Effect of Outliers in Wind Speed Assessment

In offshore wind energy assessment, the probability of a stochastic model failure for the wind speed should always be examined, since wind data are prone to the occurrence of outliers. Their presence suggests the use of more efficient and less sensitive methods. In this work, robust regression methods are introduced, described and applied in order to model the linear relationship between wind speeds obtained from two different data sources in the Aegean Sea. Moreover, an evaluation procedure, which is very frequently used in the measure-correlate-predict family of methods, is also applied for comparing the efficiency of the robust and the most commonly adopted ordinary least squares estimators for different time frames. The obtained results clearly suggest that outliers in wind data samples should not be ignored or diminished.