Statistical analysis of extreme wind speeds

Wind is a random variable that is hard to predict and can prove out to be devastating for constructions leading to great financial loses. In the Codes of Practice of many countries the design of wind loads is obligatory in regions with high winds and is especially necessary for high‐rise constructions (skyscrapers). It is also necessary to create wind prediction models that could facilitate the introduction of citizen protection plans in cases of extreme phenomena, the utilisation of wind power for electricity and keep track of climate change. Consequently, the probabilistic study of the wind speed, in order to define a certain return period constitutes a necessity nowadays. The extreme wind speed has been widely studied especially in the structural engineering area through the classical Extreme Value Theory. Many different probability distributions have been used for the modelling of extreme wind speed. In this thesis, three theoretical probability distributions, with different tail magnitude are directly adjusted to four different tail empirical samples of wind speed data. For this purpose a norm of square error is minimized, and then a comparison of the whole body is conducted. The analysis is global and is concerned both seasonal and in the total of the months. In this way the distribution that best describes the extreme wind speed is tried to be exported. The three theoretical distributions that are fitted are Pareto II with a power type tail, Log‐Normal with middle heaviness of tail and Weibull with exponential type tail.