输电线路电磁环境试验数据清理方法研究  

Data Cleaning Method of Electromagnetic Environment of Power Transmission Line

作　　者：付殷 余占清[1] 曾嵘[1] 刘磊 李敏 庄池杰[1] 

机构地区：[1]清华大学电机工程与应用电子技术系电力系统及发电设备控制和仿真国家重点实验室,北京100084 [2]南方电网科学研究院有限责任公司,广州510663

出　　处：《高电压技术》2018年第1期289-295,共7页High Voltage Engineering

摘　　要：在电磁环境试验过程中，由于周围环境会出现各种突发性干扰，因此试验数据中常常存在异常数据，需要恰当地处理方法清理异常数据。为此，针对电磁环境试验数据的自动清理方法开展研究，采用离群点空间距离实现异常值判别，根据电磁环境数据特性，提出了适合于电磁环境数据的空间距离定义方法。提出了基于聚类的电磁环境试验数据清理方法，通过对空间离群点的检测完成对电磁环境试验数据的预处理过程。结果表明，得到的处理后数据中，异常数据均得到了清除，且未改变数据的统计特征值，其中异常数据主要表现为各个数据子类中的空间离群点；该算法可提高测量数据的可信度。In the process of testing electromagnetic environment, outliers in the data sheet always need to clean because of the burst interference occurring in the test environment. Consequently, we studied an automatic data cleaning method of electromagnetic environment with the outliers detection method based on the distance. Considering the test device, the characteristics of data statistics, and the physical characteristics of the electromagnetic environment, we redefined the method of calculating the spatial distance, and put forward the outlier detection clustering. The preprocessing of the elec- tromagnetic environment test data was finished by this method. The results show that outliers have been cleared in the processed database, not changing the statistical features of the data, and the cleaned data are mainly considered as spatial outliers in each subclass. This algorithm can improve the reliability of measurement data.

关 键 词：电磁环境 试验线段 数据清理 聚类分析 离群点检测 

分 类 号：TM75[电气工程—电力系统及自动化]