基于改进快速密度峰值聚类算法的电力大数据异常值检测分析  

Abnormal value detection and analysis of power big data based on improved fast density peak clustering algorithm

作　　者：杨峰 刘胜强 YANG Feng;LIU Shengqiang(Foshan Power Supply Bureau,Guangdong Power Grid Co.,Ltd.,Foshan 528000,China)

机构地区：[1]广东电网有限责任公司佛山供电局,广东佛山528000

出　　处：《电子设计工程》2022年第3期113-116,121,共5页Electronic Design Engineering

摘　　要：针对传统聚类算法对于大数据背景下大量电力大数据异常监测过程中存在的问题,提出在电力大数据异常值检测中的快速密度峰值聚类算法的改进。对传统基于密度峰值空间聚类方法进行分析,得到传统算法在使用过程中的问题。提出了快速密度峰值聚类算法的改进,对自适应参数与聚类中心自动的选择,通过标准化局部密度与距离对大数据异常值进行评测,能够得出异常点。对本文所设计的聚类算法实现算例分析,通过算例分析表示,本文所设计的算法能够满足实际用户的需求,提高电力大数据异常值的检测精准度。Aiming at the problems of traditional clustering algorithm in the process of abnormal monitoring of a large number of power big data under the background of big data,an improved fast density peak clustering algorithm in abnormal value detection of power big data is proposed.The traditional density peak space clustering method is analyzed to get the problems in the use of the traditional algorithm.The improvement of fast density peak clustering algorithm is proposed,the adaptive parameters and clustering center are automatically selected,and the outliers of big data are evaluated by standardizing local density and distance.An example of the clustering algorithm designed in this paper is analyzed,the example analysis shows that the algorithm designed in this paper can meet the needs of actual users and improve the detection accuracy of abnormal values of power big data.

关 键 词：电力大数据 异常值检测 密度峰值聚类 聚类算法 

分 类 号：TN98[电子电信—信息与通信工程]