SUPERGEN Wind Wind Energy Technologies Phase 2 Dynamic Loading & Structures Dr Geoff Dutton Supergen Wind Phase 2 General Assembly Meeting 20 March 2014
Foundations: wave impact and floating structures Impact of wind farm control strategy on structural loads Innovative materials and manufacture VLM (aerodynamics) ABAQUS (structural dynamics)
Foundations: wave impact and floating structures
Wave interaction with structures Regular and extreme wave interaction with monopiles simulated using AMAZON-3D; wave force and run-up in good agreement with experimental and theoretical results OpenFoam tested and modified to simulate wave structure interaction making simulations more efficient. OpenFoam Amazon-3D Total cell number 970528 623844 Computing time 6 days 22 days Simulations extended to floating platforms
Experimental/numerical experiments on wave loading and run-up Range of input conditions similar to those found at Crown Estate round 3 sites Significant wave height up to 12 m. This is equivalent to a 50 year maximum wave height of 22 m. Wave buoy data and extrapolations 50 year maximum significant wave height contours Most tests concentrate on steep and near breaking waves as these are more critical and harder to model numerically.
Models and experimental set-up Experimental scale model tests in two wave tanks (Lancaster and Hull Universities) Generic monopile and Jacket structure models at 1/60 th and 1/90 th scale in 50 to 60 m deep water (equivalent depth) Tests on regular and irregular long-crested waves (uni-directional) and irregular shortcrested waves (multi-directional). Wave tank set-up at Hull University wave tank
Test results monopile regular waves The peak force on the structure in the wave direction varies with wave height and wave steepness. Semi-empirical Morison equation can give good fit to monopile tests where there is no deck impact Deck slam No deck slam Non-dimensional wave force against wavenumber Peak measured force on the structure. Modelled force using Morison equation is also shown. Wave run-up at the front of the structure causes wave impact on the horizontal deck which significantly increases the peak wave force on structure (deck slam). Around 50% increase in peak force with deck slam was observed on average.
Test results jacket regular waves Jacket structure shows consistently lower loads than equivalent monopile (mean 46% reduction) and is less sensitive to wave steepness. Jacket structure is also experiences lower run-up volume and hence reduced deck slam loads. Peak wave forces on the jacket structure are sensitive to wave direction and water height. This makes empirical models hard to use for the more complex structures.
Test results Irregular waves For irregular waves the time averaged force on the jacket structure is less than half the equivalent force on the monopile. The orientation of the jacket structure is not important. Time averaged forces are easier to predict empirically than extreme forces. Predicted and measured spectra show good similarity for irregular short-crested waves Time averaged forces on the models for irregular waves
Rogue waves Rogue waves are of particular importance to offshore wind turbines as they may occur when the turbine is operating, potentially giving extreme wave loads at the same time as peak wind loads. Tests into the formation and development of rogue waves in realistic conditions were carried out in a 50 m by 70 m wave tank in Trondheim the Marintek Ocean Basin. The tests investigated the effects of complex sea states, such as crossing bi-modal swell and wind sea. These conditions are relatively common in storms in the North Sea. Bi-modal test conditions Rogue waves up to 2.5 times the significant wave height were recorded for several test conditions. Rogue wave recorded at Marintek
Rogue waves Excess kurtosis has been shown to be linked to rogue wave activity and this is supported by the test results shown right. Empirical models linking kurtosis to directional spreading have been published (Mori, Onorato and Janssen, 2011) but have not been used for bimodal spectra. For bi-modal spectra the directional spreading can be estimated from the spectral peak or from the whole spectrum. The results fit the model well regardless of the method used for estimating directional spreading. Directional spreading (rad) Experimental results (circles) plotted against model predictions (lines), after Mori et al. (2011)
Exemplar wind farm Numerical modelling using SWAN The scale model test results have been used to validate a SWAN model (Simulating WAves Nearshore) The SWAN model was then used to estimate the turbine wave shadow at wind farm scale using SUPERGEN exemplar wind farm The tests were run with constant wind, so wind wakes are not included in the model The results show that the wave shadows behind the support structures are small and wave height recovers quickly H s values with and without wind turbines
Wave impact and floating structures Forces dependent on significant wave height and steepness Regular waves peak in-line force for jackets ~50% less than monopiles Irregular waves time-averaged in-line force for jackets ~24% less than monopiles (but note directional effects) Deck slam and run-up can significantly increase the forces
Innovative materials and manufacture
Novel materials: interlaminar toughness Selective interfacial reinforcement Veils Nano-additives Through-thickness stitching and tufting Nano silica particles Use of 3D fibre formats: braiding & weaving 3D Fabric
T-joints Objective Investigate different material configurations with aim to increase the fatigue life of T-joint in wind turbine blades Methods T-joints were made of glass fibre fabric infused with epoxy resin T-joints were modified with veil layers, tufting and 3D weaving techniques to improve the interlaminar fracture toughness T-joints with different modifications were tested under quasi-static and fatigue loading to determine the mechanical properties A finite element failure analysis model was developed using ABAQUS to simulate delamination of composite T-joints.
T-joint coupon specifications
T-joint coupon specifications
Strength (MPa) Load (N) Static test results 25 20 5000 4000 Base Veil-PA Veil-PE Tufted 15 3000 10 2000 5 1000 0 Base PA PE T 2T 3D-An 3D-4L T-joint structures 3D woven T-joints were found to have the highest pull-out strength among all modifications. Compared with the base specimen, 3D woven T-joints showed around 100% improvement. The next highest pull-out strength was the T-joint modified with a layer of polyester veil in the interface of the flange and skin with 85% strength improvement. 0 0 1 2 3 4 5 Deflection (mm)
Fatigue test results (all T-joint coupons)
Applied stress (MPa) Fatigue test results (summary) 20 15 cycles=1000 cycles=10000 cycles=100000 10 5 0 Base PA PE 3D-4L Tufted 3D-An T-joint structure Tufting / 3D-weaving significantly improve fatigue life performance of modified composite T-joints. The proper use of veil layer (e.g. polyester ) also improved fatigue life performance compared with Base specimens.
Stress 12 (MPa) Stress 11 (MPa) Stress analysis (FE) 50 40 30 Load = 2000 N Load = 2500 N Load = 3000 N Load = 3500 N Load = 4000 N 20 10 0 40 30 A B C D 0 10 20 30 40 50 60 70 80 90 Distance from point A along ABCD (mm) Load = 2000 N Load = 2500 N Load = 3000 N Load = 3500 N Load = 4000 N cracks 20 10 0 A B C D 0 10 20 30 40 50 60 70 80 90 Distance from point A along ABCD (mm)
Applied stress (MPa) Damage processes: FE simulation FE simulation shows that increasing interlaminar fracture toughness increases the strength to failure. Experimental 15 10 Gic = 600 J/m^2 Gic = 450 J/m^2 Gic = 300 J/m^2 5 0 0 1 2 3 4 5 Deflection (mm)
Load (N) FE Simulation of a tufted T-joint The unit-strip model was developed which can be used to investigate how tufted yarns affect the delamination behaviour of tufted composite T-joints 5000 4000 3000 2000 1000 Not tufted Tufted Applied load 0 0 1 2 3 4 5 Deflection (mm) Tuft yarns pull-out FE Results of Tufted T-joint Rollers are fixed Y direction movement = 0 y x Tuft yarn fractured Tuft yarn pull-out 100 mm
Innovative materials and manufacture - conclusions Static and fatigue failure of T-joints has been studied experimentally and numerically 3D woven and tufted joints can significantly improve pull-out strength and fatigue resistance Numerical modelling can represent failure for the simple geometry materials Further work needed on more complex geometries and measuring fundamental fracture properties
Impact of wind farm control strategy on structural loads VLM (aerodynamics) ABAQUS (structural dynamics)
Wake Model Unsteady formulation of the vortex lattice method Atmospheric turbulence & wake impacting modelled on a Cartesian velocity grid. Assumes frozen turbulence, propagated with the grid at the mean hub height velocity. Atmospheric turbulence modelled using method of Veers (1988). Wake modelled as mean velocity deficit similar to profile of Ainslie (1988): 1 U r b 2 = Ae U Where A and b are coefficients based on C T.
Wake Model - Turbulence Simulation of time and space varying wind field at Λ = 3. L x = D and I u = 13%. Drop in admittance for smaller lengthscales / higher frequencies. To be validated against wind tunnel experiments.
Wind tunnel experiments Investigated turbine operation in the wake of an upstream turbine. 1:250 Scale model based on the Exemplar rotor. Lateral offsets investigated: Aligned 0.25 D 0.50 D 0.75 D 1.00 D Rotor thrust and blade root bending moments recorded on downstream turbine
Coupled wake/structural model Wind field - Turbulence - Upstream wake IC VLM (aerodynamics) timestamp azimuth (blade 1) section-load ijk (ib,iz) Calculate torque, set-angle Control algorithm to specify (Ω, pitch angle velocity(ib)) STFC-RAL ABAQUS (structural dynamics) Increment timestamp =Δt) New wind speed Increment azimuth Timestamp azimuth (blade 1) Ω set angle (ib) section-disp ijk (ib,iz) section-rot ijk (ib,iz) section-vel ijk (ib,iz)
The Power Adjusting Controller (PAC) Speed Adjustment Change in Torque Change in Pitch A jacket around the main controller Fast acting torque response adjusts power A speed adjustment prevents the change in torque from being countermanded by the central controller A slower pitch response returns the turbine to an equilibrium operating point
Varying turbine output to increase energy capture and reduce loads A Power Adjusting Controller (PAC) has been developed that allows wind turbines to adjust their power output by a set amount ΔP When several wind turbines are in one another s wake, there is the potential to increase the total power output: The turbines on the windward side have their power reduced Turbines in the wake experience higher winds due to this and therefore have increased power output. Lowering the power output has the potential to reduce the loads on a wind turbine
Percentage Increase in Total Power Percentage Increase in Total Power Varying turbine output to increase energy capture and reduce loads Model of 5 wind turbines in a column, with each subsequent turbine in the wake of the previous one was simulated. Two wake models (Frandsen and Jensen) and three separation distances were used across a range of wind speeds. A variety of power reductions for the front two turbines were used and the greatest increase in total power output achieved was recorded 0.6 4 0.5 0.4 0.3 0.2 0.1 600m Frandsen 800m Frandsen 1000m Frandsen 3.5 3 2.5 2 1.5 1 0.5 600m Jensen 800m Jensen 1000m Jensen 0 6 7 8 9 10 11 12 13 0 6 7 8 9 10 11 12 13 Wind Speed (m/s) Wind Speed (m/s)
Varying turbine output to increase energy capture and reduce loads For a range of power output reductions, the reduction in the Damage Equivalent Loads (DELs) on a single wind turbine was investigated
Generator power (W) Temperature ( C) Empirical model for monitoring the temperature of the power converter Temperature change in response to power change over the full envelope 200 IGBT Diode 150 HS T buffer T max 100 50 0 1000 2000 3000 4000 5000 6000 7000 8000 9000 10000 10 x 106 - In this simulation, the PAC causes the power to increase by more than 50% of the rated power in above rate wind speed (after 4000s), and the resultant temperature is depicted. - The change in temperature takes place slowly. 8 6 4 2 - The model could be useful in preventing the temperature from crossing Tmax 0 0 1000 2000 3000 4000 5000 6000 7000 8000 9000 10000 time (s)
Control dynamics and structural loads - conclusions Vortex Lattice Wake model can represents turbulence in the wind field and operation in upstream wake Coupling of the wake model with a full structural blade model only partially successful Power Adjusting Controller can improve overall power performance and reduce damage equivalent loads Power Adjusting Controller can (temporarily) deliver higher power output with suitable monitoring/modelling of components
PhD Theses Jamie Luxmoore: Experimental studies of the interactions between waves and wind turbine support structures in intermediate depth water (to be submitted) Amirhossein Hajdaei: Extending the fatigue life of a T-joint in a composite wind turbine blade (2013) Kuangyi Zhang: Fatigue behaviour of multiaxial Glass Fibre Reinforced Plastics composites used in wind turbine blade (to be submitted) David Hankin: Wake impacting on a horizontal axis wind turbine (to be submitted)
Acknowledgements EPSRC grant no. EEP/H018662/1 SUPERGEN Wind Energy Technologies Consortium Phase 2 For further information please contact: geoff.dutton@stfc.ac.uk