Aplicação do método de Kriging para estimar campos de radiação solar: um estudo para o nordeste do Brasil Documento uri icon

  •  
  • Visão geral
  •  
  • Pesquisas
  •  
  • Identidade
  •  
  • Ver todos
  •  

tipo

  • master thesis

abstrato

  • Field estimates of solar irradiation at surface for the Northeastern territory of Brazil for the years 2008 and 2011 were made. It was used the Kriging interpolation method. An intercomparison was made using site specific values of incident solar irradiationestimated by interpolation of data collected by automatic meteorological stations (WMSs) and estimates made by the physical radiative transfer model BRAZIL-SR. The evaluation was performed using statistical indices combined with the cross validation method and the visual analysis of maps of the spatial variability of solar irradiation. The selected raw data were reformatted and organized based on a qualitative procedure and then submitted to the same data quality control adopted by the Baseline Surface Radiation Network (BSRN). In Kriging procedure, the careful adjustment of the semivariogram is critical. The measured data showed no anisotropy and trends that could otherwise have prejudiced the estimates by interpolation. The results showed that the Kriging interpolations were able to explain more than 70\% of the variability of the irradiance measured by the number of WMSs in 2010 and 60\% in 2011 Considering the whole period, the Absolute Mean Square Error (ARMSE) normalized by average of measured values had lower rates of around 5\% error, which characterizes an acceptable performance of the method showing that the deviations were minor. Furthermore, these statistical results qualified the methodology employed in the adequacy of the data used in this study. The visual analysis of maps of irradiation for the site specific points in four years showed that there was little spatial variability of the surface irradiance levels with average annual irradiance around 6000 Wh/$m^{2}$. However, all maps could presented sharp gradient in the Northern region of State of Bahia. This can be attributed to anomalies in the data measured by the WMSs that were not detected by data quality control. The BRAZIL-SR model performed better when compared to the Kriging method in estimating the site specific data. The coefficient of determination $r^{2}$ showed a value close to 0.74, showing that 74\% of the variability of the measured data was explained by the model estimates. The residues indicated that the model overestimated the values in 11 of the 14 reference points. When compared to the Kriging method, estimates showed a high degree of correlation and coefficient of determination. Regarding the ARMSE, the calculated error was below 5\%. The maps obtained by using the Kriging method showed greater spatial variability in amplitude.
  • ... mais

data de publicação

  • 2014-01-01