Comparison of Two Different Estimation Methods of Wind Speed Extreme Values

The estimation of wind speed extreme values is a topic assuming increasing importance in wind energy studies. Two different methods are compared here for that purpose, in view of safety applications. The first model is a parametric one, based upon a classical extreme value model, such as the Gumbel or the Inverse Weibull distribution. The alternative model is a "non parametric" one, based upon a stochastic characterization of the wind speed by means of a Poisson distribution. For both methods, estimates are carried out by means of Bayes estimation approach. The two approaches are compared in terms of robustness of the estimates of a proper safety index, with respect to departures from the assumed wind speed model. A large set of simulations results are discussed, as a first step towards a deeper insight to wind speed estimation methods, taking into account model uncertainty.