Research On Task Migration Modeling And Optimization Of Smart Grid Based On Edge Computing

With the vigorous development of smart grid construction,the power industry has gradually moved towards digitization and informatization.The traditional smart grid cloud data center is difficult to meet the requirements of the power industry and is not enough to fully support the digital transformation of the smart grid industry.As the key technology of 5G network in the future.Edge Computing(EC)will gradually sink the business to the edge of the network,greatly reducing the delay and energy consumption caused by processing tasks,and edge computing can provide a task migration platform for smart grid devices for computing and storage services.In the smart grid,task migration technology can migrate the complex tasks collected by the terminal equipment to the edge server through the wireless network,and complete the computing tasks with the help of the rich computing resources of the edge server,so as to solve the problem of limited computing power and battery capacity of the terminal equipment in the smart grid.In the architecture of edge computing,the edge server has faster data transmission rate and less energy consumption than the terminal equipment.It is necessary to design a reasonable task migration strategy to further improve the performance of the edge computing system.The existing traditional algorithms need complex operations to optimize the objectives,and reinforcement learning algorithm is very suitable for solving decision-making problems.The method based on deep reinforcement learning can dynamically find the optimal task migration scheme according to the current environmental state.Therefore,this paper studies the edge computing task migration strategy in smart grid.The main work is as follows:1.Aiming at the task migration strategy in the single edge server scenario,the system models are established in four aspects: scenario,network,local computing and task migration computing.Aiming at the problem model,a reinforcement learning algorithm Q learning is proposed to minimize the total energy consumption.Then,a new deep Q network(DQN)algorithm is formed by combining neural network with Q learning algorithm.Through simulation experiments,the proposed task migration strategy based on DQN algorithm can significantly reduce the total energy consumption of the system compared with Q learning algorithm in terms of different number of smart meters,different computing power of edge servers and different upload data sizes.2.Extend the system environment to the case of multi edge servers.The problem model is to minimize the weighted total cost of delay and energy consumption.On the basis of DQN,Actor network and Critical network are integrated into it to form a task migration optimization algorithm based on deep deterministic policy gradient(DDPG).According to the actual problem,set the state,action and feedback reward value for the algorithm.Finally,simulation experiments are carried out in Python platform to analyze the convergence performance,and the effects of task data upload,the number of smart meters and the computing power of edge servers on the weighted total overhead are compared.The proposed task migration optimization algorithm based on DDPG can get a better migration strategy and reduce the total weighted overhead of the system by comparing DQN algorithm,full local computing,full migration computing and other algorithms in the above different cases.