基于谱聚类的用电采集网络拓扑推断算法  

Logical Topology Inference Algorithm for Electrical Acquisition Visual Network Based on Spectral Clustering

作　　者：张小博 王婷 佟芳 徐铁军 李晖 秦浩 ZHANG Xiaobo;WANG Ting;TONG Fang;XU Tiejun;LI Hui;QIN Hao(Information &Communication Company,State Grid Qinghai Electric Power Company,Xining 810008,China)

机构地区：[1]国网青海省电力公司信息通信公司,青海西宁810008

出　　处：《电力信息与通信技术》2018年第12期39-45,共7页Electric Power Information and Communication Technology

基　　金：国网青海省电力公司科技项目"基于大数据分析的业务虚拟专用通道监管平台研究与应用"(281734)

摘　　要：针对用电采集系统使用的业务虚拟通道拓扑结构不可见、难以监测用采终端所在的网络链路问题,提出一种基于谱聚类的虚拟网络拓扑推断算法。算法将拓扑推断问题转化为顶点集划分的优化问题,利用往返时延协方差度量终端节点间链路相关性,用Normalized Cut谱聚类自底向上逐层聚类并选择最优聚类数目,实现虚拟网络拓扑结构推断。仿真实验结果与真实数据测试结果显示,算法能够高效准确地推测虚拟网络的逻辑拓扑结构,为用电采集网络的监测提供了必要的技术手段。In electricity acquisition system,the network topology is transparent for customers,which makes it difficult to monitor business terminals and the network links by users.Aiming at this problem, this paper proposes a virtual network logical topology inference algorithm based on spectral clustering, which takes use of round trip delay to measure the relativity between business terminal pairs,and then uses the normalized cut spectral clustering method to infer the logical topology of business virtual network layer by layer and obtains the most optimized clusters number.Both of the simulation experimental result and the actual network result demonstrate that the proposed algorithm can infer the logical topology accurately and efficiently.It provides necessary technical means for the monitoring of electricity acquisition system.

关 键 词：逻辑拓扑推断 虚拟网络 谱聚类 往返时延 网络层析成像 

分 类 号：TP393[自动化与计算机技术—计算机应用技术]