Progress in forecasting by neural networks

Publisher Website
Google Scholar
Abstract
The forecasting of electricity consumption by means of neural networks is reported. The neural model used is a multilayer perceptron. Learning is accomplished with a backpropagation algorithm. The neural network forecasts are made directly from the observations without any corrections. Exogeneous variables, such as temperature and nebulosity, are introduced directly as a network input. The output is always one neuron which provides forecast consumption one step ahead. The neural network results are judged to be no better than those of traditional models. Its advantages are its ability to forecast more than one step ahead and the possibility of introducing economic characteristics in the minimization criteria.

This publication has 2 references indexed in Scilit: