Aplicação de redes neurais recorrentes na previsão de geração eólica

Abstract

Abstract:

The present work aims to evaluate models, based on recurrent neural networks, for forecasting the generation of the Praia Formosa wind power plant. The database, made available by the National Electric System Operator (ONS), presents a historical series of wind generation, from the Praia Formosa plant in Ceará, in the period between 2011 and 2020. Forecasting models, based on LSTM Neural Networks (Long Short-Term Memory) and GRU (Gated Recurrent Unit), were implemented in the Python language. Results obtained from the two models were compared. We found, for a six-month horizon, that the GRU model performed better than the LSTM model.