Uncertainty quantification with dependent input data : including applications to offshore wind farms

Offshore wind energy is an alternative power source used in the Netherlands. Due to an increasing number of turbines and an increased amount of energy produced per turbine, the break-even point for profitable offshore wind farms without subsidy is near. A major issue for utilizing offshore wind farms are the uncertainties involved in the construction and operation. These appear from different sources and are complicated by the long timescales involved. It is therefore of the utmost importance to quantify them properly.

To achieve this, different aspects of uncertainty quantification are studied and combined into a framework in this thesis. The focus herein is on the case of dependent input variables which are available in the form of data rather than probability distributions. Also, a computational model to map input data to output data is assumed to be available. However, due to numerical or physical complexity, the number of available runs of this model is limited. For efficient mapping of input uncertainty to output uncertainty, the main challenge is to select suitable samples from the input data for which the model output is obtained. Additional challenges are (i) quantifying the dependencies in the data and (ii) quantifying the sensitivities in the data.

These challenges are focused on in this thesis and resulted in efficient algorithms which can easily be applied in practice. Furthermore, two applications of the framework in the domain of offshore wind energy are studied, which show the proposed framework works well and can be applied in practice.