Title: Estimation of total electricity consumption curves of small areas by sampling in a finite population.
Authors: Anne de Moliner - Universite de Bourgogne EDF (France) [presenting]
Herve Cardot - Universite de Bourgogne (France)
Camelia Goga - Universite de Bourgogne (France)
Abstract: Many studies carried out in the French electricity company EDF are based on the analysis of the total electricity consumption curves of groups of customers. These aggregated electricity consumption curves are estimated by using samples of thousands of curves measured at a small time step and collected according to a sampling design. Small area estimation is very usual in survey sampling. It is often addressed by using implicit or explicit domain models between the interest variable and the auxiliary variables. The goal is to estimate totals of electricity consumption curves over domains or areas. Three approaches are compared: the first one consists in modeling the functional principal scores with linear mixed models. The second method consists in using functional linear regression models and the third method, which is non-parametric, is based on regression trees for functional data. These methods are evaluated on a dataset of French consumption curves.