基于改进K-means聚类算法的负荷建模及应用  

Load modeling based on improved K-means clustering algorithm and its application

作　　者：白雪峰[1] 蒋国栋[2] 

机构地区：[1]哈尔滨工业大学电气工程系,黑龙江哈尔滨150001 [2]中国电力科学研究院,北京100192

出　　处：《电力自动化设备》2010年第7期80-83,共4页Electric Power Automation Equipment

摘　　要：根据数据所处区域的数据密度选择初始聚类中心以充分反映数据的分布特征;在迭代计算新的聚类中心时,利用数据与类中心的平均距离作为新一轮迭代的聚类中心以排除噪声点的影响;通过集群评估指标DBI(Davies-Bouldin Index)选择最优聚类个数,以更准确地反映数据空间分布的特征。利用改进K-means算法对某省变电所数据进行聚类分析,得到负荷特征数据,并在此基础上建立了分段3次Hermite插值函数负荷模型。算例分析表明基于改进K-means的负荷建模方法对电网仿真分析可行、有效。The data density in specific area is used to select the initial clustering center to fully display its distribution feature and the average distance from data to clustering center is used in the iteration calculation for new clustering center to eliminate to effect of noise.The DBI（Davies-Bouldin Index） is used to select the optimal number of clustering to show the distribution feature of data space more exactly.The improved K-means algorithm is applied in the data clustering for a substation and the distribution feature of its load data is analyzed,based on which a cubic Hermite interpolating polynomial load model is piecewise established.The simulative results show that the load modeling based on improved K-means algorithm for power system simulation is feasible and effective.

关 键 词：电力系统 K-MEANS 聚类分析 负荷建模 

分 类 号：TM74[电气工程—电力系统及自动化]