Efficient probabilistic analysis of offshore wind turbines based on time-domain simulations

Abstract

Offshore wind energy plays an important role in the successful implementation of the energy transition. However, without subsidies, it is not yet sufficiently competitive compared to other renewables or conventional fossil fuels. This is why offshore wind turbines have to be structurally optimised with regard to economic efficiency. One possibility to significantly increase economic efficiency is to improve the reliability or at least to assess present reliability levels precisely. For an accurate reliability assessment during the design phase, probabilistic analyses based on time-domain simulations have to be conducted. In this thesis, a methodology for a comprehensive probabilistic design of offshore wind turbines with special focus on their substructures is developed and applied. All investigations are based on time-domain simulations. This leads to more accurate results compared to semi-analytical approaches that are commonly used for probabilistic modelling at the expense of higher computing times. In contrast to previous probabilistic analyses, considering only particular aspects of the probabilistic design, this work defines a comprehensive analysis that can be split up into the following seven aspects: deterministic load model, resistance model (failure modes), uncertainty of inputs, design of experiments, sensitivity analysis, long-term extrapolation/lifetime distribution, and economic effects. For five of these aspects, scientific innovations are realised, while state-of-the-art approaches are applied for the last two aspects. First, a method is developed in order to efficiently consider soil effects in offshore wind turbine simulations. This method is integrated in a state-of-the-art model for offshore wind turbines to enhance it by considering soil characteristics. At the same time, the number of degrees of freedom is kept constant. Second, the uncertainty of the most important inputs, i.e. environmental conditions, is determined. Theoretical statistical distributions of environmental conditions - like wind speeds - are derived using real offshore measurement data. State-of-the-art approaches are improved by using more sophisticated distributions. This enables a better agreement of theoretical and empirical distributions. Moreover, a variety of environmental conditions - having been neglected so far - is taken into account. Although there is some inherent uncertainty for every input, this uncertainty influences relevant outputs (e.g. reliability) only in some cases. Therefore, third, global sensitivity analyses are applied to determine significant inputs. All other inputs are set to deterministic values to reduce computing times. Especially for non-linear systems, the developed approach is more accurate than commonly used sensitivity analyses in the field of wind energy. Since - due to computational limitations - it is not possible to simulate the entire lifetime of an offshore wind turbine, long-term extrapolations have to be applied. These extrapolations increase the uncertainty of lifetime estimations. If probabilistic inputs are used, this effect is even intensified. Hence, several improved sampling techniques are developed. They enable a significant reduction of the extrapolation-based lifetime uncertainty compared to classic approaches. Finally, the effect of variable, distributed lifetimes on the economic profitability of wind farm projects is investigated. In contrast to state-of-the-art investigations, economic aspects are not included in the engineering model, but independent economic and engineering models are combined. This enables high-quality results in both disciplines. Using the methodology for comprehensive probabilistic designs that is developed here, it is demonstrated that probabilistic analyses of wind turbines using time-domain simulations are possible. For this purpose, it is necessary to reduce computing times by applying adequate sensitivity analyses and extrapolation techniques. These approaches are developed in this thesis. Moreover, based on present findings, well-founded recommendations for efficient and realistic probabilistic simulations of offshore wind turbines are given. Finally, since the profitability of wind turbines depends significantly on the service life and lifetimes scatter substantially, deterministic approaches cannot be economically optimised. Hence, probabilistic analyses are valuable.