Colloquia

Summary

In the design and development process of an offshore wind turbine, experimental and numerical tests are crucial for assessing its performance. Due to the inherent simplifications in experimental and numerical models, the prediction of wind turbine performance with a model-scale wind turbine can never include every aspect of a real-life scenario. The uncertainties in the results of measurements and numerical simulations, and the impact that parameters have on the predictions of the performance of the wind turbine, are established by a so-called uncertainty quantification analysis.

In this thesis, an uncertainty quantification analysis will be carried out for both a full-scale and a model-scale scenario. The full-scale setup of simulations consisted of a variety of turbulent inflow fields generated with the stochastic turbulent wind field generator software TurbSim (developed by the National Renewable Energy Laboratory). The results showed that high turbulent inflow fields gave higher power output, except near the rated wind speed, where higher velocities in the distribution are cut off by the control mechanism of the wind turbine. 

At model scale, MARIN utilizes a scale model of the IEA 15MW turbine in their offshore basin. The geometry, properties, and control configuration were translated to numerical simulation tool QBlade. As MARIN is investigating the stability of offshore floating structures, the rotor thrust force is the most important performance metric. The model scale wind turbine intends to match thrust force data from the offshore basin. Consequently, a variety of custom model scale turbulent inflow fields (inspired by the wind field generation in the offshore basin) were used on this turbine.  On these results, different surrogate models were built and cross-validated. It was concluded that a Gaussian Process Regression model in combination with a rational-quadratic kernel performed with the smallest error.