Clustering Methods for Electricity Consumers: An Empirical Study in Hvaler - Norway


  • University of Oslo
  • Østfold University College
  • NTNU


The development of Smart Grid in Norway in specific and Europe/US in general will shortly lead to the availability of massive amount of fine-grained spatio-temporal consumption data from domestic households. This enables the application of data mining techniques for traditional problems in power system. Clustering customers into appropriate groups is extremely useful for operators or retailers to address each group differently through dedicated tariffs or customer-tailored services. Currently, the task is done based on demographic data collected through questionnaire, which is error-prone. In this paper, we used three different clustering techniques (together with their variants) to automatically segment electricity consumers based on their consumption patterns. We also proposed a good way to extract consumption patterns for each consumer. The grouping results were assessed using four common internal validity indexes. We found that the combination of Self Organizing Map (SOM) and k-means algorithms produces the most insightful and useful grouping. We also discovered that grouping quality cannot be measured effectively by automatic indicators, which goes against common suggestions in literature.