Search

We present an experimental study of floating wind turbine control with inclusion of tower motion feedback. The study is conducted for a 1:60 scaled-model of the DTU 10MW reference wind turbine on a design variant of the TetraSub floater. A review of control methods for floating wind turbines is provided, covering both de-tune and tower motion feedback methods. We next detail the implementation in the experimental setup that includes real-time integration and low-pass filtering of the measured tower-top acceleration. We demonstrate how the tower feedback loop is able to stabilize an otherwise unstable land-based controller. The results further show that the stability is maintained for increasing values of the feedback gain until a certain limit. For the chosen control parameters and the lab-generated wind field, which had limited low-frequency energy, we observe that the variations in rotor speed, blade pitch and platform motion are generally larger for the tower loop controller. This finding is not generalizable due to the special inflow conditions and because controller optimization is not performed. For the tower loop controller, a substantial response at the floater roll frequency is identified. This is caused by the influence of blade pitch on the aerodynamic torque in combination with the constant generator torque. Mitigation through generator torque control is proposed.

The design methods for highly nonlinear wave loads on monopile structures has over the past years been extended with methods based on pre-computed fully nonlinear wave kinematics. Yet, the slamming events of the strong sea states can not currently be predicted with these methods. We here present a simple recipe for the application of a recent pressure impulse based slamming load model in combination with fully nonlinear wave kinematics and validate the results against lab measurements of uni- and multi-directional storm sea states. The experimental slamming loads are extracted from lab measurements equivalent to 954 full scale hours. Six methods for the extraction of the slamming force are developed and analysed in detail, with a final selection of two for the further analysis. The experimental analysis shows that the frequency of slamming is larger in uni-directional sea states relative to sea states with directional spreading, and with slightly smaller force impulses.The calculated slamming frequencies from the measurements are used in the application of the numerical slamming model. It is shown that the application is straightforward and robust and involves an intuitive selection of the model inputs from the incident wave kinematics. A generally good agreement between the model and measurement distributions of the force impulse is observed. The difference between 3D and 2D slamming impulses, though, is found to be larger in the numerical model. This is traced to the numerical particle velocities in the wave crests.The pressure impulse model is next extended by assuming a predefined generic slamming force time variation and through calibration of the peak slamming force, a generally good agreement between the model and ensemble-averaged measured slamming force time series is obtained, given the uncertainty in the slamming load extraction. It is also observed that the commonly used non-dimensional slamming force peak of 2π is unrealistically large in the irregular slamming waves because of the 3D effects of small curling factors.

In this work, we analyze the stability of a concrete TLP platform with a square hull, designed for a 15 MW wind turbine. We verify the natural frequencies in the different degrees of freedom (DOFs) and two ultimate state design load cases (DLC) via two independent models, one in HAWC2+WAMIT from DTU and one in Orcaflex+AQWA from Bluenewables (BN). The decay tests show that the platform eigenfrequencies are outside of the main wave excitation range, and that the models predict the same natural periods within ±5%, with a maximum deviation 20% for the roll period (2.4 s vs. 2.23 s). In the ultimate load states (DLC 1.6 and 6.1), the platform motions are mainly driven by the wind and wave forcing. A small resonance is observed for the pitch motion, driven by the aerodynamic loads, and for the surge, driven by the difference-frequency components of the second-order wave excitation force. The tendons tension reaches a maximum of Nmax = 30 MN for DLC 6.1 (HS = 10.9 m, Tp = 14.0 s, Vref = 42.3ms−1). In both DLC 6.1 and DLC 1.6 the tendons are far from the compression limit, with a safety margin of ≈ 8 MN for the analyzed test cases.

In this work, we compare the wave and loads statistics for two different sea states with a TP = 15.0[s] on a h = 33.0[m] depth, one with a 10-year return period (HS = 7.5[m]) and one with a 100-year (HS = 9.5[m]). For each sea state, a unidirectional and a multi-directional wave realization was measured experimentally and then reproduced numerically via a fully-nonlinear potential solver. The computed wave kinematics were used to calculate loads on a stiff cylinder with a diameter of D = 7.0[m], and compared with experiments. To perform a quantitative analysis, we extracted 30-minute maxima of the free surface elevation and in-line force, and fitted a Gumbel distribution via a Bayesian methodology. The analysis of the experiments showed that the extreme forcing on a stiff cylinder was larger in the 2D sea state than in the 3D sea state. As for the crest statistics, the 2D were higher than the 3D for the milder storm, while they were quite similar for the stronger storm, likely a consequence of the increased wave breaking, limiting the maximum achievable wave crests. The reproduction of the sea states and associated loads via a fully-nonlinear potential solver was overall able to predict the main trends. However, the 3D wave crests were overestimated for the milder sea state, probably due to a too soft breaking filter. The 2D forces for the larger sea state were on the other hand underestimated, likely due to the lack of a slamming load model. The analysis of the average wave shape leading to the extreme load events showed that in the experiments the extreme events are dominated by physics linked with the particle velocity, and hence in phase with the wave elevation signal, as drag loads, slamming loads and velocity-dependent free-surface intersection loads. On the other hand, in the simulations they are more inertia dominated, hence in phase with the kinematic acceleration signal.

The estimation of extreme loads from waves is an essential part of the design of an offshore wind turbine. Standard design codes suggest to either use simplified methods based on regular waves, or to perform fully nonlinear computations. The former might not provide an accurate representation of the extreme waves, while the latter is computationally too intensive for design iterations. We address these limitations by using the fully nonlinear solver OceanWave3D to establish the DeRisk database, a large dataset of extreme waves kinematics in a two-dimensional domain. From the database, which is open and freely available, a designer can extract fully-nonlinear wave kinematics for a wave condition and water depth of interest by identifying a suitable computation in the database and, if needed, by Froude-scaling the kinematics. The nonlinear solver is validated against the DeRisk model experiments at two different water depths, $33.0 [m]$ and $20.0 [m]$, and an excellent agreement is found for the analyzed cases. The experiments are used to calibrate OceanWave3D's numerical breaking filter constant, and the best agreement is found for $\beta=0.5$. We compare the experimental static force with predictions by the DeRisk database and the Rainey force model, and with state-of-the-art industrial practices. For milder storms, we find a good agreement in the predicted extreme force between the present methodology and the standard methodologies. At the deep location and for stronger storms, the largest loads are given by slamming loads due to breaking waves. In this condition, the database methodology is less accurate than the embedded stream function method and more accurate than the WiFi JIP methodology, providing generally nonconservative estimates. For strong storms at the shallower location, where wave breaking is less dominating, the database methodology is the most accurate overall., Comment: Submitted to Marine Structures (Elsevier), 46 pages, 16 figures, 6 tables. The database associated with the publication is available at https://data.dtu.dk/articles/dataset/The_DeRisk_Database/10322033

A novel method is introduced to determine the long-term extreme value distribution of variables with an associated short-term distribution. The method is applicable to e.g. the extreme value with a return period of 10.000 years for wave height, crest height of waves as well as wave-induced response and loads on offshore wind turbine monopiles. The extreme values are determined by a combination of a joint probabilistic description of the met-ocean environment and of a recently developed database of fully nonlinear wave kinematics computations. For two benchmark monopile structures, representative for the conditions at Dogger Bank and the German Bight in the North Sea, the one-hour max value of crest height, inline force and overturning moment are computed via the Rainey slender-body force model. Convolution with the joint probability of the sea state parameters in the nondimensional space of wave steepness and Ursell parameter leads to the marginal probability distribution for these parameters up to a 10.000 year return period and properly accounts for the uncertainty of the extreme value parameters in also the short-term distributions. It is found that the best extreme value fit is generally obtained with a third-order polynomial fit within the space of steepness and Ursell parameter. Compared to the Forristall crest height distribution, the inclusion of full nonlinearity leads to larger crest heights at low return period levels. However, due to its realistic treatment of the breaking limitation, lower crests are predicted at the 10.000 year level for the two positions considered in the North Sea. With further verification of the wave kinematics in combination with load models and a thorough comparison to present engineering practice, the proposed methodology provides a robust future solution for directly estimating extreme value distributions of loads and response of offshore wind turbine monopiles.

The present paper aims to reproduce an extreme event extracted from an experimental time series relative to 2D long crested irregular waves. The experimental values are relative to the wave tank experiments at 1:80 scale, carried out on a stiff monopile within the Waveloads project. The monopile, modeled as a cylinder of constant diameter, is placed on a 1:25 slope and irregular waves are generated at the inlet with a mechanical paddle. In order to reproduce the time series and the consequent loads on the structure, a reconstruction of the wave elevation using a super position of 1st order components is done, the components are then imported to the aero-servo-elastic code 3Dfloat and the wave kinematics at the inlet for the computation fluid dynamics (CFD) simulations are obtained using 1st order formulation. Successively, the kinematics are imported as boundary conditions into the CFD commercial code STAR-CCM+. The free surface is modelled with the volume of fluid formulation and the k-ω turbulence model with the Menter formulation is used. In the experiments, a monopile of 0.75cm diameter (6m in full scale) is placed a distance of 7.75m (600m in full scale) from the inlet. The water depth at the model is 0.26cm (20m in full scale). Two different wave events of 25s extracted from two different time series are reproduced and compared with the experiments. The simulated wave elevation is measured at different locations and compared with the experiments. Good agreement between the simulations and the experiments is found.

Monopiles are nowadays the preferred substructure type for bottom-fixed offshore wind turbines at shallow to intermediate water depths. At these locations, the large waves that contribute to extreme loads are strongly nonlinear. Therefore they are not easily reproduced via the simple engineering models who are commonly used in the offshore industry. In the current approach, we develop a design pattern which improves this standard methodology. To retain nonlinearity in the force computations, we have precomputed a number of wave realizations by means of a potential fully-nonlinear code (OceanWave3D), for a wide span of nondimensional water depths and significant wave heights. The designer can then extract a wave kinematics time series from the precomputed set, scale it by the Froude law, and couple it with a suitable force model to compute loads. To complete the picture, slamming loads are calculated by means of the so-called pressure impulse model, recently developed at DTU. Rather than computing the time series of the slamming load, the model uses a few parameters, all except one determinable from the incident wave to calculate the pressure impulse. First comparisons with experimental results, obtained in the framework of the DeRisk project, are promising. The force and the wave elevation statistics from the precomputed simulations are in good agreement with the experiments. Some discrepancies are present, due to an imperfect scaling and to the differences in the physical and numerical domains. The computed loads from the slamming model match the experimental ones quite closely, when the wave celerity is extracted as the ratio between the time gradient and the x-wise space gradient of the surface elevation.

Spurious harmonics generation can occur in wave basins when strongly nonlinear waves are generated with linear wave maker theory. For regular long crested waves, the phenomena is known to cause a beating pattern for the amplitude of the higher harmonics, while for irregular waves the effect is often thought to average out. We here report on recent experimental results for spurious harmonic generation for highly nonlinear irregular waves. Long crested waves at two depths are considered. We find a systematic distortion of the spectral shape which is non-uniform in space. The effect occurs in the linear and super harmonic region of the spectrum and is strongest at the smaller depth. We analyze the phenomenon through re-computations in a fully nonlinear wave model and the further application of a four-phase harmonic separation. We next extend the analysis to allow separation of the third- and first-order content of the first-harmonic wave field, by computation at two different amplitudes. The numerical results resembles the measured spectrum closely and the harmonic separation shows that the spectral distortion in the super-harmonic region is to large extent a property of the second-harmonic wave field. Further, the match between the first-harmonic spectrum and the original target spectrum is shown to improve strongly by the removal of the extracted third-order contribution. The observed spectral distortion in the linear frequency region can thus be explained by this third-order field. Returning to the second-order wave field, we demonstrate that the extracted second-order wave field with associated spectral distortion can be reproduced by application of the Sharma and Dean theory with inclusion of the reconstructed spurious wave field from the linear wave generation. Finally, the effect of the spurious second-order harmonic generation on the measured forces is discussed and quantified. We find that the force peak exceedance probability is less influenced by the spurious waves than the spectral shape. We link this to the smaller phase speed of the spurious waves, that makes them arrive later than the main wave packet and thus leads to a limited enhancement of the local peaks.

The aero-servo-hydro-elastic Finite-Element-Method code 3DFloat is tailored for nonlinear, coupled time-domain simulations of offshore structures in general and offshore wind turbines in particular. This article describes the theory behind the structural model, aerodynamic and hydrodynamic load modules, control system and coupling with an optimizer. The verification and validation history includes the IEA OC3/OC4/OC5 projects, two wave tank tests and participation in commercial projects. Current development examples include implementation of advanced hydrodynamics in the DIMSELO project, implementation of soil/structure interaction super-elements in the REDWIN project, and optimization of large rotors with sweep in an industry project.

The present paper aims to show the potential of Computational Fluid Dynamics (CFD) solvers for surfboard design and its applicability by comparing three different surfboards with minimal changes in design. In fact, surfboard manufacturing routines are moving towards more controlled and reproducible manufacturing processes, in particular Computer numerically controlled (CNC) shaping techniques. As a consequence, three dimensional (3D) computer models of the boards start to be available, and can be imported in Computational Fluid Dynamics (CFD) programs. This opens up a new design methodology, where the performances of the different shapes can be studied and quantitatively evaluated, highlighting details that would be otherwise impossible to identify from a field test. The commercial CFD code STAR-CCM+ is used in the present work to compare the performance of three different surfboards, with different curvature at the bottom and different tail shapes. In the simulations, an Unsteady Reynolds Navier Stokes (URANS) approach is used, with the volume of fluid (VOF) method as free surface discretization method and the k-omega-SST turbulence model as numerical closure of the RANS equations. CFD proved to be a valid tool to compare the performances of the different shapes, bringing into light subtle but important differences between the designs. In particular, the static simulations showed that the rocker affects the performances by increasing the lift but also the drag of the board, also generating higher forces in maneuvering conditions. On the other hand, the tail shape did not affect the performances of the board in the analyzed cases.

Detailed wake measurements with two similar model wind turbines, with a diameter of 0.9m, were performed in the wind tunnel. A single turbine arrangement and a tandem setup, where a wind turbine was operated in the wake of an upstream wind turbine, were tested. Measurements of the cross section in the near and far wake as well as measurements in axial direction are presented. A single hot-wire and a five-hole probe were used. The application of the five-hole probe enabled to measure a three-dimensional flow field providing detailed inspections of the rotation of the wake. A strong tower wake and a non-uniform velocity field and turbulence distribution were observed.

DIMSELO is a Competence Building for Industry project (KPN), granted by the Norwegian Research Council under the ENERGIX program, which ran from 2013 to 2017. The project's objective was to quantify the consequences of using advanced sea loads modeling in integrated simulations of offshore wind turbines in shallow to intermediate waters. During the project, engineering hydrodynamics load and wave kinematics models of increasing complexity and fidelity were chosen from the literature and implemented. The effect of different model combinations on the substructure loads was was tested for three reference turbines: a DTU 10 MW rotor positioned on a monopile at a water depth of h = 25 m, on a second monopile at h = 35 m and on a jacket, also at h = 35 m. In this paper, the fatigue loads in a production case for the h = 25 m monopile was calculated via three different load models: the well-known Morison model, the Rainey nonlinear force model and the McCamy-Fuchs linear diffraction model. The models were coupled to kinematics coming both from linear irregular waves and second order irregular waves. The comparison showed that using the McCamy-Fuchs diffraction theory reduced the predicted fatigue damage by 15% with respect to a base case where the Morison load model was used. Nonlinear wave kinematics and nonlinear force models influenced the force calculations but did not alter the total fatigue damage, since the load cases with high wave steepness were less likely to happen. Content from this work may be used under the terms of theCreative Commons Attribution 3.0 licence. Any further distribution of this work must maintain attribution to the author(s) and the title of the work, journal citation and DOI. Published under licence by IOP Publishing Ltd

For the third time a blind test comparison in Norway 2013, was conducted comparing numerical simulations for the rotor Cp and Ct and wake profiles with the experimental results. As the only large eddy simulation study among participants, results of the Technical University of Denmark (DTU) using their in-house CFD solver, EllipSys3D, proved to be more reliable among the other models for capturing the wake profiles and the turbulence intensities downstream the turbine. It was therefore remarked in the workshop to investigate other LES codes to compare their performance with EllipSys3D. The aim of this paper is to investigate on two CFD solvers, the DTU's in-house code, EllipSys3D and the open-sourse toolbox, OpenFoam, for a set of actuator line based LES computations. Two types of simulations are performed: the wake behind a signle rotor and the wake behind a cluster of three inline rotors. Results are compared in terms of velocity deficit, turbulence kinetic energy and eddy viscosity. It is seen that both codes predict similar near-wake flow structures with the exception of OpenFoam's simulations without the subgrid-scale model. The differences begin to increase with increasing the distance from the upstream rotor. From the single rotor simulations, EllipSys3D is found to predict a slower wake recovery in the case of uniform laminar flow. From the 3-rotor computations, it is seen that the difference between the codes is smaller as the disturbance created by the downstream rotors causes break down of the wake structures and more homogenuous flow structures. It is finally observed that OpenFoam computations are more sensitive to the SGS models.