基于MDS-WFCM的日负荷曲线聚类方法  

Clustering Method of Daily Load Curve Based on MDS-WFCM

作　　者：杨邓 杨俊杰 YANG Deng;YANG JUN-jie(School of Electronic and Information Engineering,Shanghai University of Electric Power,Shanghai 200090,China;Shanghai Dianji University,Shanghai 201306,China)

机构地区：[1]上海电力大学电子与信息工程学院,上海200090 [2]上海电机学院,上海201306

出　　处：《计算机仿真》2023年第4期103-107,294,共6页Computer Simulation

基　　金：国家自然科学基金(61202369,61401269,61572311);上海市科技创新行动计划地方院校能力建设项目(17020500900);上海市教育发展基金会和上海市教育委员会“曙光计划”(17SG51)。

摘　　要：电力负荷曲线聚类是挖掘电力消耗大数据的主要方法。为解决现有研究中聚类方法在运行效率、聚类质量等方面存在的不足,提出了一种基于多维标尺(MDS)及加权模糊C均值聚类(WFCM)的日负荷曲线聚类方法。首先使用MDS将负荷曲线数据降维处理,然后利用最小二乘法原理(LSM)确定降维指标数目以及改进熵权法确定指标权重,最后采用WFCM聚类方法对日负荷曲线聚类分析。算例分析表明,所提方法在聚类效率、质量以及算法鲁棒性方面均优于传统方法。Power load curve clustering is the main method for mining big data of power consumption.In order to solve the shortcomings of the existing clustering methods in terms of operation efficiency and clustering quality,this paper proposes a daily load curve clustering method based on multi-dimensional scale(MDS)and weighted fuzzy Cmeans clustering(WFCM).Firstly,the MDS was used to reduce the dimension of load curve data.Then,the least square method(LSM)was used to determine the number of dimension reduction indicators,and the improved entropy weight method was used to determine the weight of indicators.Finally,the WFCM clustering method was used to cluster analysis the daily load curve.Numerical examples show that the proposed method is superior to traditional methods in clustering efficiency,quality and algorithm robustness.

关 键 词：多维标尺 最小二乘法 改进熵权法 鲁棒性 

分 类 号：TP301.6[自动化与计算机技术—计算机系统结构]