东北大学学报:自然科学版 ›› 2020, Vol. 41 ›› Issue (8): 1070-1074.DOI: 10.12068/j.issn.1005-3026.2020.08.002

• 信息与控制 • 上一篇    下一篇

基于AP-LOF离群组检测的配电网连接验证

司方远1, 韩英华2, 赵强3, 汪晋宽1   

  1. (1.东北大学 信息科学与工程学院, 辽宁 沈阳110819;2.东北大学秦皇岛分校 计算机与通信工程学院, 河北 秦皇岛066004;3.东北大学秦皇岛分校 控制工程学院, 河北 秦皇岛066004)
  • 收稿日期:2019-08-24 修回日期:2019-08-24 出版日期:2020-08-15 发布日期:2020-08-28
  • 通讯作者: 司方远
  • 作者简介:司方远(1992-),男,河北迁安人,东北大学博士研究生; 韩英华(1979-),女,吉林白山人,东北大学教授; 汪晋宽(1957-),男,辽宁沈阳人,东北大学教授,博士生导师.
  • 基金资助:
    国家重点研发计划项目(2016YFB0901900); 国家自然科学基金资助项目(U1908213); 中央高校基本科研业务费专项资金资助项目(N182303037); 东北大学秦皇岛分校校内基金资助项目(XNB201803).

Verification of Distribution Network Connectivity Based on AP-LOF Outlier Group Detection

SI Fang-yuan1, HAN Ying-hua2, ZHAO Qiang3, WANG Jin-kuan1   

  1. 1.School of Information Science & Engineering, Northeastern University, Shenyang 110819, China; 2.School of Computer & Communication Engineering, Northeastern University at Qinhuangdao, Qinhuangdao 066004, China; 3.School of Control Engineering, Northeastern University at Qinhuangdao, Qinhuangdao 066004, China.
  • Received:2019-08-24 Revised:2019-08-24 Online:2020-08-15 Published:2020-08-28
  • Contact: HAN Ying-hua
  • About author:-
  • Supported by:
    -

摘要: 现有配电网连接验证工作将可疑异常值视为具有二元属性的独立个体,因此难以有效识别和验证具有高度内在相关性的局部离群组.针对这一问题,提出了基于AP-LOF离群组检测的配电网连接验证方法.通过引入近邻传播(affinity propagation,AP)聚类方法,将待校验台区用户聚类为多簇,并基于局部离群因子(local outlier factor,LOF)算法对所有簇心进行离群点检测,从而准确识别出台区内的离群组用户.以某电力公司实际用户电压数据进行算例分析,结果证明了AP-LOF算法在配电网连接验证中的适用性和有效性.

关键词: 电压数据分析, 配电网连接验证, 局部离群组检测, 近邻传播聚类, LOF算法

Abstract: In the existing methods for the verification of distribution network connectivity, the suspicious outliers are usually regarded as independent individuals with binary attributes, which is difficult to effectively identify and validate local outlier groups which are correlated with each other. Therefore, a verification method for distribution network connectivity is proposed based on AP-LOF outlier group detection. Users are clustered into multiple clusters by introducing affinity propagation (AP) clustering, and all of the cluster centers are then detected by the local outlier factor (LOF) algorithm. In this way, the outlier groups can be accurately identified. The actual user voltage data of a power company are used in the case study, and the results demonstrate the applicability and effectiveness of the AP-LOF algorithm in the verification of distribution network connectivity.

Key words: voltage data analysis, distribution network connectivity verification, local outlier group detection, affinity propagation clustering, LOF algorithm

中图分类号: