采用双层聚类的区域用电形态分析

Analysis of Regional Electricity Consumption Using Bi-level Clustering

应用大数据技术对用户信息进行用电行为分析是电力公司开展精细化服务的基础。为此,提出一种基于双层聚类的区域用电形态分析方法。首层聚类利用SAX降维结合马尔科夫转移矩阵对电力用户的用电行为规律性进行区分,内层的自适应聚类再利用余弦相似度作为聚类指标分别对两类用户的典型日负荷曲线做进一步划分。最后,将对聚类的结果展开分析,通过与用户数据集中行业信息的比对以及与其他聚类方法聚类效果的比较,验证算法的有效性。

The use of big data technology to analyze the power consumption behavior of user information is the basis for power companies to carry out refined services. This paper proposes a method of regional power morphology analysis based on bi-level clustering. The first-level clustering uses SAX dimensionality reduction combined with Markov transition matrix to distinguish the regularity of power users ＇ electrical behavior. The inner layer adaptive clustering reuse cosine similarity is used as a clustering indicator,which is the typical day for two types of users,and the load curve is further divided. Finally,the results of clustering are analyzed,and the validity of the algorithm is verified by comparison with the industry data in user data sets and the clustering results with other clustering methods.