Geração de curva de colina para turbina hidrelétrica usando rede neural artificial

Abstract

Within a hydroelectric plant, the hill curve is an important graph to perform value predictions in specific circumstances. Thus, three regression models are proposed being Feedforward, Linear Regression and XGBoost. The data used to train the nets were obtained through of a reduced-scale model, totalize 1243 data. So noticing that among the three models, the Feedforward model demonstrated greater generalizability. thus a better alternative among the three proposed networks.