Characterization of electric consumers through an automated clustering pipeline
Clustering analysis of daily load profiles represents an effective technique to classify and aggregate electric users based on their actual consumption patterns. Among other purposes, it may be exploited as a preliminary stage for load forecasting, which is applied in the same way to consumers in the same cluster. Several clustering algorithms have been proposed and developed in the literature, and the choice of the most appropriate set of clustering parameters is crucial for ensuring reliable results. In this paper, an automated service, suited for repeated clustering analysis, is presented. The pipeline is able to process a generic time series dataset and is easily adjustable to test other clustering input parameters; therefore, it may be utilized to find the best set of parameters with the specific dataset. Moreover, it facilitates repeated characterization on real-time load profiles with the aim of detecting sudden changes of consumers behaviors and variable external conditions, which influence the real power forecasting activity on a short temporal scale.