Estudo de distribuições de probabilidade para descrição de Dados de frequência de velocidade de vento e de potência Eólica

Abstract

In this study, a statistical analysis was developed for the determination, between three probability distribution functions (Weibull, Gamma and Nakagami), of the best alternative to wind velocity and wind power data representation. The treatment and the parameterization of the data, the charts elaboration and the statistical performance tests were all developed in R programming language in RStudio software. The data was measured in 2009 and 2010 by an anemometric station of metereologic data collection based in the city of Petrolina. The data gathered by Rede SONDA of environmental data passed through a treatment to remove measurement errors resulting in a 92.7% effectiveness post-treatment. For the probability distribution functions parameterization, two deterministic mehtods were used: the empirical method and the maximum-likelihood method. The calculated curves were compared to the empirical measured data histograms for both studied properties and, through three statistical performance tests, it was determined that, for the chosen time, the Weibull distribution was the best alternative, between the three, for wind speed and power density data representation since it showed the best overall performance in representation for both chosen deterministic methods in all applied performance tests, with a R2 of 0.987 versus 0.925 and 0.972 for Gama and Nakagami distributions, respectively. Besides that, it was noted a slightly higher performance from the empirical mode to the maximum-likelihood method for the three studied distribution curves, with a difference of until 2.6%. in R2 test