面向分布式光伏状态实时感知的边缘缓存与计算策略  

Edge Caching and Computing Strategy for Distributed Photovoltaic Real-Time Sensing

作　　者：崔炳荣 李德建 刘亮[1] 蒋名扬 孙毅 陈恺[2] CUI Bingrong;LI Dejian;LIU Liang;JIANG Mingyang;SUN Yi;CHEN Kai(Beijing Smart-Chip Microelectronics Technology Co.,Ltd.,Beijing 100089,China;School of Electrical and Electronic Engineering,North China Electric Power University,Beijing 102206,China;School of Integrated Circuits,Tsinghua University,Beijing 100084,China)

机构地区：[1]北京智芯微电子科技有限公司,北京100089 [2]华北电力大学电气与电子工程学院,北京102206 [3]清华大学集成电路学院,北京100084

出　　处：《智慧电力》2023年第8期67-74,共8页Smart Power

基　　金：国家重点研发计划资助项目(2022YFB2402901);北京智芯微电子科技有限公司实验室开放基金项目资助(SGSC0000SJQT2207178)。

摘　　要：针对分布式光伏部署分散、电力通信网难以支撑光伏业务数据实时接入与处理的问题,提出基于双层边缘计算的分布式光伏数据在线处理策略。首先,构建数据接入、边缘缓存与计算模型,并构建长期网络服务开销最小化的目标函数。其次,针对分布式光伏数据规模随机性强、优化函数难以直接求解问题,提出基于随机对偶次梯度法的在线任务卸载与代码缓存策略,通过将多时隙耦合问题解耦为单时隙独立问题,以在线得到近似最优解。最后,算例验证所提策略的数据处理速率优化性能。To cope with the problem that current electric communication network cannot support the real-time sensing and execution of distributed PV data,an online data execution strategy based on two-tier edge computing is investigated.Firstly,data access,edge caching and computing model are proposed,an objective function is formulated to minimize the long-term network service cost.Then,for the network is stochastic and the multiple-time slots problem cannot be solved directly,a stochastic dual-subgradient method-based online code caching and task offloading strategy is proposed,which can decompose the multi-time slots problem into many independent single time slot problems for online obtaining approximate optimal solution.Finally,simulations are given to verify the data execution effectiveness of the proposed method.

关 键 词：代码缓存 5G边缘计算 分布式光伏 数据在线处理 随机对偶次梯度法 

分 类 号：TM929.5[电气工程—电力电子与电力传动]