UTILIZATION OF ARTIFICIAL NEURAL NETWORKS FOR FORECASTING OF ELECTRICITY GENERATION AT UHE PEDRA DO CAVALO, BAHIA, BRAZIL

The aim of this study is to forecast the energy generation of the Pedra doCavalo Hydroelectric Power Plant (HPP) in Bahia, Brazil, using artificial neuralnetworks (ANNs). Historical data of operational variables from the plant were used toforecast production over 2-year and 4-month horizons. Test blocks in the ANN wereconducted with variations in the number of layers, neurons, and data sets. Using twoyears of hourly reservoir level data from seven different points on the ParaguaçuRiver, it was possible to estimate electric generation four months ahead, achieving a74% efficiency rate with a neural structure of three layers and ten neurons each. Theresults demonstrate the potential and efficiency of using ANNs for forecastingproduction scenarios in hydroelectric plants.