Probabilistic Models for Wind Actions

Abstract

In common cases the wind actions may be expressed as a product of time invariant and time variant components. The time invariant components, originated primarily from the pressure factor, gust factor and roughness, are represented by the product a single factor. The time variant component is caused by the basic wind pressure dependent on square of wind speed. Four types of distributions are considered here to describe involved random variables: three parameter Weibull distribution WB3, general extreme value distribution GEV, three parameter lognormal distribution LN3 and the Gumbel distribution GU, which is a special case of GEV. Available data indicate that the statistical parameters of annual wind speed depend significantly on local conditions. The average coefficient of variation is about 0.12, the skewness around 0.3. The time invariant component C has the coefficient of variation about 0.2 and skewness around 0.35. The annual wind pressure w, has the coefficient of variation about 0.3 and skewness 0.8. Then the 50 years extreme wind pressure w has the coefficient of variation 0.23 and skewness 0.56. Considering the above mentioned data it is shown that both the annual wind speed and wind pressure can be best described by LN3 that provide similar predictions as lower bounded WB3 distribution. Further research should be primarily focused on probabilistic models for the annual wind speed and the time invariant components.