文章摘要
林君豪,张焰,祝锦舟,等.基于宏微观特征分层聚类的配电网拓扑相似性分析方法[J].电力系统自动化,2019,43(13):80-88. DOI: 10.7500/AEPS20181017006.
LIN Junhao,ZHANG Yan,ZHU Jinzhou, et al.Topology Similarity Analysis Method for Distribution Network Based on Hierarchical Clustering of Macroscopic and Microscopic Features[J].Automation of Electric Power Systems,2019,43(13):80-88. DOI: 10.7500/AEPS20181017006.
基于宏微观特征分层聚类的配电网拓扑相似性分析方法
Topology Similarity Analysis Method for Distribution Network Based on Hierarchical Clustering of Macroscopic and Microscopic Features
DOI:10.7500/AEPS20181017006
关键词: 配电网拓扑相似性  宏微观特征  核密度估计  Kullback-Leibler散度  改进分层聚类
KeyWords: topology similarity of distribution network  macroscopic and microscopic features  kernel density estimation  Kullback-Leibler divergence  improved hierarchical clustering
上网日期:2019-04-26
基金项目:国家高技术研究发展计划(863计划)资助项目(2015AA050203)
作者单位E-mail
林君豪 上海交通大学电子信息与电气工程学院, 上海市 200240 linjunhao@sjtu.edu.cn 
张焰 上海交通大学电子信息与电气工程学院, 上海市 200240  
祝锦舟 上海交通大学电子信息与电气工程学院, 上海市 200240  
赵腾 全球能源互联网发展合作组织, 北京市 100031  
苏运 国网上海市电力公司, 上海市 200437  
摘要:
      提出一种基于宏观与微观拓扑特征分层聚类的配电网拓扑相似性分析方法。首先,构建涵盖宏微观拓扑特征的指标集作为拓扑相似性分析的依据,包括宏观层面的网络社团特性和节点度相关性指标以及微观层面的线路功率传输性能和负荷分布密度指标;针对传统拓扑相似性分析方法难以对变量数不同的微观拓扑特征进行细粒度分析的问题,提出基于核密度估计的微观拓扑特征描述方法和基于Kullback-Leibler散度的微观特征相似性量度方法;再以基于标准离差率改进的谱聚类算法对宏微观拓扑特征集进行分层聚类,实现配电网拓扑相似性分析。算例分析表明,所提出的配电网拓扑相似性分析方法在表征配电网结构特性的能力及相似性分析效果方面相对于传统方法更优。
Abstract:
      A topology similarity analysis method for distribution network based on hierarchical clustering of macroscopic and microscopic topology features is proposed. Firstly, the index set covering macroscopic and microscopic topology features is constructed as a basis for topology similarity analysis, including network community characteristics and node degree correlation indices at the macroscopic level, and line power transmission performance and load distribution density indices at the microscopic level. Since it is difficult for traditional topology similarity methods to make fine-grained analysis on microscopic topology features with different numbers of variables, a microscopic topology feature description method based on kernel density estimation and a microscopic feature similarity measurement method based on Kullback-Leibler divergence is proposed. The improved spectral clustering algorithm based on coefficient of variation is used for hierarchical clustering of macroscopic and microscopic topology feature sets, to realize topology similarity analysis for distribution network. Case study shows that the proposed topology similarity analysis method of distribution network is superior to the traditional method in the ability of representing structural characteristics of distribution networks and the effect of similarity analysis.
查看全文(Free!)   查看附录   查看/发表评论  下载PDF阅读器