Comparative Analysis of Spatial Techniques for Global Horizontal Solar Irradiance Estimation in the Brazilian Context

Solar irradiance information is one of the crucial factors for feasibility studies of photovoltaic plants. In that sense, this work presents a comparative performance analysis of a set of hourly spatial global horizontal irradiance (GHI) estimation methods in the Southeast region of Brazil, considering: GHI data from satellites (Landflux, SARAH-2, and NASA), spatial interpolation techniques (Ordinary Kriging, Inverse Distance Weighting and Nearest Neighbor Extrapolation) and estimation using artificial neural networks (Multilayer Perceptron). The results showed that the approximation obtained by satellite data outperformed the other methods in terms of average precision (nRMSE), especially in capturing temporal variations. In addition, it was possible to outline the behavior of each technique as a function of the distance from terrestrial data available for estimation.