Exploration of Techniques for Long-Term Wind Speed Prediction

Wind generation in Brazil and in the world has been growing and, with it, studies on its source as well. Thus, in order to support the planning of the operation, long-term wind speed forecasts are made using two classes of models: SARIMA and Neural Networks, in order to test the robustness of the models. Two sites were analyzed, the Praia Formosa Wind Farm in Ceará (CE) and the Elebrás Cidreira 1 Wind Power Plant in Rio Grande do Sul (RS). For both parks, the best models were statistical models, reaching RMSE of 1,142 and 1,730, respectively; while the best model of Neural Networks, a network with a hidden layer of 100 neurons, had the RMSE of 3,913 and 4,186, respectively.