Revista Colombiana de Estadística
Print version ISSN 0120-1751
Our goal is to model, with forecasting aims, the daily electricity demand in a southeast colombian region through a non-parametric regression model implementation. We consider some "calendar variables" such as time of the day, day of the week, month, and year, among others, on the estimation process. Data come from an electricity distribution local company and are taken from Valencia (2005). Available data go from January 2001 to November 2004. These data show such a complicated behavior that it becomes hard to model using classical parametric models. Since exploratory analysis suggested the existence of an electricity demand daily typical curve, we used non-parametric models instead. For comparison purposes, we made use of some other methodologies including ARIMA models and the insertion of macroeconomic variables. Statistical processing was run using R.
Keywords : Smoothing; Non-parametric regression; ARIMA models.