Densidade de potência eólica a partir de ensembles de modelos climáticos regionais no Nordeste do Brasil para o passado recente e futuro

Abstract

The significant increase in electricity consumption and consequent rise in greenhouse gas emissions has led to a heightened exploration of clean energy sources worldwide. In Brazil, particularly in the Northeast region (NEB), there has been notable development in wind and solar energy. Therefore, the main goal of this thesis is to perform high-resolution ensembles from the CORDEX project using simple and robust statistical techniques as well as machine learning to project wind speed and wind power density across different areas of the NEB under two climate scenarios for the near future (2041-2060) and distant future (2081-2099), comparing them to the recent past (1986-2005). The thesis is organized in the form of articles. In the first article, the relationship between wind speed and other climatic variables was evaluated. Principal Component Analysis (PCA) was used to correlate wind speed with air temperature, relative humidity, and precipitation using the Regional Climate Model (RCM) RegCM4.7_HadGEM2-ES for Rio Grande do Norte, in the recent past from 1986 to 2005. The results utilized only PC1 as it accounted for 75.74% of the data variability and had an eigenvalue above 1. Wind speed showed a strong negative correlation with precipitation and relative humidity, -0.91 and -0.94 respectively. In the second article, wind power density was calculated for areas in the NEB during the recent past (1986-2005). Initially, the ECMWFERA5 reanalysis was spatially validated with observed data. Subsequently, the RCMs RegCM4.7, RCA4, and Remo2009 were validated against the reanalysis, and overestimation values above 1m/s were observed in several NEB areas during the dry season (July to December). Four areas in the NEB were selected for study based on the number of wind farms and different climatic aspects: northern Ceará (N-CE), northern Rio Grande do Norte (N-RN), the Borborema Plateau (Borborema), and central Bahia (C-BA). The RCMs with the best statistical indices in each area were chosen to perform the wind power density calculation. NRN and N-CE were the areas that showed the highest wind power density in the recent past. Finally, in the third article, arithmetic mean and other robust techniques were applied to perform ensembles in the same areas chosen in the previous article and the same period (recent past). The arithmetic mean and convex combination techniques showed high underestimation of wind speed for the first semester, except in the C-BA area, and high overestimation for the second semester. The other techniques accurately represented (median values close to) the ECMWF-ERA5 reanalysis in all studied NEB areas. Based on the Taylor Diagram and Willmott’s index of agreement and the standard deviation ratio, the technique that best represented the reanalysis in all NEB study areas was PCR. Using this technique, wind power density was calculated for both the recent past and the near and distant futures under RCP 2.6 and 8.5 scenarios. It was considered that there is stability in wind power density values for the near and distant futures. Therefore, the N-RN area showed the highest wind power density in the NEB, followed by N-CE, Borborema, and C-BA, both for the recent past and for the near and distant futures.