Reinforcement Learning Based Resource Allocation For Cloud Edge Collaboration Fault Detection In Smart Grid

Smart power grid can meet the various demands of modern people on production and living,and improve the intelligent level of production and living.The normal operation of smart power grid is the basis of maintaining production and life.The fault monitoring of intelligent power grid becomes more and more important.With the development of artificial intelligence technology and cloud computing,the design of anomaly monitoring system based on deep learning with the powerful computing power of cloud has become the trend of future technology development.On the one hand,due to the large overall capacity of monitoring data,on the other hand,data transmission via the Internet has certain delay and delay jitter,which will lead to the cloud platform bandwidth load is too large,the real-time system is poor,and the delay model is not accurate and other problems.In view of these problems,this paper has done the following work:1.The current situation of fault detection and resource allocation research at domestic and abroad is summarized,and introduces the basic concept,composition,principle and application of deep reinforcement learning.2.Since the resource allocation problem studied in this paper is a combinatorial optimization problem,a solution framework of combinatorial optimization problem based on deep reinforcement learning is proposed.The solution framework adopts the deep reinforcement learning method.Under the framework of actor-critic,the model is trained by the pointer network and optimized by the strategy gradient method.Then the framework is used to solve the two-dimensional packet knapsack problem.3.A smart grid anomaly monitoring system assisted by edge computing is proposed.The embedded equipment assisted by lightweight neural network is placed near the edge of the monitored device to realize real-time monitoring of the sampled data.In addition,considering the limited communication,computing power and different detection accuracy of embedded devices,this paper proposes an optimal allocation method of communication and computing resources based on deep reinforcement learning,which greatly speeds up the solution time,maximizes the throughput of the system,and improves the communication efficiency and resource utilization of the system.Finally,through the analysis and comparison of the trained models,it is found that the algorithm in this paper can efficiently solve the resource scheduling problem,so as to realize the allocation and scheduling of resources in the case of real-time changes of Internet delay.