Aplicação de redes neurais artificiais à previsão do preço da energia elétrica para distintas zonas de mercados desregulamentados

Abstract

The estimation of the energy price plays a crucial role in the current model of commercialization of energy in many countries. Better estimation capacity makes it possible to identify appropriate strategies for market players. Thus, this work aims to determine a methodology to estimate point values and intervals (maximum and minimum) for a day for the Pennsylvania - New Jersey - Maryland energy market through Data Mining, where they will be considered Attribute Selectors and Artificial Neural Networks. In this sense, the responses of neural networks of the Multilayer Perceptron type and of Recurrent Neural Networks will be analyzed, considering different topologies. Keywords: Energy market, Artificial neural networks, Energy Price, Time-series forecasting.