PREDICTION MODELS OF ELECTRICITY DEMAND WITH TIME DATA FOR URUGUAY

The impossibility of storing electrical energy makes predicting an indispensable tool for an efficient management of electricity production. Providing models of electricity energy demand gives to the producer entity a more accurate knowledge of the market and allows reducing the uncertainty in decision making. The objective of this work is to develop a forecasting system for the short term, based on articulated models with daily schedules prediction models. To this aim, two methodological approaches of modeling are evaluated. The first one is based on the estimation of a univariate ARIMA-IA model to a single series of energy demand. This model incorporates the effects of the special days, atypical events and a SARIMA component. The second methodology is based on the estimation of 24 models, one for each hour. Finally, the individual performance of each predictive model is evaluated and contrasted with the results of a daily predictive model.