https://hal-ifp.archives-ouvertes.fr/hal-04033056Peyrard, C.C.PeyrardEDF - EDFRobaux, FabienFabienRobauxEDF - EDFBorràs Nadal, AdriàAdriàBorràs NadalIFPEN - IFP Energies nouvelles - IFPEN - IFP Energies nouvellesJoulin, Pierre-AntoinePierre-AntoineJoulinIFPEN - IFP Energies nouvelles - IFPEN - IFP Energies nouvellesMayol, Maria-LauraMaria-LauraMayolIFPEN - IFP Energies nouvelles - IFPEN - IFP Energies nouvellesEldevik, S.S.EldevikDNV GLGuiton, MartinMartinGuitonIFPEN - IFP Energies nouvelles - IFPEN - IFP Energies nouvellesCousin, AlexisAlexisCousinIFPEN - IFP Energies nouvelles - IFPEN - IFP Energies nouvellesBenoit, M.M.BenoitEDF - EDFDimitrov, N.N.DimitrovDTU - Danmarks Tekniske Universitet = Technical University of DenmarkLovera, AnaïsAnaïsLoveraEDF - EDFFereira, C.C.FereiraDTU - Danmarks Tekniske Universitet = Technical University of DenmarkAero-servo-hydroelastic model uncertaintyHAL CCSD2022[SDE.IE] Environmental Sciences/Environmental EngineeringCouëdel, NadineHIghly advanced Probabilistic design and Enhanced Reliability methods for high-value, cost-efficient offshore WIND - HIPERWIND - 101006689 - INCOMING - 2023-03-16 19:22:452023-03-17 04:19:242023-03-16 19:32:00enReportsapplication/pdf1This report presents the work related to Deliverable 3.3. It is dedicated to the estimation of Aero-Servo-Hydro-Elastic (ASHE) model uncertainty for what concerns the hydrodynamic loading (section 4) and the aerodynamic loading (section 5). Indeed, the models which are commonly used to predict the loads on the main components of an Offshore Wind Turbine (OWT), given some conditions of wind and waves, simplify the physics to keep the computation time affordable. Note that the influence of submarine current is not considered in this report for the sake of simplification. These “engineering” models are thus introducingepistemic (i.e. which could be reduced by increasing knowledge) model uncertainties on their resulting outputs which are estimated here through comparisons with the results of higher fidelity models.The results are documented for the two offshore case studies of HIPERWIND: the 2.3 MW turbine on a monopile within the Teesside (United Kingdom) wind farm, and the IEA15MW wind turbine on UMaine semi-submersible floater in South Brittany site (France). The latest is a modified version of the original design which was proposed by NREL and UMaine [1.][2.]. The changes are mainly on the tower to avoid a risk of resonance with natural frequencies too close to the 3P frequency [3.]. In this report, we also present an additional modification with clump weights on the mooring lines to adapt their dynamics to the extreme 50-year returning period waves of South Brittany.Summary of hydrodynamic loading uncertainty:The hydrodynamic loadings have been investigated on both floating and bottom fixed foundations:- On the bottom fixed side, the focus was made on the damage induced by the hydrodynamic loads on two monopiles at the Teesside location: the original 2.3MW monopile and a larger one, designed within HIPERWIND based on the DTU 10MW reference turbine. A comparison has been made between a base case engineering hydrodynamic method (based on the EDF R&D DIEGO solver and the MacCamy & Fuchs correction) and more reliable hydrodynamic load estimation methods, based on Potential Flow solvers. An extensive set of sea conditions have been used to perform the comparisons, covering the full Teesside Hs-Tp scatter diagram. The conclusions suggested that variations of +/- 10% of the hydrodynamic Damage Equivalent Load (DEL) can be obtained due to simplification made by the application of the MacCamy-Fuchs correction. In addition, a tendency of the engineering model to overestimate by 5 to 10% the DEL due to the high-frequency loads has been found, in a large range of wave periods. For high and long waves however, the engineering model underestimates quite significantly the DEL (-10% to -50%). We have also noted that these levels of discrepancies between engineering models and more accurate method is in the same order as the uncertainty obtained when using different stretching models. Recommendations on the hydrodynamic DEL uncertainty level have been made.- On the floating side, the focus was made on the level of damping applied on the floating foundations and in particular on the drag loads. The University of Maine 15MW floater has been used and several uncertainty sources have been investigated using CFD:o solver: OpenFoam and neptune_cfd have been used by IFPEN and EDF R&D respectively,solver: OpenFOAM® and neptune_cfd have been used by IFPEN and EDF R&D respectively,o kinematic conditions: fixed body in waves and motion of the body in still water,o extraction of the drag coefficients: 2 methods have been applied, the first one based on least square and the order one on harmonics identification.Five zones on the floater have been selected for drag coefficient estimation: the side column, the central column, the pontoons, and the base of the columns. For each zone, the drag coefficients are obtained for a large range of Keulegan-Carpenter numbers, allowing to characterize the damping for a large range of conditions. Recommendations are made for the drag coefficient choice and their expected variations.Summary of aerodynamic loading uncertainty:An exhaustive comparison between the usual Blade Element Momentum (BEM) approach and a more accurate Vortex-based model has been conducted for both the Teesside fixed, and the South Brittany floating OWT. The BEM approach is provided for 3 different software: Deeplines WindTM (DLW, IFPEN), Diego (EDF) and HAWC2 (DTU). The input space is 3D for the fixed case with mean speed, turbulence intensity and yaw misalignment. For the floating case, the input space is 6D. It is composed of 3 parameters for wind: wind speed, the absolute wind direction and turbulence intensity; and 3 parameters for waves: wave height, wave period and wave direction (6D). The outputs concern the production, Damage Equivalent Loads (DEL) at the blade root, forces along the blades and forces integrated over the rotor. From an initial space-filling Design of Experiments (DoE), an iterative procedure has been chosen to define the points at which costly Vortex simulations are done, in order to reduce the uncertainty on the difference between BEM and Vortex for both production and design outputs. It is exploiting the characteristics of Gaussian Processes computed for each output. The first contribution to the model uncertainty is estimated from the differences between the 3 BEM models. Then it is found that outside of a domain of both high turbulence and high wind speed, which is rarely encountered, the differences between BEM and Vortex remain small, particularly for the fixed OWT. Another contribution to the model uncertainty is that the BEM vs Vortex differences are higher for the floating wind turbine. Even though, the 6D input space for the South Brittany case is sparsely explored due to the high computational cost, differences between the two approaches can be noticed, specifically for the large floating OWT case. A major part of these differences may however be due to the use of controller setting chosen for BEM, which is thus non optimal for the Vortex model. Due to this controller influence, the estimated model uncertainty is an upper bound of the true uncertainty, which is more sensitive to the wind parameters than to the wave parameters.Research Significance:The results presented in this report provide a comprehensive estimate of the ASHE model uncertainty for both the aerodynamic loads and the hydrodynamic fluid-structure interaction. They are provided for both a fixed and a floating case, and by considering their distribution in the input parameters representing the environmental conditions. This information is useful for the designer to interpret the results of the ASHE simulations keeping in mind the usual “engineering” model limits. From a more quantitative point of view, the uncertainty can also be used to replace pure a priori distributions in uncertainty reduction study, for example to compute reliable design, with the limit of the extrapolation to other case studies. This will be for instance the case in the next work of HIPERWIND which will test and develop different approaches for both Ultimate and Fatigue Limit State Analysis.