Visualizing 2665 electric power load curves an a single A4 sheet of paper

The load curves of a panel of 205 clients of EDF are systematically collected and archived in a centralized database. Among other applications, these curves are used to understand the power consumption behaviour of these clients, in order to fix the price more acutely. A way of analyzing the consumption of these clients is to perform automatic clustering of the different load curves and give an interpretation to the clusters. In this paper, we investigate the use of new techniques to perform this clustering process: the Kohonen maps. Combining this method with new visualization techniques provides a much better understanding of large datasets of load curves.