基于时间序列提取的数字化配电网异常检测方法  

Anomaly Detection Method of Digital Distribution Network Based on Time Series Extraction

作　　者：周亮 毛峻青 ZHOU Liang;MAO Junqing(State Grid Shanghai Economic Research Institute,Shanghai 200002,China)

机构地区：[1]国网上海市电力公司经济技术研究院,上海200002

出　　处：《微型电脑应用》2024年第12期301-304,共4页Microcomputer Applications

摘　　要：针对配电网电力数据异常会直接导致电力网络运行故障的问题,提出一种结合时间序列提取和维诺图异常检测(VOD)的数字化配电网异常电力数据检测算法,通过改进的分段线性表示(PLR)完成配电网电力数据降维和编号,并结合VOD实现电力数据异常检测。提取配电网电力数据时间序列后,正常序列和异常序列的关键点分别有28个和26个,正常序列和异常序列的缩减程度分别为13%和12%。对配电网电力额数据异常点检测表明,点序号6和18为异常点,相应样本序列段分别为120~162和312~348。所提出的数字化配电网电力数据异常检测方法具有较高的有效性,其能预测配电网信息物理系统异常。Aiming at the problem of the abnormal power data in distribution network may directly lead to power network operation failure,an abnormal power data detection algorithm of digital distribution network combined with time series extraction and Voronoi outlier detection(VOD)is proposed.The dimensionality reduction and numbering of distribution network power data are completed through the improved piecewise linear representation(PLR),and the abnormal power data detection is realized combined with VOD.After extracting the time series of distribution network power data,there are 28 and 26 key points of normal sequence and abnormal sequence,respectively,and the reduction degrees of normal sequence and abnormal sequence are 13%and 12%,respectively.The detection of abnormal points of distribution network power amount data shows that point No.6 and No.18 are abnormal points,and the corresponding sample sequence segments are 120 to 162 and 312 to 348,respectively.The proposed anomaly detection method of digital distribution network power data has high effectiveness,which can predict the anomaly of distribution network information physical system.

关 键 词：配电网 时间序列 维诺图 VOD 异常数据 

分 类 号：TM23[一般工业技术—材料科学与工程]