Statistical evaluation study for different wind speed distribution functions using goodness of fit tests

Modeling wind generation for use in many power system applications requires a massive database of historical wind speeds so that the stochastic nature of the wind at a particular site can be accurately analyzed. The alternative is to use reliable estimates of a probability distribution function (PDF) that can preserve the variable characteristics of wind speed and generate the desired synthetic data. This paper presents a statistical evaluation study for different collections of PDFs in order to find the best model to precisely reflect the variable characteristics of the wind at a particular site. The most commonly used PDFs, along with some advanced PDFs, have been verified against the observed wind data based on consideration of two well-known goodness of fit statistical tests. A further case study is conducted in order to evaluate the impact of sample size on the selection of the best-fit PDFs. From a variety of candidate PDFs, the results indicate that the Generalized Logistic and Dagum distributions are the PDFs that best maintain the main characteristics of the observed wind data.