Pronóstico de Demanda Eléctrica con Pocos Datos Históricos

GARCIA, GUILLERMO; LUIS RAUL TORRES; CATUOGNO, GUILLERMO; LARREGAY, GUILLERMO

San Miguel de Tucuman

Congreso; ARGENCON 2018; 2018

IEEE Argentina

This paper presents an application of artificial neural networks for short-term forecasting of electricity demand. A feed-forward architecture with 97 entries, two hidden layers of 100 and 70 neurons respectively, and 48 output data are used. The input data to the network are the temperature and energy demand of the day before the one you want to forecast, and the day of the week you want to predict.At the output, the predicted demand for the next day is obtained. The demand and temperature data are sampled every 30 minutes. The proposed method was designed using the tools of the MATLAB Neural Network Toolbox, and the results were analyzed by working with 21 data of the electrical demand of the center area of the Argentine electrical system.