Offshore floating wind turbines are one of the newest technologies in the renewable markets today. The world’s first floating wind farm, the Hywind Scotland Pilot Park, was commissioned in October 2017 and has been competitive with fixed bottom offshore wind turbines. There is a global push to make more renewable energy available, but less desire to have wind turbines cluttering the coastline. Floating wind turbines enable the developer to take advantage of unused offshore space, at depths where traditional fixed bottom structures are impractical and at locations that do not spoil the vista of the coastline. This thesis project aims to develop a working mooring system at depth of 600 m in the Norwegian North Sea, and then investigates the possibility of shared anchors in a wind park with this mooring system. The DTU 10MW reference wind turbine atop a classic spar substructure is used. First, the mooring system at 320 m is tested under decay and environmental loads. Then a chain-polyester-chain mooring line with a bridle was developed for 600 m so that the surge offset is limited to <60 m for 3 load cases. A simplified model of the wind turbine was then developed for these three load cases. The simplified model was then used to create a wind farm arrangement with 5-6 turbines each. Each wind farm varied in layout and in the number of shared anchors. It was found that while the mooring system designed passes the surge offset and natural frequency requirements, and the normal ULS safety class, it failed the high safety class in some cases. For shared anchors with multidirectional loads, the resultant force on the anchor is significantly less as long as the lines are distributed equally around the anchor point. The resultant force does not increase with two lines 120± apart. The footprint of a single turbine with the designed mooring is larger than the footprint of the entire Hywind Scotland Farm, so suggestions are made for improvement and further work.

In the simulation of floating wind turbines, a traditional rigid floater assumption becomes less valid while pursuing large size floating wind turbines with steel-efficient floaters. Up to date, hull flexibility still cannot be efficiently incorporated into aero-servo-elastic-hydro simulation tools, and the possible influence of hull flexibility has not yet been well-understood. Consequently, it is necessary to identify the significance of hull flexibility and the possible effect of it. Recent researches have been investigating the influence of hull flexibility on substructural internal load, global responses and dynamics of the system. However, little has been done from a tower design perspective. Moreover, tower design for a floating foundation has also been seldom documented. To fill the knowledge gap, two research questions are defined: What is the difference in tower design with a floating foundation? and What is the effect of hull flexibility on tower design? To answer the first research question, a FEM model with rigid hull is built based on four floating concepts designed for DTU 10MW wind turbine. The tower fore-aft bending natural frequencies are compared between fixed foundation and floating foundation. The second research question is answered by developing a FEM model with flexible hull based on a spar-buoy concept. The rigid hull model and the flexible hull model are compared by implementing structural analysis and fatigue damage estimation under waves load. The result shows that the 1st tower bending natural frequency increases significantly(except for TLP) from a fixed foundation to a floating foundation, making it difficult to achieve a soft-stiff tower design. Furthermore, it is indicated that hull flexibility can decrease the 1st tower bending natural frequency, and the magnitude varies with different tower designs. A stiff-stiff tower decreases more while a soft-stiff decreases less. Lastly, the fatigue damage estimation implies that a soft-stiff design can be lack of fatigue strength to survive from waves load. In conclusion, a soft-stiff tower design is difficult for large size floating wind turbine partly due to the increase in 1st tower bending natural frequency from fixed foundation to floating foundation, and partly because of strength requirement for fatigue load. As for a stiff-stiff tower design, without considering hull flexibility, there is a high uncertainty in the 1st tower bending natural frequency. As a result, for large size floating wind turbines, inclusion of hull flexibility is necessary for the tower design.

With the Paris climate accords signed in 2016, most countries have committed themselves to ambitious climate targets during the next decades. One of these targets is a dramatic increase in the overall energy portfolio's market share of renewable energies. This increase in renewable market share will, for a large part, consist of newly built offshore wind farms. In turn, this rise in offshore wind energy projects is expected to be especially dramatic in northern regions, where high and constant wind speeds prevail. However, as offshore wind farm projects move further north, additional challenges need to be faced. One of these is the technical challenge to design offshore wind farms for possible encounters with drifting sea ice. To tackle this challenge, a proper understanding of the mechanics associated with encounters of drifting sea ice with offshore wind turbines is essential. Such encounters are currently primarily understood phenomenologically, and the associated models simulating these encounters – or ice-structure interactions – therefore are phenomenological as well. Moreover, most ice-structure interaction models are fundamentally one-dimensional, whereas ice-structure interactions are generally not one-dimensional. This mismatch holds especially for ice-structure interactions with offshore wind turbines, where wind loads are generally misaligned with ice loads causing highly two-dimensional ice-structure interaction problems. Therefore, the first half of this work sets out to extend one of the industry-leading phenomenological one-dimensional models – the Hendrikse (2017) model – to a two-dimensional environment. The ultimately developed Zero-friction contact Area variation Model By Omnidirectional Numerical Ice (ZAMBONI) attempts to do so by introducing practical extensions rather than introducing new assumptions. Nevertheless, one extension does entail a shift from current one-dimensional ice-structure interaction models. Namely, the assumption that ice experiences neither friction at the ice-structure interface nor internal shear forces. Consequently, much of the correctness of this model hinges on this extension. To assert the correctness of ZAMBONI. A comprehensive verification campaign is performed as well as a simple order-of-magnitude validation campaign. Although both confirm the extensions' correctness, further validation is required, especially concerning the zero-friction principle. Upon developing and discussing this two-dimensional ice-structure interaction model, the second half of this work couples ZAMBONI to an offshore wind turbine model to gain further insight into ice-structure interactions. These dynamically coupled two-dimensional simulations serve two purposes. Firstly, to compare one- and two-dimensionally simulated load cases of aligned ice and wind. Secondly, to perform newly simulable load cases of misaligned ice and wind. Four primary findings are discussed. Firstly, as hypothesized, introducing a disturbing wind load lowers the ice-structure contact area, causing smaller loads and displacements due to ice loads. This effect is especially well observable for misaligned wind loads and low far-field ice velocities. Secondly, a new ice-structure interaction regime is observed where ice and structure synchronize in the structure's first bending mode. This synchronization occurs most dominantly for two-dimensional ice. Thirdly, frequency lock-in occurs solely in the second bending mode and is terminated at lower ice indentation speeds for two-dimensional than for one-dimensional ice. Finally, small ice-wind misalignments, which are most common, appear highly similar to load cases of fully aligned ice and wind.

Wind power to cold climate sites is attractive because of favorable wind conditions and low population density. However, icing of wind turbine blades remains one of the main challenges for cold climate sites. Ice formation on wind turbine blades causes several problems, such as the loss of power production, unbalanced loads in drive train and ice throw. Most of the existing detection methods need special sensors installed and would be expensive to achieve a satisfied accuracy. In this research, an alternative model-based ice on blade detection method is proposed. There are mainly two advantages of the method. Firstly there are no additional measurements are needed. Secondly the detection method can be implemented for any kind of blade aerodynamic changes, not only ice on blade. The basis of the mode-based ice-detection method is a reliable linearized wind turbine model. The accuracy of the model is mainly influenced by the number of degrees of freedom (DOFs) and the number of operation points (OP). The most efficient model is the one with less DOFs and OP but without losing much accuracy. The relative importance of DOF and OP is revealed and suggestions are given for an optimum choice of DOF and OP for different quantities. Specifically, for power production properties, increase the number of OP is more efficient. While, for blade and tower related properties, increase the number of DOF is a better choice. The ice on blade influence of aerodynamic force, power production and structure loads is studied for three ice conditions, which are start ice, light ice and moderate ice. Based on the ice on blade influence and the reliable linearized wind turbine model, the concept of model-based ice-detection is proposed as follows: Firstly ice on blade changes the aerodynamics and the blade mass. Therefore it can be considered as one wind turbine system changed to a different system. That means the outputs of the two systems, with ice and without ice, are different, even the inputs are exactly the same. From the ice-detection point of view, if the difference of outputs between iced system and cleaned one is much larger than modeling error, ice on blade can be detected. With the proposed model-based detection method, the ice-detection capability of different output quantities is studied. It has been found that the power production related quantities usually have the highest ice-detection capability, while the blade and tower properties have relatively lower capability. In order to verify the ice-detection method for a broad working conditions, implementations are performed for above and below rated wind speed conditions. Successful detections have been achieved for both of them. For possible measurement errors in reality, they are efficiently considered by introducing uncertainty factors. The ice-detection capabilities with measurement error have been analyzed for three ice conditions and detected based on different quantities. Although a drop in the detection capability has been observed if considering the measurement error, all the quantities still can detect the ice successfully to some degree.

The report summarises the development of a frequency domain Multi-Unit Floating Platform (MUFP) model and a parametric design optimization using the model. The model returns the response statistics of a MUFP due to environmental wind and wave loading using input variables of column separation, column diameter and draft. The model is developed with careful consideration to ensure compatibility with gradient-based optimizers. The low fidelity tool is capable of quickly traversing a design space to land on optimal substructure dimensions. The program uses the input variables to calculate the geometry, buoyancy, mass and stiffness matrices of the MUFP. The design of a mooring system is considered outside the scope of this report. As a result, the stiffnesses in surge, sway and yaw are filled with placeholder stiffnesses. The hydrodynamic coefficients of the three MUFP columns are calculated in HydroD. Smaller drag components such as bracing are assumed to be less influential and are therefore ignored. To maintain computational efficiency during the optimization, the hydrodynamic coefficients are interpolated using surrogate models. The rated power of the MUFP is 10 MW, this is generated using two 5 MW turbines. The aerodynamic loading is accounted for by taking the Power Spectral Density (PSD) of a thrust force time series generated for the NREL 5 MW reference turbine in SIMA. The thrust force time series is taken as the sum of two individual SIMA simulations run in a 280m x 200m wind field. The response statistics of the MUFP are calculated in the frequency domain. This is done by assuming the MUFP motion can be modelled by a Gaussian distribution. The Gaussian distribution is used to derive the zero-crossing periods and the expected number of cycles in a 3-hour timeseries. These values are used to calculate the probability of the maximum wave amplitude. The Most Probable Maxima (MPM) is then found by equating this probability to a Rayleigh distribution. The MUFP is tested against two load cases at rated wind speeds. The first load case includes uni-directional head-on wind and wave loading. The second load case is designed to test the MUFP weathervaning properties. This is done by simulating head-on wind loading, a 90 degree wave heading and a 3 degree yaw misalignment. The optimization is run using the SciPy SLSQP Minimize function. The objective function is the total steel mass and constraints are set on the static heave displacement in addition to the MPM pitch and roll rotations expected in a 3-hour timeseries. The optimization was run three times. Two solutions were found at the global minimum while the third was found at a less optimal local minimum. A comparison of the MUFP optimized dimensions is made against a 10 MW WindFloat design. The comparison revealed the optimization solution was considerably smaller than the WindFloat. Although the physical properties of the MUFP do provide benefits such as a reduction in the aerodynamic pitching moment arm, it is concluded that the reduction in size is a result of the calculation process underestimating the MUFP response statistics.