Desing and optimization of artificial neural network models for solar resource assessment

Resumen

The aim of the thesis is the design and development of artificial neural network models for solar resource assessment, deriving reliable GHI and DNI estimates over large areas. Satellite imagery and reanalysis products, covering the whole globe or extensive areas, are used as input variables. The first two models generate daily GHI estimates. The first one uses ECMWF ERA-Interim data as input variables. The second one uses Meteosat-9 images, with better spatial and temporal resolution. The other two models are artificial neural network ensemble models for estimating hourly GHI and DNI respectively, using eleven Meteosat-9 spectral channels. Both models have been validated in a large region, covering mainly Europe and part of Africa and Middle East.