Forecasting Electricity Demand by Neural Networks and Definition of Inputs by Multi-Criteria Analysis

The planning of efficient policies based on forecasting electricity demand is essential to guarantee the continuity energy supply for consumers. Some techniques have used specific procedures define input variables, which can be particular each case study. However, definition independent and casual variables still an issue explored. There a lack models that could help selection correlate criteria level importance integrated with artificial networks, directly impact quality. This work presents model integrates multi-criteria approach provides relevant neural networks forecast in countries. It consider particularities application. To demonstrate applicability time series consumption from southern region Brazil was used. dependent inputs by were selected using traditional method called Wrapper. As result this application, ELECTRE I possible recognize temperature average evaporation as explanatory variables. When included predictive models, observed more consistent results together better than linear models. Radial Basis Function Networks Extreme Learning Machines stood out potential perform forecasting.