallel, and the execution time of device n is: ， The total energy consumption of new energy electric field equipment n in time slot t is: ， The task of each new energy electric field equipment is C nm There is a certain execution time , let task C nm Completion time for ,in, Represents task C nm The start time of For Task C n and Task C n+1 The dependency relationship between tasks C n+1 Need to be in Task C n After completion, execute Task C n+1 Start time Should meet , i.e. task C n+1 The start time of task C is no earlier than n completion time.

5. A multi-task edge computing resource scheduling method as claimed in claim 4, characterized in that: The construction of the maximum experience quality optimization problem includes: The quality of experience (QoE) of a new energy electric field device n is determined by the task delay and energy consumption of the device. The following optimization objectives are defined: ， in, represents the QoE penalty factor of the new energy electric field device n at time slot t, involving task delay and local energy consumption, weighted parameter represents the QoE preference factor of the user in time slot t; The optimization variables of the system in time slot t are defined as , , , the optimization problem is formally expressed as: ， in, represents the maximum computing resources of device n, represents the delay requirement of the task, represents the used migration cost, represents the available migration cost, constraint C1 represents the equipment task scheduling ratio coefficient constraint, constraint C2 represents the equipment resource constraint, constraint C3 represents the equipment delay requirement, constraint C4 represents the service migration cost limit, and constraint C5 represents the execution dependency between the new energy electric field equipment tasks.

6. A multi-task edge computing resource scheduling method as claimed in claim 5, characterized in that: The problem is converted into an optimization problem without long-term constraints by constructing a virtual migration cost optimization queue, which includes: Design a virtual migration cost optimization queue and define the evolution representation of the virtual migration cost optimization queue of the n slot of the new energy electric field equipment at time slot t; The Lyapunov function is defined to describe the sum of squares of all virtual migration cost optimization queue backlogs during the tth time slot. To ensure the stability of the system, As conditional Lyapunov drift; The drift penalty algorithm is used to obtain the Lyapunov drift penalty function. , the Lyapunov queue optimization technique is used to transform the original optimization problem into an optimization problem without long-term constraints, and the objective function is expressed as: ， Among them, the positive control parameters V and Used to dynamically balance the trade-off between processing latency performance and migration cost consumption. represents the delay of the kth type of data of the nth node at time t.

7. A multi-task edge computing resource scheduling method as claimed in claim 6, characterized in that: The joint optimization of task scheduling and resource allocation for the optimization problem without long-term constraints using a deep reinforcement learning algorithm includes: The dynamic edge server scenario is described using a discrete-time Markov decision process (MDP). The state of the MDP is the communication state of the available edge servers around the new energy electric field equipment, and the energy consumption when the SE schedules tasks to the edge servers. According to the current system state, the mobile scheduling module takes action, and the agent performs actions based on the observed state and obtains rewards from the environment, where the rewards include the energy consumption of the system and the penalty for violating the delay constraint; The PPO algorithm consists of an action network and an evaluation network. The action network is divided into two parts, corresponding to the parameters and , the action network is based on the state Output Action and interact with the environment; the evaluation network calculates the state value based on the state information; To evaluate actions The performance of the advantage function is introduced , and calculate the objective function of the evaluation network, using the objective function of the evaluation network Back propagation updates the Critic network; The objective function of PPO constrains the update range by comparing the changes between the current strategy and the old strategy; Update the Actor network using backpropagation of the objective function.

8. A multi-task edge computing resource scheduling system, using a multi-task edge computing resource scheduling method as described in any one of claims 1 to 7, characterized in that: include: The scheduling model building module is used to build a new energy electric field mobile edge computing scheduling model by combining various parameters of the new energy electric field equipment and the mobile edge computing coverage; An ordered set construction module, used to construct an ordered set auxiliary task scheduling based on the new energy electric field mobile edge computing scheduling model and the dependency relationship between new energy tasks; An optimization problem construction module is used to maximize the experience quality of the new energy electric field data acquisition system by migrating tasks to edge nodes, and to construct a maximum experience quality optimization problem under the conditions of satisfying the data task processing delay and energy constraints of the new energy electric field system; An optimization problem conversion module, used for decoupling the constraints of long-term migration cost in the optimization problem of maximizing the quality of experience by using Lyapunov optimization, and converting the problem into an optimization problem without long-term constraints by constructing a virtual migration cost optimization queue; The resource optimization scheduling module is used to use a deep reinforcement learning algorithm to jointly optimize task scheduling and resource allocation for the optimization problem without long-term constraints, so as to realize resource scheduling of multi-task edge computing.

9. An electronic device, comprising a memory and a processor, characterized in that: The memory is used to store computer-executable instructions, and the processor is used to execute the computer-executable instructions. When the computer-executable instructions are executed by the processor, the steps of the method according to any one of claims 1 to 7 are implemented.

10. A computer-readable storage medium having computer-executable instructions stored thereon, characterized in that: When the computer executable instructions are executed by a processor, the steps of any method described in claims 1 to 7 are implemented.

Images