This document summarizes a project to study the effects of surface erosion and roughness on wind turbine performance. Wind tunnel tests were conducted on airfoils with simulated surface contamination, and computational models were developed to predict airfoil and turbine performance degradation from roughness. Testing showed roughness reduced maximum lift and increased drag on airfoils. Modeling predicted a 4-8% reduction in annual energy capture for a 5MW reference turbine from representative levels of airfoil roughness. The study aims to improve understanding of roughness effects to help mitigate performance losses in wind turbines.
1. 1
Surface Erosion and
Roughness Effects on
Airfoil and Wind
Turbine Performance
C.P. (Case) van Dam
Department of Mechanical &
Aerospace Engineering
University of California, Davis
2014 Wind Turbine Blade Workshop
Albuquerque, NM
27 August 2014
2. 2
Contributors
• Sandia National Laboratories
– David Maniaci
– Mark Rumsey
– Matt Barone
• Texas A&M
– Ed White
– Robert Ehrmann
– Ben Wilcox
• UC Davis
– Chris Langel
– Ray Chow
– Owen Hurley
3. 3
Outline
• Background
• Project outline
– Field Measurements
– Wind tunnel testing
– Computational modeling
• Model validation
• Airfoil results
• Turbine performance
effects
– NREL 5-MW rotor
• Conclusions & Next steps
Source: Mayda
4. 4
Windplant Loss Categories
Walls & Kline (2012)
• Power losses can be as much as 20-30% in
state of the art windplants:
– Wake losses
– Turbine availability
– Balance of Plant (BOP) availability
– Electrical
– Environmental
– Turbine performance
– Curtailment
5. 5
Windplant Loss Categories
Walls & Kline (2012)
• Power losses can be as much as 20-30% in
state of the art windplants:
– Wake losses
– Turbine availability
– Balance of Plant (BOP) availability
– Electrical
– Environmental
– Turbine performance
– Curtailment
6. Environmental - Turbine Performance
6
Losses
• These losses typically range from 1% to 10%
• Impact on rotor aerodynamics:
– Icing
• Glaze
• Hoar
– Blade soiling
– Blade erosion
– Drop in air density (high temperature)
– Turbulence, shear, etc.
8. 8
Background - I
• Early stall controlled, constant speed wind
turbines were severely affected by blade
surface contamination and erosion. Large
performance losses resulted (40% in peak
power, ≥ 20% in energy capture).
• Development and introduction of blade
section shapes that were less roughness
sensitive mitigated this issue.
• Issue was focus of Wind Energy Conversion
System Blade-Surface Roughness Workshop
at NREL on April 20-21,1993.
9. 9
Blade Contamination
Moroz & Eggleston (1993)
• Surface soiling induced loss
in power for fixed pitch, stall
controlled rotors was big
problem
• Surface contamination
caused by insect
contamination, dust, erosion
of gel coat
• Surface roughness causes
reduced sectional lift curve
slope and maximum lift
coefficient, and increased
sectional drag
• Effect greater for stall
controlled rotors than pitch
controlled rotors
10. 10
Blade Contamination
Tangler (1993)
• Surface contamination
induced loss in power was
problem for stall controlled
rotors
• Aerostar blade uses NACA
4415-4424 airfoils
• NREL blade uses S805A,
806A, 807 airfoils
• NREL airfoils designed to
have less (maximum) lift
sensitivity to surface
roughness
• Tests show reduced loss in
turbine power due to
surface roughness for
NREL blade
11. 11
Background - II
• Effect of (small) roughness:
– It may cause premature transition from laminar to turbulent
boundary layer state
– It may cause boundary layer separation
– It may cause flow unsteadiness
– It removes energy from flow (increased skin friction)
– Effect depends on:
• Roughness height
• Roughness chordwise location
• Roughness density
• Pressure gradient
• Unit Reynolds number
• Mach number
12. 12
Background - III
• Variable speed, variable pitch turbines started to supersede the
constant speed, fixed pitch turbines and this significantly
mitigated the problem.
• However, a resurgence of the surface roughness problem has
occurred:
– More awareness as a result of improved windplant performance
analysis methods
– Higher maximum thickness-to-chord ratio (t/c) blade sections
– Higher lift-to-drag ratio (L/D) blade sections
– Higher Reynolds numbers
• Combination of high density altitude and blade surface
roughness can be especially troublesome.
• Because of size of turbines, blade washing is often cost
prohibitive.
• Detailed knowledge of loss mechanisms is still missing.
• Computational tools to analyze roughness sensitivity of airfoils
are missing.
13. 13
Surface Roughness and Erosion
Project
• Effects of Surface Contamination and Erosion on
Wind Turbine Performance
• Project started in April 2012
• Team:
– Sandia National Laboratories, Albuquerque
– Texas A&M University
– University of California, Davis
• Tasks:
– Field measurements of surface roughness and erosion
– Wind tunnel testing of effect of surface roughness and
erosion on airfoil performance
– Development of computational roughness model to account
for effect on aerodynamic performance of airfoils, blades,
rotors
– Correlate wind tunnel and CFD results
14. 14
Wind Tunnel
• Oran W. Nicks Low Speed Wind
Tunnel at Texas A&M
• Closed return tunnel
• Test section 7 ft × 10 ft
• Maximum velocity of 90 m/s
• Blockage of 4.8%
• Turbulence intensity of 0.25%
• Maximum Rec = 3.6 × 106 based
on loading at maximum lift
conditions
• Maximum Rec = 5.0 × 106 to α =
4°
Model installed in wind tunnel
freestream
16. pressure side suction side pressure side suction side
16
Distributed Roughness
Random insect distribution with 3% coverage. Random insect distribution with 15% coverage.
20. 20
Computational Modeling - I
• OVERFLOW-2
– Overset, multigrid, compressible Reynolds-averaged Navier-Stokes flow
solution method
– Semi-public domain
– Method newly developed Roughness Model has been coded into
• Reynolds averaged Navier-Stokes Equations
– Remove turbulent fluctuations from flow equations. All eddy scales are
ignored and mean flow can then be resolved with coarser computational
grid.
• Turbulence Modeling
– To properly account for turbulent fluctuations, there must be a way to
approximate the effect of the removed scales. In RANS methods, these
fluctuations are accounted for in the Reynolds stress terms
– Surface roughness has a prominent effect on this process
• Transition Modeling
– Baseline turbulence models must either assume fully laminar or “fully”
turbulent. Need additional correlation to automate switch between laminar
and turbulent.
– Surface roughness has a prominent effect on this process
22. 22
Computational Modeling - III
• Existing transition model Langtry-Menter:
– Recently developed
– Two variable model
• Local momentum thickness parameter, transition onset when
local momentum thickness ≥ critical momentum thickness
• Intermittency parameter governs growth turbulent kinetic
energy from transition onset to fully turbulent
• Roughness model adds 3rd variable to Langtry-
Menter transition model:
– Roughness amplification parameter (Ar)
• Turbulence model modified to account for surface
roughness effects
– Currently based on Wilcox
23. 23
Roughness Variable (Ar) Distribution
• There is a direct correlation
between distribution of Ar and
skin friction due to dependence
on wall shear stress (τw)
Flat plate flow, Re = 1.34 million, Ma = 0.30
Top: Distribution of Ar variable along flat plate
Bottom: Corresponding skin friction distribution
Ar Rough Wall Boundary
= f (k+ )
k+ =
U!ks
"
ks
= Roughness Height
U!
=
!w
#
Cf
=
!w
1
2 #U2
24. 24
Initial Validation Cases
• Flat plate with distributed sand-grain
roughness of varying heights (Feindt, 1956)
– Zero pressure gradient
– Adverse pressure gradient
• NACA 0012 with leading edge roughness
(Kerho & Bragg, 1997)
• Texas A&M tunnel, NACA 633-418
– Clean
– Distributed roughness
25. 25
Effect of Roughness Height on
Skin Friction
Flat plate, zero-pressure gradient, Feindt (1956)
Re
k =
!U
k
k
μ
26. 26
Comparison of Measured and
Predicted Effect of Roughness on
Transition
Flat plate, zero-pressure gradient, Feindt (1956)
Re
k =
!U
k
k
μ
27. 27
Comparison of Measured and
Predicted Boundary Layer Profiles
NACA 0012, Re = 1.25 × 106, α = 0°
1/2 in. roughness strip applied at s = 4 mm (x/c = 0.0018 - 0.0191)
• Wind tunnel measurement from Kerho & Bragg (1997)
• Slight lag in boundary layer development at early stations
• Profiles match well at later stations
28. 28
Comparison of Measured and
Predicted Boundary Layer States
NACA 0012, Re = 1.25 × 106, α = 0°
29. 29
Comparison of Measured and
Predicted Drag Polars
NACA 633-418, Clean surface, Re = 1.6 × 106, Texas A&M tunnel
30. 30
Comparison of Measured and
Predicted Transition Location
NACA 633-418, k/c = 170 × 10-6 @ x/c = -0.12:0.04, Re = 1.6 × 106,
Texas A&M tunnel
31. 31
Comparison of Measured and
Predicted Transition Location
NACA 633-418, k/c = 170 × 10-6 @ x/c = -0.12:0.04, Re = 2.4 × 106,
Texas A&M tunnel
32. 32
NREL 5-MW Rotor
• Geometry based on
6MW DOWEC rotor
– Conceptual off-shore
turbine design
– ECN (Energy Research Centre
of the Netherlands)
• Rotor diameter
truncated and hub
diameter reduced
33. 33
NREL 5-MW Rotor
• Rotor diameter =126 m
• Specific power = 401 W/m2
• 12.1 RPM
• 3 m hub diameter
• 61.5 m blade length
• 4.7 m max chord
• 13.3° inboard twist
• 3 m/s cut-in speed
• 25 m/s cut-out
• 12 m/s rated speed
34. 34
Performance Prediction Using
Computational Roughness Model
• Six different airfoil profiles
• Airfoils analyzed using OVERFLOW-2 in both
“clean” and “rough” configuration
• Roughness applied from 5% chord on lower to
5% chord on upper surface
• Height of roughness set at k/c = 240 × 10-6 ( k
= 0.24 mm or 0.001 in. for a chord of 1 m)
• Corresponds to relatively heavy soiling
36. Effect of Blade Roughness on Turbine
36
Power
WT-Perf, NREL 5-MW turbine, Roughness height k/c = 240 × 10-6
Percent power loss due to
degradation
Gross power loss due to
degradation
37. Effect of Blade Roughness on Turbine
37
Performance
NREL 5 MW turbine, Roughness height k/c = 240 × 10-6
Change in
Annual
Energy
Capture (%)*
Turbine
Capacity
Factor (rough)*
Turbine
Capacity
Factor (clean)*
Mean wind
speed at hub
height (m/s)
5.5 0.194 0.186 -4.26
6.0 0.241 0.231 -3.82
6.5 0.287 0.278 -3.43
7.0 0.334 0.323 -3.08
8.0 0.420 0.409 -2.80
8.5 0.459 0.449 -2.52
* = based on Raleigh distribution
40. 40
Conclusions
• Comprehensive study on effect of blade surface
erosion and soiling on wind turbine performance is
being conducted:
– Field measurements of blade erosion
– Wind tunnel testing (NACA 633-418)
– Computational modeling of surface roughness
• Study is providing significant aerodynamic insight into
surface roughness effects
• Newly developed model allows for specifying
roughness and analyzing impact on airfoil/blade/rotor
performance
• Computational modeling and wind tunnel studies will
be published in two Sandia reports in fall 2014
41. 41
Next Steps
• Near term:
– Implement improvements in computational roughness
model:
• Pressure gradient effect
• Distributed roughness density effect
– Calibrate/validate computational roughness model against
Texas A&M wind tunnel results
• Longer term:
– Evaluate (experimentally and computationally) roughness
sensitivity of higher t/c and higher L/D section shapes
– 3D RANS modeling of roughness effect on rotor
performance
– Implement boundary modifiers (VGs) in RANS and study
their effectiveness mitigating surface roughness effects
– Develop lower-order tool to evaluate surface roughness
effects and optimize boundary layer modifiers (size, location)
42. 42
Acknowledgements
• U.S. Department of Energy
• Sandia National Laboratories
• Warren and Leta Giedt Endowment
• National Science Foundation GK-12 RESOURCE
program