基于函数型数据分析和k-means算法的电力用户分类（英文）  

Electricity Consumer Archetypes Study Based on Functional Data Analysis and K-Means Algorithm

作　　者：张欣[1] 高卫国[1] 苏运[2] 

机构地区：[1]复旦大学数学科学学院,上海市杨浦区200433 [2]上海市电力公司,上海市虹口区200437

出　　处：《电网技术》2015年第11期3153-3162,共10页Power System Technology

基　　金：Projected Supported by the National High Technology Research and Development Program of China(863 Program)(2015AA050203);National Talents Training Base for Basic Research and Teaching of Natural Science of China(J1103105)~~

摘　　要：为了对大量电力用户的稀疏、不规律的日耗电量数据进行特征分析,并对用户进行分类,文章提出一种函数性数据聚类分析方法。首先,应用kernel方法将离散的电量数据还原成连续曲线;然后,受Sobolev空间距离的启发,定义了新的函数距离,用于k-means算法进行聚类。以某城市10 000户居民538天的实际用电数据进行实验,得到了用户在不同距离和聚类个数下的聚类原型。实验结果显示,由于选取的用户主要是城市居民,其用电模式比较相似:大高峰时段主要在6—9月,小高峰时段主要在1—2月,日消耗波动较小。而不同用户类别的主要区别体现在用电量的范围上:低耗电用户整体低于13 k W?h/天,高耗电用户接近100 k W?h/天。In this paper, a functional cluster method to analyze features of sparse and irregular longitudinal electricity data and to cluster all users is proposed. Firstly, kernel method is applied to estimate daily continuous curves of discrete data. Then, inspired by distance in Sobolev space, new distances for functional data usable in k-means algorithm are proposed. Based on experiment, electricity consumer archetypes for all users in different functional distances and cluster numbers are calculated. Result shows that, because users in experiment are mainly city residents, their consumption patterns are similar： big peak period is between June and September, small peak is between January and February, and consumption fluctuations aren＇t very intense. However, main difference is in consumption ranges： low-consumption user＇s daily consumptions are lower than 13 k W？h overall, while high-consumption users use almost 100 k W？h every day.

关 键 词：函数性数据分析 K-MEANS kernel方法 智能电表 数据分析 

分 类 号：TM714[电气工程—电力系统及自动化]