Utilización de técnicas de agrupamiento para la obtención de curvas de carga representativas de consumidores tipo residencial, comercial e industrial: utilización de la técnica de agrupamiento jerárquico para la obtención de curvas de carga representativas de consumidores de tipo residencial, comercial e industrial.

Abstract

This curricular integration work deals with obtaining representative daily load curves, using the hierarchical grouping technique. This technique is applied to the set of daily load curves product of the active power consumption measurements of 236 clients, to obtain representative daily load curves, in addition two adequacy indices are calculated: clustering dispersion or CDI (Clustering Dispersion Indicator) and mean index or MIA (Mean Index Adequacy). The analysis of the daily load curve allows the identification of consumption patterns responsible for the variations in the electrical demand, in order to carry out the electrical planning in the short, medium and long term. The hierarchical clustering technique consists of obtaining representative daily load curves, which are the center of each cluster performed by the algorithm. This algorithm assigns the daily load curves to a specific group, based on the calculation of the weighted Euclidean distance between the center load curve of the group and the analyzed daily load curve, if the distance is less than a previously established threshold, the daily load curve will be assigned to said group and its center will be recalculated, otherwise a new group will be created in which the new daily load curve is also the center of said group. Starting from the active powers consumed by 236 clients belonging to the electricity company CNEL-Azogues Business Unit to obtain daily load curves joining the average peak powers recorded for each hour, forming the database to be analyzed. Before entering the daily load curves into the algorithm, they are separated by Monday to Friday, Saturday, Sunday and by type of consumer. Each daily load curve is normalized to its maximum recorded active power. The algorithm of the hierarchical clustering technique is developed in Matlab and includes the display of representative daily load curves for three types of consumers: residential, commercial and industrial. To check the goodness of the algorithm, the values ​​of the MIA and CDI indices are analyzed.