基于局部离群点的低压台区线损异常检测  

Treatment of line loss anomaly in low voltage station area based on local outliers

作　　者：周慷 高林飞 陆惠惠 ZHOU Kang;GAO Lingfei;LU Huihui(State Grid Shanghai Electric Power Company,Shanghai 201600,China)

机构地区：[1]国网上海市电力公司松江供电公司,上海201600

出　　处：《自动化与仪器仪表》2022年第8期76-79,共4页Automation & Instrumentation

基　　金：国家自然科学基金(51877527)。

摘　　要：为提高低压台区线损异常检测方法的数据采集成功率与检测准确率,提出基于局部离群点的低压台区线损异常检测方法研究。分析低压台区基本结构,获取线损数据集,依据基于密度的局部离群算法完成离群数据点的定义。对数据集进行预处理,通过计算信息熵来判断其是否具有离群属性,通过计算其加权距离,计算各对象间的局部可达密度,继而通过计算各对象的离群因子,通过与离群因子阈值比较完成离群属性的判断,完成低压台区线损异常的检测。对比实验结果显示:该方法可大幅提高低压台区的线损异常检测的数据采集成功率与检测准确率,经过适当治理后其异常情况骤减,改善了该台区的线损情况,提高供电质量和供电效率。In order to improve the data collection success rate and detection accuracy of the line loss anomaly detection method in the low-voltage station area,a research on the line loss anomaly detection method in the low-voltage station area based on local outliers is proposed.The basic structure of the low-voltage station area is analyzed and the line loss data set is obtained.According to the density-based local outlier algorithm,the definition of outlier data points is completed.The data set is preprocessed,the information entropy is calculated to determine whether it has outlier attributes,the weighted distance is calculated,the local reachable density between each object is calculated,and then each object is calculated by calculating the weighted distance.The outlier factor is compared with the outlier factor threshold to complete the judgment of the outlier attribute,and complete the detection of abnormal line loss in the low-voltage station area.The results of comparative experiments show that this method can greatly improve the data acquisition success rate and detection accuracy of abnormal line loss detection in low voltage platform area.After proper treatment,the abnormal situation is sharply reduced,which improves the line loss situation in the station area,and improves the power supply quality and power supply efficiency.

关 键 词：局部离群点 低压台区 线损 异常 治理 供电台区 

分 类 号：TP207[自动化与计算机技术—检测技术与自动化装置] TM769[自动化与计算机技术—控制科学与工程]