Previsão de demanda a médio prazo aplicada em dados reais do sistema de distribuição: uma comparação entre RNA e Lógica Fuzzy

Abstract

Demand forecasting is an important tool to support decision-making in the planning of power systems, providing essential information to aid specialists in the electricity sector in the allocation of available resources. Methods based on computational intelligence have been used in demand forecasting for more than twenty years. The Artificial Neural Networks (ANN) and Fuzzy Logic are among the most widely used techniques. In this study, we developed two demand forecasting systems for a real substation by means of an RNA and a fuzzy inference system. The case at hand studied the Cajazeiras substation, located in Paraíba/Brazil and its active power values, between the years 2008 and 2013, obtained by measurements of a system data acquisition (SCADA), which formed the time series data.