A hybrid model based on time series models and neural network for forecasting wind speed in the Brazilian northeast region

This paper aims to define a methodology capable of providing accurate wind speed monthly average predictions in the Brazilian Northeast region. Hybrid models involve a combination of time series models (with the exogenous variables of pressure, temperature and precipitation as inputs) with artificial intelligence. Wind power generation is growing in many parts of the world, and this growth is a result of the large number of research focused on the economical and environmental benefits. One particular line of research that may have contributed to this overall growth is the prediction of local wind speed, that is, aiming to understand and thus predict the wind regime of a given region. The hybrid model proposed in this paper was efficient in reducing statistical errors, especially when compared to traditional models, it produced the lowest percentage error between the observed and the adjusted series, of only about 8%. Finally, it is important to highlight that through this work, decision makers will have a guarantee to explore the local wind potential, allowing for the possibility of predicting future wind speed, and thus giving them the ability to plan the demand for electricity generated from wind power.