间隙统计在电力系统不良数据辨识中的应用  

Application of Gap Statistic in Identifying Bad Data of Power Systems

作　　者：吴京秋[1] 杨伟[2] 

机构地区：[1]南京工程学院工程基础实验与训练中心,江苏南京211167 [2]南京理工大学动力工程学院,江苏南京210094

出　　处：《南京工程学院学报（自然科学版）》2007年第3期28-33,共6页Journal of Nanjing Institute of Technology(Natural Science Edition)

摘　　要：将基于间隙统计(gap statistic)聚类算法的数据挖掘技术应用于电力系统不良数据的辨识中.将聚类内围绕聚类均值的欧氏平方距离和的对数与相应的参考值的数学期望做比较,对量测数据进行聚类个数的确定.仿真结果表明,与传统的状态估计方法相比,此方法可避免残差污染和残差淹没.The technology of data mining based on clustering-algorithm gap statistic was applied to the identification of bad data in power systems. By comparing the logarithm of Euclidean square distances around means clustering with mathematical expectation of corresponding reference values, the optimal number of clusters was accordingly determined. The simulation results indicated that gap statistic is superior to traditional state estimation as to preventing residual contamination and misidentification.

关 键 词：间隙统计 数据挖掘 不良数据辨识 聚类分析 

分 类 号：TM712[电气工程—电力系统及自动化]