#Introduction

Offshore wind energy is an important part of the current energy system. As the decarbonisation of the energy system progresses, floating wind becomes more important, allowing to exploit resources located in deep water.

The bottleneck of this technology is currently the cost. The platform and the mooring system provide sufficient stability to resist the overturning moment. For floating wind turbines, they make up a proportionally larger portion of the initial investment when compared with bottom fixed structures.

Traditionally, catenaries are used for the mooring systems. They provide restoring forces via their weight. This mechanism becomes inefficient for larger platforms, leading to very large catenaries that are difficult to produce and install. Leaner mooring systems can be installed if taut or semi-taut concepts are used, where the lines are under tension and the restoring force is provided by the tension force itself.

To reduce the size of the mooring system components and the associated fatigue loads, peak load reduction devices have been used. They normally have a non-linear stiffness, and in case of extreme event can reduce the maximum load at the expense of a larger displacement of the structure.

In the framework of the EU Project FloatFarm, we investigate numerically the benefits of using peak load reduction devices on two different floating platforms with an IEA-15MW wind turbine. For the WindCrete concrete spar-buoy, we use DTU's in-house aeroelastic tool HAWC2. For the semi-submersible platform VolturnUS from UMaine, we use OrcaFlex.

#Methodology

We use the environmental conditions from Deliverable 1.3 of the CoreWind project [1], with a water depth of 200 m and a 10-min omnidirectional average wind speed of 12.26 m/s.

For both platforms, we start with a baseline catenary mooring design, with 6 mooring lines (i.e. three couples at 120 ° radial separation). The WindCrete spar also has a yaw bridle that increase the stiffness in the yaw degree of freedom. Both platforms are tested on three load cases, a fatigue one (DLC1.2), an extreme operational condition (DLC1.6) and an extreme non-operational condition (DLC6.1). We then re-design the mooring system with synthetic mooring lines, in order to achieve similar natural frequencies of the system, and finally rerun the simulations on the full set of environmental conditions.

In the third phase, we introduce the peak load reduction device, either in the catenary or in the synthetic line system, and evaluate the same environmental conditions for the two platforms and the associated fatigue loads.

#Expected results

We conclude on the strengths and drawbacks of the different mooring systems related to the platform types. In particular, we expect a drop of the mass of the mooring system when going from catenary to synthetic lines, with a possible increase of the overall fatigue due to peak loads. With the peak load reduction device, we expect a reduction of the fatigue on the mooring line system.

[1] V. Arramounet and F. Borisade, Public design and FAST models of the two 15MW floater-turbine concepts. Deliverable 1.3 of the CoReWind EU project, 2020

Single-point mooring (SPM) concepts might be adopted for floating offshore wind turbines (FOWT) to take advantage of its passive weathervaning ability. The main objective of this mooring concept is to allow for a reduction of mooring loads, but, in the case of FOWTs, other benefits may arise, such as lower stability demands for the floater and the possibility to dispense the turbine’s yaw adjustment system. In the early stages of the design of a SPM-FOWT, a static analysis regarding the expected equilibrium heading angles under environmental conditions representative of the installation site must be performed. As a result of the different directions and intensities of wind, sea/swell waves and current, the equilibrium headings often involve a misalignment of the wind with respect to the rotor. Previous studies done in the University of São Paulo with a deep-water FOWT designed to operate offshore Brazil has indicated a passive wind-rotor misalignment generally clustered within 0-20°. Cases above 20° were infrequent, but they did exist in certain weather conditions with the turbine operative. In this previous analysis, rotor loads were precomputed with OpenFAST, using the Blade Momentum Algorithm (BEM) for different speed and wind direction w.r.t. the rotor axis. Nonetheless, BEM has been validated primarily for yaw angles within ±50° and, for larger inflow angles, recent AeroDyn updates added Glauert’s skew momentum correction. Yet, higher fidelity approaches such as computational fluid dynamics (CFD) and vortex methods (VM) remain preferred. A second limitation is that BEM does not capture the net yaw moment with sufficient fidelity. A third issue arises when the net moment is calculated in the so-called Region III, where the pitch controller alters the blade aerodynamics, and the aligned equilibrium at 0° becomes unstable and bifurcates into stable headings at ±θ with increasing wind speed. The present work primarily aims to address BEM limitations in predicting yaw moments at large inflow angles. The cOnvective LAgrangian Filaments (OLAF) based on VM implemented in OpenFAST is adopted to compute thrust and moments over yaw angles 0-±90° and wind speeds 3-25 m/s" . An effort is made to validate OLAF by reviewing the literature and replicating a published case study using multiple independent software packages. This yields robust rotor–moment surfaces across the SPM-relevant operating envelope. Although OLAF is slower than BEM, it remains far less expensive than CFD while providing predictions that closely align with CFD for these conditions. The resulting rotor–moment surfaces can be integrated into an equilibrium heading model and embedded within a hull–mooring optimization loop to rapidly predict steady heading and rotor moments. In this way, the optimization framework can generate an optimized hull–mooring system for comparison with an optimized spread‑mooring configuration.

The projected expansion of floating offshore wind into large-scale farms will demand a volume of anchors that significantly exceeds the current capacity of the supply chain. Anchor optimisation is critically needed to reduce the scale of individual anchors, and thereby the overall demand, improving system efficiencies across manufacturing, transport, and installation. Current anchor design principles have evolved from the oil and gas sector and borrow conservative safety factors from this sector, driving a very cautious approach that is not efficient for anchoring systems for farm scale floating offshore wind. Different limit states and risk profiles could be tolerated if the risk of anchor failure were accurately determined and quantified.

More efficient anchoring systems could be adopted if the emerging concept of Whole-Life Geotechnics were integrated into the anchor design approach. Whole-Life Geotechnics captures the long-term evolution of seabed properties and anchor resistance to pull-out by considering variable loading from environmental conditions (winds and waves) acting on the floating offshore wind turbine. These loads, that are transmitted to the anchoring system, cause changes in the seabed strength around the anchor, and can lead to potential increases in the long-term anchor capacity as a result of seabed consolidation. Increased through-life anchor capacity could lead to a 30% reduction in the required anchor size compared to in traditional anchor design, where single peak loads and extreme storm events are used to size the anchor.

A key risk to anchor failure within the whole-life approach is if large loads occur early in life, before the seabed has had time to strengthen from loading on the anchor. However, this is offset by a rise in anchor reliability in later life. Evaluating anchor failure risk over a floating wind turbine’s lifetime requires generating multiple statistically representative metocean time series, computing corresponding anchor load histories, and modelling the resulting seabed strength changes due to these loads.

This presentation will describe a novel probabilistic anchor design framework that has been developed to enable the design of anchors with accompanied quantified probabilities of anchor failure. The probabilistic anchor design framework is based on (1) a surrogate model that can rapidly generate many realisations of anchor loading history as a function of environmental conditions and (2) a seabed evolution model that tracks temporally varying seabed properties that affect anchor capacity as a function of anchor loading history.

The machine learning techniques that were used to train the surrogate model from 1,297 Orcaflex time-domain response simulations of an IEA 15 MW floating offshore wind turbine on the Volturn US-S semisubmersible platform will be discussed. The trained surrogate model was also used to predict a variety of entire wave-by-wave lifetimes of anchor loading histories to investigate the required anchor sizes for a range of probabilities of anchor failure. A comparison will be made between the probabilistic anchor design framework’s predicted required anchor sizes for given probabilities of failure, and the required anchor size from standard anchor design practice.

This study presents a comparative analysis of structural reliability methods for estimating the failure probability of nylon mooring lines in floating wind turbines, using case study data for the extreme tension load for a floating wind turbine in the Gulf of Maine and experimental test data for the tensile yarn strength of nylon. The First-Order Reliability Method (FORM), standard Monte Carlo simulation (in both X- and U-space), and Importance Sampling are evaluated under two levels of parameter uncertainty: (1) a low-uncertainty case (Resistance CoV = 2%, Load CoV = 4%) and (2) a higher-uncertainty case (Resistance CoV = 5%, Load CoV = 10%).Results show strong agreement among all the different methods, validating their applicability in relation to mooring line reliability assessment. However, computational analysis highlights the comparative inefficiency of crude Monte Carlo simulations, where the required sample size scales as N ≈ 400/Pf, where Pf is the lifetime failure probability of the mooring line. This work, conducted as part of the NYMOOR project, demonstrates that advanced methods such as FORM and Importance Sampling provide computationally efficient and practical alternatives for reliability-based design of floating wind turbine mooring systems at the individual line level.

Reducing Computational Cost in Floating Wind Design Through Multi-Fidelity Bayesian Optimisation
Ander Aristondo, Iñigo Urteaga, Markel Penalba, Vincenzo Nava and Miguel Esteras

The design of floating offshore wind turbines (FOWTs) presents a complex multi-disciplinary challenge, requiring the integration of aerodynamics, hydrodynamics, structural mechanics, and control systems. Among the critical subsystems of FOWTs, taut mooring systems—particularly those employed in Tension Leg Platforms (TLPs)—play a pivotal role in ensuring platform stability under stochastic marine conditions. However, the computational burden associated with high-fidelity simulations of these systems severely limits the feasibility of traditional optimisation techniques. This study addresses this challenge by implementing a Multi-Fidelity Bayesian Optimisation (MFBO) framework tailored for the design of taut mooring systems in floating wind applications.
MFBO is a sample-efficient global optimisation strategy that leverages surrogate modelling to reduce the number of expensive high-fidelity evaluations. The framework employs an autoregressive Gaussian Process (GP) model to capture the statistical relationships between multiple fidelity levels, enabling coherent predictions of the high-fidelity objective function. The optimisation process is guided by a Multi-Fidelity Expected Improvement (MFEI) acquisition function, which selects both the next sampling location and the fidelity level to query, balancing exploration, exploitation, and computational cost.
The methodology is first validated on a synthetic benchmark—the one-dimensional Forrester function—demonstrating that MFBO achieves comparable optimisation performance to single-fidelity Bayesian optimisation (SFBO) while reducing computational cost by over 50%. The surrogate model effectively integrates low- and mid-fidelity data to inform high-fidelity predictions, with the autoregressive GP showing superior robustness and accuracy compared to single-fidelity counterparts.
Subsequently, the MFBO framework is applied to a simplified engineering use-case: the optimisation of tendon stiffness and fairlead position in a TLP mooring system. The design objective is a composite cost function comprising peak tendon tension, platform pitch angle, and material cost. Two fidelity levels are defined: a low-fidelity static model and a high-fidelity time-domain dynamic simulation, both evaluated under severe environmental conditions representative of rated turbine operation. The MFBO algorithm successfully identifies optimal design parameters using only seven high-fidelity evaluations, supported by 37 low-fidelity samples. The resulting design achieves a balanced trade-off between structural performance and cost, with the surrogate model capturing the dominant trends in the design space and refining predictions in promising regions.
The results confirm the applicability of MFBO to offshore wind system design, where computational efficiency is paramount. The framework mimics the standard engineering design workflow—initial exploration using simplified models followed by targeted refinement with high-fidelity simulations—thus aligning with industry practices. Future work will extend the implementation to more complex design problems and explore alternative multi-fidelity surrogate models to address cases with lower inter-fidelity correlation. Additionally, further investigation into acquisition function tuning and sample rate ratios will be conducted to enhance the robustness and generalisability of the approach.