tecnico lano POLI di MI tecnico Short Course on Wind Energy - Multidisciplinary Design of Wind Turbine Blades -

Transcription

1 di MI tecnico POLI tecnico lano Short Course on Wind Energy - Multidisciplinary Design of Wind Turbine Blades - Carlo L. Bottasso Politecnico di Milano, Italy November 2011

2 Presentation Outline

3 Holistic Design of Wind Turbines - Annual Energy Production (AEP) - Noise - - Generator (RPM, weight, torque, drive-train, ) - Pitch and yaw actuators - Brakes - Pitch-torque control laws: - Regulating the machine at different set points depending on wind conditions - Reacting to gusts - Reacting to wind turbulence - Keeping actuator duty-cycles within admissible limits - Handling transients: run-up, normal and emergency shutdown procedures - GE wind turbine (from inhabitat.com) - Loads: envelope computed from large number of Design Load Cases (DLCs, IEC-61400) - Fatigue (25 year life), Damage Equivalent Loads (DELs) - Maximum blade tip deflections - Placement of natural frequencies wrt rev harmonics - Stability: flutter, LCOs, low damping of certain modes, local buckling - Complex couplings among rotor/drivetrain/tower/foundations (off-shore: hydro loads, floating & moored platforms) - Weight: massive size, composite materials (but shear quantity is an issue, fiberglass, wood, clever use of carbon fiber) - Manufacturing technology, constraints

4 Holistic Design of Wind Turbines Current approach to design: discipline-oriented specialist groups Lengthy loops to satisfy all requirements/constraints (months) Different simulation models Data transfer/compatibility among groups There is a need for multi-disciplinary optimization tools, which must: Be fast (hours/days) (on standard desktop hardware!) Provide workable solutions in all areas (aerodynamics, structures, controls) for specialists to refine/verify Account ab-initio for all complex couplings (no fixes a posteriori) Use fully-integrated tools (no manual intervention) They will never replace the experienced designer! but would greatly speed-up design, improve exploration/knowledge of design space

5 Holistic Design of Wind Turbines Focus of present work: integrated multi-disciplinary (holistic) constrained design of wind turbines, i.e. optimal coupled sizing of: Aerodynamic shape Structural members (loads, aero-servo-elasticity and controls) Constraints: ensure a viable design by enforcing all necessary design requirements Applications: Sizing of a new machine Improvement of a tentative configuration Trade-off studies (e.g. performance-cost) Modifications to exiting models Previous work: Duineveld, Wind Turbine Blade Workshop 2008; Fuglsang & Madsen, JWEIA 1999; Fuglsang, EWEC 2008; etc.

6 Optimization-Based Multi-Level Blade Design Cost: AEP Aerodynamic parameters: chord, twist, airfoils Cost: AEP/weigh (or cost model if available) Macro parameters: rotor radius, max chord, tapering, Cost: Blade weight (or cost model if available) Structural parameters: thickness of shell and spar caps, width and location of shear webs Controls: model-based (selfadjusting to changing design)

7 Preliminary design Design refinement Fully automated links

8 Fine level: 3D FEM Coarse level: 2D FEM section & beam models Definition of sectional design parameters - ANBA 2D FEM sectional analysis - Computation of 6x6 stiffness matrices - Definition of geometrically exact beam model - Span-wise interpolation When SQP converged SQP optimizer min cost s.t. constraints Constraints: - Maximum tip deflection - 2D FEM ANBA analysis of maximum stresses/strains - 2D FEM ANBA fatigue analysis - Compute cost (mass) - Definition of complete HAWT Cp-Lambda multibody model - DLCs simulation - Campbell diagram DLC post-processing: load envelope, DELs, Markov, max tip deflection Automatic 3D CAD model generation by lofting of sectional geometry Automatic 3D FEM meshing (shells and/or solid elements) Update of blade mass (cost) Analyses: - Max tip deflection - Max stress/strain - Fatigue - Buckling Verification of design constraints Constraint/model update heuristic (to repair constraint violations)

10 Cp-Lambda highlights: Geometrically exact composite-ready beam models Generic topology (Cartesian coordinates+lagrange multipliers) Dynamic wake model (Peters- He, yawed flow conditions) Efficient large-scale DAE solver Non-linearly stable time integrator Fully IEC compliant (DLCs, wind models) Rigid body Geometrically exact beam Revolute joint Flexible joint Actuator

11 ANBA (Anisotropic Beam Analysis) cross sectional model (Giavotto et al., 1983): Evaluation of cross sectional stiffness (6 by 6 fully populated) Recovery of sectional stresses and strains Compute sectional stiffness of equivalent beam model Compute cross sectional stresses and strains Rigid body Geometrically exact beam Revolute joint Flexible joint Actuator

12 Measurement noise Wind

13 Blade parameterization: Chord and twist shape functions deform a baseline configuration Constraints: Noise constraint (V tip): regulation in region II1/2 Torque-TSR stability Max chord Richer shape with fewer dofs

14 Modeling: Extract reduced model from multibody one Linearize reduced model Synthesize controller: Compute LQR gains Update process: Update cross sectional models Compute beam stiffness and inertial properties Analyses: DLCs (IEC61400: load envelope, fatigue DELs) Eigenfrequencies (Campbell diagram) Stability Compute constraints: Max tip deflection Frequency placement Update multibody model Analyses: Eigenfrequencies Max deflection Fatigue Max stresses/strains Buckling Update non-structural mass Analyses: Transfer loads from multibody to cross sectional models Recover sectional stresses and strains Compute cost function: Weight Compute constraints: Stress/strains safety margins

15 Structural Blade Modeling Cross section types Sectional structural dofs Spanwise shape functions Location of structural dofs and load computation section Load computation section Twisted shear webs Straight webs Caps extend to embrace full root circle

16 3D FEM Modeling 3D CAD with solid and shell (no offsets) meshing directly from preliminary design optimizer: Webs + Web core + Spar caps + LE & TE reinforcements Internal skin + Skin core + External skin = Complete model Analyses

17 FEM Analyses and Update of Aeroelastic Models Stress/strain/fatigue: - Fatigue constraint not satisfied at first iteration on 3D FEM model - Modify constraint based on 3D FEM analysis - Converged at 2 nd iteration Peak stress on initial model Buckling: - Buckling constraint not satisfied at first iteration - Update skin thickness - Update trailing edge reinforcement strip - Converged at 2 nd iteration Fatigue damage constraint satisfied Increased skin thickness Increased trailing edge strip

18 Parameter: radius, max chord, etc. Example: tapering Example: AEP over weight

19 2MW 45m Wind Turbine Blade Currently undergoing certification at TÜV SÜD CNC machined model of aluminum alloy for visual inspection of blade shape Design developed in partnership with Gurit (UK)

20 di MI tecnico POLI tecnico lano An Exercise in Blade Design: Integrated Active and Passive Load Control in Wind Turbines

21 Need for Load Mitigation Trends in wind energy: Increasing wind turbine size Off-shore wind To decrease cost of energy: Reduce extreme loads Reduce fatigue damage Limit actuator duty cycle Ensure high reliability/availability

22 Active Load Mitigation: Pitch Control Individual blade Pitch Control (IPC) Inner loop (collective pitch): regulation to set point and alleviation of gust loads Outer loops (individual pitch): reduction of - Deterministic (periodic) loads due to blade weight and non-uniform inflow - Non-deterministic loads, caused by fast temporal and small spatial turbulent wind fluctuations Uniform wind Turbulent wind

23 Active Load Mitigation: Predictive LiDAR-Enabled Pitch Control LiDAR: generic model, captures realistically wind filtering due to volumetric averaging Receding Horizon Control: model predictive formulation with wind scheduled linear model, real-time implementation based on CVXGEN Non-Homogeneous LQR Control: approximation of RHC, extremely low computational cost LiDAR prediction span Reduced peak values Reduced peak values Reduced peak to peak oscillations

24 Active Load Mitigation: Distributed Control Flow control devices: TE flaps Microtabs Vortex generators Active jets (plasma, synthetic) Morphing airfoils (Credits: Risoe DTU) (Credits: Risoe DTU) (Chow and van Dam 2007) (Credits: Smart Blade GmbH)

25 Active Load Mitigation: Limits and Issues Pitch control: Limited temporal bandwidth (max pitch rate 7-9 deg/sec) Limited spatial bandwidth (pitching the whole blade is ineffective for spatially small wind fluctuations) Distributed control: Alleviate temporal and spatial bandwidth issues Complexity/availability/maintenance Sensor-enabled control solutions: Complexity/availability/maintenance Off-shore: need to prove reliability, availability, low maintenance in hostile environments

26 Passive Load Mitigation Passive control: loaded structure deforms so as to reduce load Two main solutions: - Bend-twist coupling (BTC): exploit anisotropy of composite materials - Swept (scimitar) blades Angle fibers in skin and/or spar caps Potential advantages: no actuators, no moving parts, no sensors Other passive control technologies (not discussed here): - Tuned masses (e.g. on off-shore wind turbines to damp nacelle-tower motions) - Passive flaps/tabs -

27 Present study: Objectives Design BTC blades (all satisfying identical design requirements: max tip deflection, flap freq., stress/strain, fatigue, buckling) Consider trade-offs (load reduction/weight increase/complexity) Identify optimal BTC blade configuration Integrate passive BTC and active IPC Exploit synergies between passive and active load control Baseline uncoupled blade: 45m Class IIIA 2MW HAWT

28 Fully Coupled Blades 1. Identify optimal section-wise fiber rotation Consider 6 candidate configurations Skin angle Spar-cap angle BTC coupling parameter: α = K BT K B K T

29 Fully Coupled Blades: Effects on Weight Spar-caps: steep increase Skin: milder increase Spar-cap/skin synergy Stiffness driven design (flap freq. and max tip deflection constraints): Need to restore stiffness by increasing spar/skin thickness

30 Fully Coupled Blades: Load Reduction Spar-cap/skin synergy: good load reduction with small mass increase

31 Fully Coupled Blades: Mechanism of Load Reduction

32 Fully Coupled Blades: Effects on Duty Cycle Less pitching from active control because blade passively selfunloads Much reduced life-time ADC

33 Partially Coupled Blades 2. Identify optimal span-wise fiber rotation: 5 candidate configurations Reduce fatigue in max chord region Avoid thickness increase to satisfy stiffness-driven constraints

34 Partially Coupled Blades: Effects on Mass Fully coupled blade Too little coupling

35 Partially Coupled Blades: Effects on Loads F30: load reduction close to fully coupled case

36 Partially Coupled Blades: Effects on Duty Cycle Best compromise: similar load and ADC reduction as fully coupled blade, decreased mass

37 Integrated Passive and Active Load Alleviation Individual blade pitch controller (Bossaniy 2003): Coleman transform blade root loads PID control for transformed d-q loads Back-Coleman-transform to get pitch inputs Baseline controller: MIMO LQR

38 Integrated Passive and Active Load Alleviation Two IPC gain settings: 1. Mild: some load reduction, limited ADC increase 2. Aggressive: more load reduction, more ADC increase Five blade/controller combinations: BTC: best coupled blade + collective LQR IPC1: uncoupled blade + mild IPC BTC+IPC1: best coupled blade + mild IPC IPC2: uncoupled blade + aggressive IPC BTC+IPC2: best coupled blade + aggressive IPC

39 Integrated Passive/Active Control: Effects on Loads Synergistic effects of combined passive and active control

40 Integrated Passive/Active Control: Effects on Duty Cycle Not significant: ADC very small here Same ADC as baseline (but great load reduction!)

41 Conclusions of Blade Design Exercise Optimization-based blade design tools: enable automated design of blades and satisfaction of all desired design requirements BTC passive load control: - Skin fiber rotation helps limiting spar-cap fiber angle - Partial span-wise coupling limits fatigue and stiffness effects Reduction for all quality metrics: loads, ADC, weight Combined BTC/IPC passive/active control: - Synergistic effects on load reduction - BTC helps limiting ADC increase due to IPC (e.g., could have same ADC as baseline blade with collective pitch control) Outlook: Manufacturing implications of BTC and partially coupled blades Passive distributed control and integration with blade design and active IPC control

43 Comprehensive Wind Turbine Modeling Multibody models: - Blades - Tower - Nacelle - Drive train - Actuators Wake models: Induced velocity Induced swirl Wake dynamics Wake turbulence (stiffness and mass spanwise distributions) Aerodynamic models: Tower/rotor interference Nacelle, nacelle/rotor interference Airfoil aerodynamic models: Lift, drag and moment Stall and dynamic stall 3D root (Coriolis) effects Tip, root and hub losses Unsteady aerodynamics Wind models: Direction Spatial distribution Temporal fluctuations Upstream wind turbine wakes Terrain orography effects Mechanical properties of foundations Hydro-dynamic models: Wave and current loads Floatation Mooring lines (airfoil lift, drag and moment, and dynamic stall loops)

44 Comprehensive Wind Turbine Modeling

45 Comprehensive Approach to Wind Turbine System Identification Fidelity of overall model depends on fidelity of sub-models & of their couplings Fidelity of sub-models depends on: Their ability to capture the relevant physics Correct tuning of their parameters Divide and conquer approach: identify all sub-models from specific observations Identification of specific sub-models Blade structural properties Airfoil properties Wake properties Drive-train properties Validated comprehensive model Sub-model specific experiments and measures

46 Identification of Blade Properties 1. Ground test experimentation Load deflection tests Laser scan of loaded blade 5. Improved model 2. Natural freqs and static deflections Aeroelastic simulations Control laws 4. New identified blade properties 3. System identification Constrained optimization Goal: match static deflection and modes Unknown: corrective function of the blade distribution properties Blade model: FEM POLI-Wind Cost function: Research WLS and ML Lab Optimization algorithm: SQP

48 WT 2, the Wind Turbine in a Wind Tunnel Goal: aeroelastically-scaled wind tunnel model of the Vestas V90 wind turbine with individual blade pitch and torque control Applications: Testing of advanced control laws and supporting technologies Testing of extreme operating conditions Tuning of mathematical models Aeroelasticity and system identification of wind turbines Multiple wind turbine interactions Off-shore wind turbines (moving platform actuated by hydro-structural model) 13.8x3.8m, 14m/s, civil section: Turbulence < 2% With turbulence generators = 25% 13m turntable 4x3.8m, 55m/s, aeronautical section: Turbulence <0.1% Open-closed test section Civil-Aeronautical Wind Tunnel of the Politecnico di Milano Aeroelastically scaled blades (70g, 1m) Pitch actuator: Zero backlash gearhead Built-in encoder Main shaft with torque meter Pitch actuator electronics Rotor sensor electronics Slip ring Torque actuator: Planetary gearhead Torque and speed control Conical spiral gears

49 WT 2, the Wind Turbine in a Wind Tunnel Good aerodynamic performance even at low Reynolds (and currently being further improved) Blockage correction verified by RANS CFD

50 The Politecnico di Milano Wind Tunnel

51 The Politecnico di Milano Wind Tunnel Turn-table 13 m Turbulence (boundary layer) generators Low speed testing with vertical wind profile Multiple wind turbine testing (wake-machine interaction) High speed testing Aerodynamic characterization (C p -TSR-β & C F -TSR-β curves)

52 Design of an Aero-elastically Scaled Composite Blade Carbon fiber spars Objective: size spars (width, chordwise position & thickness) for desired sectional stiffness within mass budget Cost function: sectional stiffness error wrt target (scaled stiffness) Constraints: lowest 3 frequencies Wind Turbine Design Airfoil cross section Carbon fiber spars for desired stiffness Sectional optimization variables (position, width, thickness) Span-wise shape function interpolation Chordwise Position Width Rohacell core with grooves for the housing of carbon fiber spars Thickness Film of glue to close pores and ensure smooth finish ANBA (ANisotropic Beam Analysis) FEM cross sectional model: Evaluation of cross sectional stiffness (6 by 6 fully populated matrix)

53 Off-Shore Aero-Elastic Model Goal: aeroelastically-scaled wind tunnel model of off-shore wind turbine with individual blade pitch and torque control Applications: Testing of control laws Damping enhancement controllers Load-reducing controllers Floating platform effects on stability 6 DOF moving platform Platform motion WT response Proof of concept, 2 DOF hydraulic actuation (prescribed) Real-time PC running mathematical model of wet part of offshore machine (hydro-elastic model)

55 Individual Higher Harmonic Pitch Control Inner loop: regulation to set point and alleviation of gust loads Middle loop: higher harmonic reduction of deterministic (periodic) loads on blades or shaft/tower (mainly due to blade weight and non-uniformity of spatial distribution of wind over rotor disk) Outer loop: non-deterministic loads, caused by fast temporal and small spatial scales 2.5MW wind turbine: Uniform wind Turbulent wind

56 Observed Yaw Angle [ ] Flap root [Nm] Hub Wind [m/sec] Wind Observers Goal: obtain an instantaneous description of the wind field over the rotor disc Applications: Feed forward and/or scheduled control laws Simple low cost alternative to LiDAR (but no prediction capabilities) Collection of accurate wind field data using wind turbines as wind sensors Blade loads x Load sensors Wind Observer States describing wind field Real Observed Time [sec] Wind Turbine Blade load sensors Approach: Demodulate blade loads Wind spatial field is contained in load phasing and magnitude

58 Stability Analysis of Wind Turbines The small amplitude perturbation response of a wind turbine around a trim condition can be described by a linear time periodic (LTP) system Approaches: Perturb about a trim point Identification of (Periodic) ARX + periodic exogenous inputs Obtain damping factors (LTP extension) for modes of interest Quality of identified model First blade edgewise mode Rotor in-plane mode Loss of stability of first collective in-plane rotor mode for a 80m span blade

59 Conclusions Simulation: key enabler for design and optimization Model validation: key enabler for reliable predictions System identification: synergy between full scale and scaled testing Field (full-scale) testing Validated mathematical models Wind tunnel (scaled) testing Upscaling Vast field of ideas/technologies that will be explored/developed: Advanced blades Advanced controls Offshore wind (figures from University of Stuttgart, GE, NASA, other sources)

60 Acknowledgements Thanks to the POLI-Wind team! Special thanks to M. Bassetti, P. Bettini, M. Biava, F. Campagnolo, S. Calovi, S. Cacciola, F. Cadei, G. Campanardi, M. Capponi, A. Croce, G. Galetto, L. Maffenini, P. Marrone, M. Mauri, S. Rota, G. Sala, A. Zasso of the Politecnico di Milano Funding provided by Vestas Wind Systems A/S, Clipper Windpower, Alstom Wind, DOE National Renewable Energy Laboratory, Italian Ministry of Education, University and Research