Overview and performance assessment of the clustering methods for electrical load pattern grouping

Abstract

In the current structure of the electricity business, distribution and supply services have been unbundled in many jurisdictions. As a consequence of unbundling, electricity supply to customers is now provided on a competitive basis. In this context, the electricity suppliers need to get accurate information on the actual behaviour of their customers for setting up dedicated commercial offers. Customer grouping on the basis of consumption pattern similarity is likely to provide effective results. This paper provides an overview of the clustering techniques used to establish suitable customer grouping, included in a general scheme for analysing electrical load pattern data. The characteristics of the various stages of the customer grouping procedure are illustrated and discussed, providing links to relevant literature references. The specific aspect of assessing the performance of the clustering algorithms for load pattern grouping is then addressed, showing how the parameters used to formulate different clustering methods impact on the clustering validity indicators. It emerges that the clustering methods able to isolate the outliers exhibit the best performance. The implications of this result on the use of the clustering methods for electrical load pattern grouping from the operator’s point of view are discussed.