Time Series Forecasting Methods and Hybrid Modeling both Applied on Monthly Average Wind Speed for Regions of Northeastern Brazil.
Datasets usually provide raw data for analysis. This raw data often comes in spreadsheet form, but can be any collection of data, on which analysis can be performed.
Abstract The objective of this work is to perform time series forecasts of wind speed in terms of monthly averages in the Brazilian Northeast. The following time-series models, Auto-Regressive Integrated Moving Average (ARIMA) and Holt-Winters (HW) were tested, as well as computational artificial intelligence with the use of Artificial Neural Networks (ANN). In addition, two hybrid models were tested, the first with a combination of the ARIMA and ANN models, which work in the literature, and the second is an attempt to combine HW and ANN models. The adjusted series obtained by the hybrid models, are efficient to follow the profile of the observed series of the study regions, with similarities to the data observed in terms of maxima and minima, thus indicating the capacity of the models to represent seasonalities. The calculation of error statistics involving the hybrid (HW and ANN) model obtained the lowest values in Fortaleza, São Luís and Natal, for example, with percent error values of 3.80%, 4.91% and 2.85%, respectively. The reduction of the statistical variables of errors by the hybrid models when comparing the use of the models (ARIMA, HW and ANN) separately may influence the predictions of expected wind velocities.