Load forecasting with Neural Networks for Antioquia-Choco region.

This paper describes a short term load forecasting system based on artificial neural networks and a data pre-processing technique. The forecasting is performed according to the type of week is going to be predicted, based on a given classification. As result of the pre-processing, database are obtained and used to train the neural networks. These networks are evaluated later to generate weekly prediction. The multi layer perceptron (MLP) network presented better results. The neural network inputs are hour demand and day of the classified week and the output is the load forecasting.