基于多变量自动回归的电力大数据异常值检测平台设计  

Design of Power Big Data Outlier Detection Platform Based on Multivariate Automatic Regression

作　　者：刘涛[1] 李英俊[1] 邢峰[1] 乔斌强 刘斌[1] LIU Tao;LI Ying-jun;XING Feng;QIAO Bin-qiang;LIU Bin(Ulanqab Electric Power Supply Bureau,Inner Mongolia,Ulanqab 012000 China)

机构地区：[1]乌兰察布电业局,内蒙古乌兰察布012000

出　　处：《自动化技术与应用》2022年第10期93-96,共4页Techniques of Automation and Applications

基　　金：内蒙古电力(集团)有限责任公司2019年科技项目(WD-ZXZB-2019-SC0402-1273)。

摘　　要：当前电力大数据异常值检测方法存在检测准确率低、耗时较长等问题,因此提出基于多变量自动回归的电力大数据异常值检测平台的设计方法。首先构建由数据源、数据收集、实时计算、数据管道、数据存储以及数据层组成的电力数据异常值检测平台。通过多变量自动回归方法完成电力数据挖掘以及异常数据的检测,并根据基本数据和异常数据线性组合生成电力数据评价值,并对评价值进行累积分布处理,最后利用高斯分布解决电力数据异常值稀疏的问题,实现数据异常值检测。实验结果表明,所设计平台检测准确率较高且耗时较短,具有可靠性。The current power big data outlier detection methods have the problems of low detection accuracy and long time-consuming.Therefore, a design method of power big data outlier detection platform based on multivariable automatic regression is proposed.Firstly, a power data outlier detection platform is constructed, which consists of data source, data collection, real-time calculation,data pipeline, data storage and data layer. The power data mining and abnormal data detection are completed by multivariable automatic regression method, and the evaluation value of power data is generated according to the linear combination of basic data and abnormal data, and the evaluation value is processed by cumulative distribution. Finally, the problem of sparse outliers in power data is solved by using gaussian distribution, and outliers detection is realized. The experimental results show that the designed platform has high detection accuracy, short time consumption and reliability.

关 键 词：多变量自动回归 大数据 异常数据 

分 类 号：TP311.13[自动化与计算机技术—计算机软件与理论] TP393.08[自动化与计算机技术—计算机科学与技术]