基于随机矩阵理论的数据建模方法  

Data Modeling Method Based On Random Matrix Theory

作　　者：连鸿波 陈旸 李家睿 LIAN Hongbo;CHEN Yang;LI Jiarui(Songjiang Power Supply Company,State Grid Shanghai Power Supply Company,Shanghai 201617,China)

机构地区：[1]国网上海市电力公司松江供电公司,上海201617

出　　处：《电工技术》2021年第13期69-73,共5页Electric Engineering

摘　　要：随机矩阵理论作为统计学处理复杂系统和高维数据的理论工具,已越来越多地与大数据技术结合,应用于电力系统分析。首先介绍了随机矩阵的基础定义,其次介绍了随机矩阵单环定理,并依此建立了DAS量测数据的随机矩阵模型,采用平均谱半径作为其线性特征值统计量,根据平均谱半径的变化感知原始随机矩阵中数据的变化。采用IEEE 33节点配电网络作为仿真分析模型,可发现当配电网无故障稳定运行时平均谱半径随时间序列的变化情况,为后续探究配电网故障的诊断方法打下了理论基础。As a theoretical tool for statistically processing complex systems and high-dimensional data,random matrix theory has been increasingly combined with big data technology and applied to power system analysis.Firstly,the basic definition of random matrix is introduced.Secondly,the random matrix single ring theorem is introduced,and a random matrix model of DAS measurement data is established based on this.The average spectral radius is used as its linear eigenvalue statistic,and the data changes in the original random matrix are perceived according to the change of the average spectral radius.The IEEE 33-node power distribution network is used as the simulation analysis model.It is found that the average spectral radius changes with time series when the distribution network is running stably without faults,which lays a theoretical foundation for the subsequent exploration of the diagnosis method of the distribution network fault.

关 键 词：随机矩阵 配电网 数据建模 特征谱半径 

分 类 号：TM933[电气工程—电力电子与电力传动]