Spanish Renewable Energy Generation Short-Term Forecast
Abstract
This work presents and compare four short-term forecasting methods for one day-ahead hourly values of Spanish solar and wind energy generation. From the four models analyzed, two are based on ARIMAX statistical methods and the other two are based on computational intelligence methods as Non-linear Autoregressive eXogenous Neural Networks (NARX). Both forecasting methodologies use the same numerical weather prediction data (NWP), consisting for solar energy, on solar irradiation and for wind power, on wind speed. The NWP data is combined in the model with the installed power of the different generation technologies in Spain. In addition to the NWP data, the models are fed with the aggregated solar and wind energy generation in hourly steps provided by the Spanish Transport System Operator (TSO). The obtained results by the forecasting methods and by the different energy generation technologies are compared using different error metrics such as: MBE, RMSE, MAE, MAPE and MADPE.
Keywords
Forecasting, Solar irradiation, Solar forecasting, Wind forecasting, Energy market, Arimax, Neural networks, Narx, Time series