Mapping of the Solar Irradiance in the UAE Using Advanced Artificial Neural Network Ensemble

Accurate spatial and temporal solar irradiance mapping is important for a wide range of applications related to efficient utilization of solar-based energy harvesting technologies. An improved artificial neural network (ANN) ensemble framework is proposed to estimate the solar irradiance variables from satellite data acquired using the Spinning Enhanced Visible and Infrared Imager (SEVIRI) instrument onboard the Meteosat Second Generation (MSG) satellite. The cloud-free and cloudy observations were clustered in two separate case studies, and for each case, two ANN ensemble models were trained; one for predicting the diffuse horizontal irradiance (DHI) and the other for predicting the direct normal irradiance (DNI). The global horizontal irradiance (GHI) was then computed from DHI and DNI estimates for each cloud condition. The proposed methodology was also applied in a second scheme, where the input and output variables, for each case study at each cloud condition are preprocessed using the Box-Cox transformation. The training and testing of the models were performed using spatially and temporally independent data. The proposed models produced significantly improved generalization ability and superior performance when compared with results from a previous study dealing with solar mapping in the United Arab Emirates (UAE).