Three-stage clustering procedure for deriving the typical load curves of the electricity consumers

Load profiling based consumer characterization is a two-stage procedure: during the first stage the daily load curves of each consumer are grouped in a certain number of clusters. The average load curve is the normalized load profile of each cluster. For each consumer, a load profile is chosen and a new clustering takes place leading to the formation of consumer classes. This paper proposes one additional stage, the “pre-clustering” step in order to optimize the whole load profiling procedure. We employ the family of the hierarchical agglomerative algorithms in order to group the daily load curves into classes. The performance of the algorithms in formulating well separated and compact clusters is checked by an adequacy measure. The proposed analysis is suitable when there is an availability of large load data samples that are gathered through the smart metering infrastructure and stored at the utilities´ databases.