基于多视角数据与社区发现的典型用电负荷模式挖掘研究  

Research on Mining Typical Power Load Patterns Based on Multi-view Data and Community Detection Method

作　　者：魏伟 韩颖 刘怡君 张伟[3] WEI Wei;HAN Ying;LIU Yijun;ZHANG Wei(School of Management,Zhengzhou University,Zhengzhou 450001,China;Business School,the University of Melbourne,Victoria,Melbourne 3010,Australia;School of Ecology and Environment,Zhengzhou University,Zhengzhou 450001,China)

机构地区：[1]郑州大学管理学院,河南郑州450001 [2]墨尔本大学商学院,澳大利亚维多利亚州墨尔本3010 [3]郑州大学生态与环境学院,河南郑州450001

出　　处：《郑州大学学报（理学版）》2023年第4期68-74,共7页Journal of Zhengzhou University:Natural Science Edition

基　　金：河南省哲学社会科学规划项目(2020CZH009);国家自然科学基金项目(72001191);中国科技智库青年人才计划项目(20220615ZZ07110090)。

摘　　要：电力系统为满足各类用户的用电需求,需掌握用户端用电负荷曲线的典型模式。当前电力负荷模式挖掘过程中往往使用单视角数据进行聚类分析,未考虑不同视角数据内部度量信息差异和复杂的计算。针对这一问题,从三个粒度获取多视角用电负荷数据并进行预处理,分别计算相似度矩阵,使用相似度网络融合算法构建融合相似度矩阵,选用Leiden社区发现算法对用户群体进行划分,然后动态识别各个子社区的典型用电负荷曲线并进行趋势分析。实验结果表明,算法能够稳定划分用户群体并识别各个子社区的典型用电负荷模式曲线,结合用户基本信息,得到各个子社区的峰值模式,从而为电力系统采取个性化措施以满足不同用户群体的需求提供指导。To meet the electricity demand of various customers,the power system should understand the typical pattern of the user′s power load curve.While the single-view data was often used for cluster analysis,which could not consider the difference in internal measurement and complex calculations.The multi-perspective electricity load data was collected from three granular perspectives and processed pretreatment.The similarity matrix was first calculated and the fusion similarity matrix was constructed by the similarity network fusion algorithm.The Leiden algorithm was used to divide the user groups and identify the typical power load curve of each sub-community dynamically,the trend analysis was followed.The experimental results demonstrated that the sub-community could be classified and the typical power load curve could be identified reliably by the algorithm.The peak pattern for each sub-community could be found when combined with the basic information of users and thus could provide guidance for power system to take personalized measures to meet the needs of different user groups.

关 键 词：多视角数据 社区发现 相似度网络融合 模式识别 典型用电负荷曲线 

分 类 号：TM714[电气工程—电力系统及自动化]