Forecast of Short-term Solar Irradiation in Brazil Using Numerical Models and Statistical Post-processing
Abstract
This work aims at establishing a methodology to get reliable solar irradiation forecasts for the Brazilian Northeastern region by using WRF model together with statistical methods for post-processing data. The key issue is concerning how to deal with the diversity of typical climate features occurring in the region presenting the largest solar energy resource in Brazil. The solar irradiance forecasts for 24h in advance were obtained using the WRF model. In order to reduce uncertainties, cluster analysis technique was employed to find out areas presenting similar climate features. Comparison analysis between WRF model outputs and observational data were performed to evaluate the model skill in forecasting the surface solar irradiation. After all, post-processing of WRF outputs were performed using artificial neural networks and multiple regression methods in order to refine short-term solar irradiation forecasts.