电子设计工程
Electronic Design Engineering
第 28卷
Vol.28
第 18期
No.18
2020年 9月
Sep. 2020
收稿日期:2019-12-11 稿件编号:201912093
基金项目:国家电网公司科技项目(520932190025);国网上海电力公司管理咨询项目(62093216009G)
作者简介:蒋勇斌(1970—),男,上海人,工程师。研究方向:电力营销。
用户复杂用电作为衡量一个地区工业生产的重
要指标,与各个地区经济发展具有密切关系。在经
济快速增长期间,投资力逐渐加大,人们对电量需求
度增强,用电量随之上升
[1]
。用户复杂用电量如实反
映不同地区行业发展差异,为电力企业不同地区资
源配置优化奠定基础
[2]
。随着近几年用户用电量需
求的不 断 增 加,使 得 电力用 户 数据存 在 安 全问题 。
基于 k-means 聚类算法的用户复杂用电特征挖掘方法
研究
蒋勇斌,赵 炜,曹晶晶,周 丹
(国网上海市电力公司金山供电公司 上海 200540)
摘要:用户用电情况随着电网技术发展变得更加复杂,同时产生大量用电特征。以往采用基于神
经网络挖掘方法和基于 CURE 算法的挖掘方法受到噪声数据影响,导致挖掘精准度较低,针对该问
题,提出基于 k-means 聚类算法的用户复杂用电特征挖掘方法。在 k-means 聚类算法中,研究用户
复杂用电特征挖掘原理,并对数据进行清洗、集成、规约变换预处理,避免噪声干扰。利用信息熵
原则聚类矩阵规整特征点,根据复杂用电特征,通过簇类决策用电特征点,计算聚类簇之间距离,
获取用电特征信息增益,完成用户复杂用电特征挖掘。通过实验对比结果可知,该方法挖掘精准
度最高为 99%,为用户提供更好优质服务。
关键词:k-means 聚类;用户复杂用电;特征;挖掘
中图分类号:TN927.2 文献标识码:A 文章编号:1674-6236(2020)18-0011-05
DOI:10.14022/j.issn1674-6236.2020.18.003
Research on the mining method of complex power consumption characteristics
based on k⁃means clustering algorithm
JIANG Yong⁃bin,ZHAO Wei,CAO Jing⁃jing,ZHOU Dan
(State Grid Shanghai Jinshan Electric Power Supply Company,Shanghai 200540,China)
Abstract: With the development of power grid technology,the power consumption of users becomes
more complex,and a large number of power consumption characteristics are produced. In the past,the
mining methods based on neural network and cure algorithm are affected by noise data,which results in
low mining accuracy. To solve this problem,the k ⁃ means clustering algorithm is proposed. In the k ⁃
means clustering algorithm,we study the mining principle of complex power consumption characteristics
of users,and clean,integrate and preprocess the data to avoid noise interference. Using the principle of
information entropy to cluster matrix regular feature points,according to the complex power consumption
characteristics,the power consumption feature points are determined by cluster class,the distance
between clusters is calculated,the power consumption feature information gain is obtained,and the
complex power consumption feature mining of users is completed. The experimental results show that the
mining accuracy of this method is 99%,which provides better quality services for users.
Key words: k⁃means clustering;complex power consumption of users;features;mining
-- 11