摘要
提出一种双层聚类的分析方法,分别利用K-means和余弦相似度来进行内层和外层聚类。算法能有目的地选取初始聚类中心,并且能自动确定最优聚类数,解决了随机选取初始聚类中心所造成的问题。另外针对高维负荷曲线双层聚类计算速度慢的问题,在外层聚类完成后先降维,再进行内层聚类。对比应用了降维技术的双层聚类算法和K-means算法的信息损失和计算速度。算例仿真结果表明,双层聚类在降维技术上的应用效果优于单纯的K-means算法,结合主成分分析降维的双层聚类算法可以取得最佳效果和较快的聚类速度。
A method of two-layer clustering is proposed,which uses K-means and cosine similarity to cluster the inner and outer layers respectively.The algorithm can select the initial clustering center intentionally,and can automatically determine the optimal clustering number,which solves the problem of randomly selecting the initial clustering center.In addition,Aiming at the problem that the two-layer clustering of the high-dimensional load curve is slow,it is dimensioned after the outer cluster is completed and then the inner layer is clustered.The information loss and calculation speed of the two-layer clustering algorithm and K-means algorithm with reduced dimensionality are compared.The simulation results show that the double layer clustering is better than the simple K-means algorithm in the dimension reduction technique,and the double-layer clustering algorithm with reduced principal component analysis can achieve the best effect and faster clustering speed.
作者
宁光涛
陈明帆
林强
周航
黄亮
高玉洁
NING Guangtao;CHEN Minfan;LIN Qiang;ZHOU Hang;HUANG Liang;GAO Yujie(HaiNan Power Grid Corporation, Hainan Haikou,57020)
出处
《自动化与仪器仪表》
2018年第5期18-23,共6页
Automation & Instrumentation
关键词
双层聚类算法
降维技术
电力用户分类
bilayer clustering method
dimensionality reduction
power user classification
作者简介
宁光涛（1961-），男，本科，高级工程师，主要从事电网规划及生产运行研究工作。
