The probabilistic analysis of extreme wind speed grounds the wind hazard assessment and the design of safe structures under wind loads. In wind engineering, this usually translates into predicting the design wind speed having a specified return period, a task that traditionally relies on Extreme Value Analysis (EVA) under stationarity assumptions. However, this paradigm has been challenged within the last decades by land surface anthropogenic modifications and climate change. To overcome the above modelling assumption, in this MSc Thesis a specific framework based on Bayesian inference, specifically a Markov Chain Monte Carlo algorithm for Non-stationary Extreme Value Analysis (NEVA), is adopted for the analysis of extreme wind speeds.