基于数据划分的工业电力负荷曲线聚类研究  

Research on Industrial Power Load Curve Clustering Based on Data Division

作　　者：潘明明[1] 田世明[1] 魏娜 赵嵩正[2] 王莉芳[2] 吴磊[3] Pan Mingming;Tian Shiming;Wei Na;Zhao Songzheng;Wang Lifang;Wu Lei(China Electric Power Research Institute, Beijing 100192, China;School of Management, Northwestern Polytechnical University, Xi'an Shaanxi 710072, China;State Grid Tianjin Electric Power Co., Tianjin 300010, China)

机构地区：[1]中国电力科学研究院,北京100192 [2]西北工业大学管理学院,陕西西安710072 [3]国网天津市电力公司,天津300010

出　　处：《电气自动化》2019年第4期24-26,67,共4页Electrical Automation

基　　金：面向随机性电源的多元负荷主动响应及预测控制技术研究与应用(SGTJDK00DWJS1700034)

摘　　要：工业电力用户作为电力需求大客户,为了对其负荷曲线聚类,研究其负荷模式,基于某生态城工业电力用户负荷曲线数据,比较了划分聚类和层次聚类中较常用和传统的五类算法,提出了基于数据划分的层次聚类算法,分析了基于数据划分的层次聚类、传统层次聚类和划分聚类算法的差异。运用基于数据划分的层次聚类算法对该生态城工业电力用户负荷曲线聚类。研究表明:基于数据划分的层次聚类算法与传统层次聚类算法相比,可以用较短的聚类时间得到较好的聚类结果;与K-means算法相比,当数据规模较小时,聚类时间增加,但聚类质量大幅度提升;当数据规模较大时,该算法比K-means算法的聚类时间更短。To cluster load curves of industrial power consumers as heavy buyers of electricity, this paper studied their load mode. Based on the data acquired from the load curves of industrial electricity consumers in a certain eco-friendly area, it compared five commonly used traditional algorithms belonging to partition clustering and hierarchical clustering categories, proposed hierarchical clustering algorithm based on data division, and analyzed the difference among the hierarchical clustering based on data division, traditional hierarchical clustering and partition clustering. Hierarchical clustering algorithm based on data division was used to cluster the load curves of the industrial electricity consumers in the eco-friendly area. The research indicated that, compared with traditional hierarchical clustering algorithm, the hierarchical clustering algorithm based on data division could achieve better clustering results within shorter clustering time. In comparison with K-means algorithm, in the case of small data scale, clustering time would increase while clustering quantity would be improved greatly, and in the case of big data scale, this algorithm would need a shorter clustering time than K-means algorithm.

关 键 词：负荷曲线 数据划分 层次聚类 划分聚类 工业用户 

分 类 号：TM714[电气工程—电力系统及自动化]