ARIMA vs. Neural networks for wind speed forecasting

Abstract

In this paper an ARIMA model is used for time-series forecast involving wind speed measurements. Results are compared with the performance of a back propagation type NNT. Results show that ARIMA model is better than NNT for short time-intervals to forecast (10 minutes, 1 hour, 2 hours and 4 hours). Data was acquired from a unit located in Southern Andalusia (Peñaflor, Sevilla), with a soft orography (10 minutes between measurements). This feature is which makes performance of the ARIMA model and the NNT very similar, so a simple forecasting model could be used in order to administrate energy sources. The paper presents the process of model validation, along with a regression analysis, based in real-life data.