The demand for renewable energy is unquestionable. Global climate trends have been fomenting the urge for renewable, low carbon-footprint technologies. Wind energy is one of the most prominent alternatives to conventional fossil fuel energy conversion. While the interest on wind energy dates back to the 1980 decade, offshore wind only became highly relevant later in the 2000’s, with the realization of the potential within the offshore wind resource. The fact that wind depends on the Sun, makes it a virtually innite source of energy. In the case of offshore wind energy, it is remarkable the contribution and role of engineering practices and procedures to enhance the cost-reliability ratio for the harnessing systems. In this sector, research is expected to be the key driver of development up to 2050. While improvement of the techniques applied in the sector is on high demand in order to unlock new breakthroughs that will enable wind energy to become progressively more competitive, the development of new innovative practices is no less important.
In this thesis, design techniques for offshore wind turbines are evaluated and discussed. The work presented is strongly focused on the probabilistic assessment for structural design. A particular focus of the work is also directed to the cost of the design procedures, which recurrently hinder optimization at the design level. Different contributions are presented in the following document. Extreme waves are evaluated using different probability models, motivated by an identied lack of consistency in the approach to wave extrapolation. Three probability models, Weibull, Exponential and Generalised Pareto, are analysed in a Peak-over-threshold analysis. An innovative choice of threshold methodology is introduced. This procedure relates to the density function shape. A logarithmic transformation to signicant wave data is applied jointly with the Generalised Pareto model. Five t indicators are used to compare the results, showing no evidence to reject any of the models studied. Stress-cycle fatigue design and its probabilistic basis are also discussed. Stress-cycle fatigue has been a widely studied problem in the wind engineering sector. Different probabilistic approaches are compared to assess it. Sample size influence on the loading spectra approximation is investigated and a bootstrapping procedure is applied to quantify the loading sample uncertainty. The large design effort associated with the stress-cycle fatigue assessment motivates then the application of meta-modelling techniques for fatigue design. Gaussian process predictors, also known as Kriging models, are researched as an alternative for efficient stress-cycle fatigue assessments. Due to their inherent probabilistic character, these are also evaluated as potential quantiers of uncertainty. A new indicator for space reduction considering uncertainty is presented in a global sensitivity analysis. Gaussian process predictors are applied using different methods to dene the design of experiments. An innovative search function is introduced, that relates to the problem of stress-cycle fatigue analysis.
The developments presented are supported by an extensive literature review. Application of the design procedures is studied on a 5MW monopile baseline turbine. Results from the work developed showed that some engineering practices still demand improvements in regard to their probabilistic comprehension. Are examples; the analysis of the probability model support; design approach and its probabilistic relation to the problem; the size of representative design samples; or demand for notions of improvement that relate to the physical problem when meta-modelling. It is of relevance to highlight that development on uncertainty assessment for the wind energy sector has been steadily occurring through small contributions from multiple authors that spend signicant efforts in order to comprehend what are the sources of error for each of the design steps. Aligned with this, the proposed work adds new insights to the fields of offshore wind turbine analysis and uncertainty characterization. -> Read more about the research by Rui in TRUSS