基于软件定义网络的电网边缘计算资源分配

针对电网节点中边缘服务器资源受限问题,提出了一种基于软件定义网络的边缘计算框架。采用深度确定性策略梯度(Deep Deterministic Policy Gradient, DDPG)强化学习算法对边缘服务器中的计算、存储资源进行合理分配。首先建立应用于电网中的基于软件定义网络的边缘计算模型,得到边缘服务器计算存储资源及任务时延的约束条件,分析得到需处理的MINIP问题。使用Tensorflow搭建仿真环境并执行强化学习算法,实现电网边缘节点对边缘服务器存储计算资源的最优利用。结果表明,强化学习算法中的回复值呈上升趋势,采用DDPG进行分配的系统总延时更低。

SDN-based MEC resource allocation of a power grid

To solve the problem of limited resources of edge servers in grid nodes, an edge computing framework based on a software-defined network is proposed. The Depth Deterministic Strategy Gradient (DDPG) reinforcement learning algorithm is used to allocate the computing and storage resources reasonably in edge servers. First, an edge computing model applied into the software-defined grid network is established. This is to obtain the edge node computing and cache resources and task delay constraints, in order to analyze the MINIP which needs to be solved. TensorFlow is used to build a simulation environment and execute the reinforcement learning algorithm to realize the optimal utilization of edge server storage computing resources by grid edge nodes. It shows that the rewards of reinforcement learning increase, and the total delay of the system using DDPG allocation is lower.This work is supported by the Science and Technology Project of State Grid Corporation of China (No. (SGHEXTOOGCJS2000167) and the National Natural Science Foundation of China (No. 61971190).