一种基于语义聚类的典型日负荷曲线选取方法  

A semantic clustering method for selecting the typical day load curve

作　　者：孟令奎[1] 段红伟[1] 黄长青[1] 孙琤 

机构地区：[1]武汉大学遥感信息工程学院,湖北武汉430079 [2]中兵勘察设计研究院,北京100053

出　　处：《华北电力大学学报（自然科学版）》2013年第1期43-48,共6页Journal of North China Electric Power University：Natural Science Edition

基　　金：国家科技支撑计划项目(2011BAH16B08)

摘　　要：将典型日负荷曲线的选取问题转化为基于统计学习的多元分类问题,利用概率潜在语义分析模型(PLSA)进行问题求解。方法首先通过K均值聚类和负荷曲线时段划分形成观测特征词和目标文档,通过阈值计算获得特征词-目标共生矩阵;然后基于Davies-Bouldin指标计算PLSA模型的最佳主题数目,并对模型参数求解获得每个目标文档中特征词的潜在主题;最后依据电力负荷曲线与特征词的对应关系形成新的聚类,并采用选取策略获得各聚类的典型日。实验表明,方法能够较好的反映节假日、气候等因素的影响,典型日选取合理可行。A method of transforming the typical day load curve selection problem into the multiple classification problems based on statistical learning is proposed. The Probabilistic Latent Semantic Analysis （PLSA） is used to solve the problem. Firstly, observed characteristic words and target documents are formed by K mean clustering and load curves＇ division, and the characteristic words-target co-occurrence matrix is obtained based on threshold calculation; secondly, based on the Davies-Bouldin index, the best topic number of PLSA model is calculated, and the model＇s parameters are solved to get the potential topic of each characteristic word in target documents; finally, on the basis of the correspon- dence between the load curves and characteristic words, new clusters are formed, then the typical days of each cluster are selected by using strategies. The experiment shows that this method can better reflect the factor effect, such as holidays, climate and other factors, and the typical daily selection is reasonable and feasible.

关 键 词：概率潜在语义分析模型 典型日负荷曲线 Davies—Bouldin指标 

分 类 号：TM76[电气工程—电力系统及自动化]