Loads Measurement and Validation of a 2.1-kW HAWT on a Fiberglass Composite TowerScott Dana, Rick Damiani, Jeroen van Dam
Small Wind Conference 2016
June 13th – June 15th
2
Motivation
Ongoing efforts to improve system dynamic modeling and prediction
Assess common industry practice loads derivation methods
Evaluate IEC 61400-2 Simplified Loads Approach
Address fatigue loads more comprehensively
3
Design Load Derivation Methods
1. Full-Scale Loads Measurements (Field)
• A 2.1-kW, free yaw, variable-speed, downwind machine
• An AnemErgonics, LLC 18-m multisection fiberglass composite tower
• Meteorological data recorded at hub height
• Tower strain measured at 0.8 m above tower hinge plate
• Full bending bridges in two orthogonal directions
2. Aeroelastic Modeling (FAST)
• FAST v 7.02 utilized
• Turbulent wind files, 3 seeds, with reference wind speeds of 2, 4…,24, 26 m/s
• Turbulence intensity of 18%
• Air density of 1.0 kg/m3 to match site conditions
• Tower drag not modeled
3. Simplified Loads Approach (SLA)
• IEC 61400-2 design load cases (DLCs) A (fatigue), H, and I (ultimate loads)
• Input parameters derived from the FAST model, preliminary test results, and turbine dimensions
• SLA load components used to determine the tower-base bending moment
4
Full-Scale Loads Measurements Approach
X (approximate prevailing wind direction)
Y
285°
15° (from true North)
105°
195°
X Bending Bridge
Y Bending Bridge
My
Mx
Tower-Base Bending
Bridge Location
Primary
Anemometer
Reference
Anemometer
Temperature
and Pressure
Wind Vane
• Measurement sector of 223° to 333 ° of true north following IEC 61400-12-1
• Measurement methods and loads processing following IEC 61400-13
• Coordinate transformation applied to loads
Test article used for full-scale loads measurement with identification of tower strain gage locations
(Photo by Rick Damiani, NREL)Met tower and instrumentation used for field test campaign (Photo by Scott Dana, NREL)
Tower Strain full bridge orientation and loads configuration
5
Measurement Data Acquisition System
DAS Enclosure
• NI cDAQ Chassis and Input Modules
• Sampled at 35 kHz• Stored at 100 Hz
• 1,332 10-Minute data files within measurement sector for Loads Calculations
• 3,482 10-Minute data files for Fatigue Calculations (all operating data)
Communication one-line for tower loads DAS
6
Simplified Loads Approach – Maximum Bending Moments
0
5
10
15
20
25
30
35
40
45
Load Case H Load Case IB
en
din
g M
om
en
t [k
Nm
]
Rotor Drag
Nacelle Drag
Tower Drag
Component-drag contributions to the total combined tower-base bending moment for DLCs H and I
SLA DLCs for tower analysis:
• IEC 61400-2 DLC H
• Extreme wind speed of 59.5 m/s• A normally parked turbine
• IEC 61400-2 DLC I
• 42.5 m/s wind speed• Maximum exposure (yaw mechanism
failure)• A ‘no-yaw-error’ configuration in
this case
80 kNm
47 kNm
• Total Moment = 80 kNm
• Total Moment = 47 kNm
DLC H develops the greatest bending moment Tower Design Driver
7
FAST and Field Bending Moments Results
0 5 10 15 20 25 30-10
-5
0
5
10
15
20
25
30
35
40
Bendin
g M
om
ent
[kN
-m]
0 5 10 15 20 25 30-25
-20
-15
-10
-5
0
5
10
15
Wind Speed [m/s]
Bendin
g M
om
ent
[kN
-m]
Loads extrapolation of the field-measured loads is required for a proper comparison
with SLA results — a future work item of this study.
• Deviation between Field and FAST is likelya result of the actual turbine controlsystem mitigating loads at high windspeeds (18 m/s and above)
Ten-minute statistical comparison of the Field and FAST tower-base ultimate loads in the fore-aft (top) and side-to-side (bottom) directions
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 10
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1
Wind Speed [m/s]
Mom
ent
[Nm
]
Perpendicular to Wind Moments, Mx: 10 Minute Statistics
Field Max
Field Mean
Field Min
FAST Max
FAST Mean
FAST Min
• For the captured wind speeds FASTand field loads are within the loadenvelope of DLCs H and I
• Less than 47 kNm (DLC I)
8
Tower-base 1-Hz lifetime (20 years); DELs at zero mean for Field, FAST, and SLA with S-N curves with inverse exponent values (m) of 3, 6, and 10
Fatigue Loads Comparison – Damage Equivalent Loads
0
2
4
6
8
10
12
14
Field Fore-Aft FAST Fore-Aft SLA Fore-Aft
DEL
Be
nd
ing
Mo
me
nt
[kN
m]
m = 3 m = 6 m = 10
m = 3
m = 10
Cycles
Load
Ran
ge
NSLANeqNFAST
SSLA
SFAST
Illustration of the effect of a chosen S-N curve slope on the DEL for a high load range (SFAST) at a low number of cycles (NFAST) and vice versa (SSLA, NSLA)
• Three different S-N curves
• Rayleigh distribution of wind speed
• 20-year design life
• SLA DELs by IEC 61400-2 DLC A
• Peak-to-peak thrust load
• 20-yr # of cycles @ rated rotor speed
Significant effect of material properties S-N curve exponent!
9
Fatigue Loads Comparison Continued
0 5 10 15 20 25 300
5
10
15
20
25
30
10 m
in D
EL [
kN
m]
Wind Speed [m/s]
Fore-aft Field
Fore-aft FAST
104
106
108
1010
1012
1014
0
5
10
15
20
25
30
35
40
Load R
ange [
kN
m]
Cumulative Cycles
Fore-aft Field
Fore-aft FAST
SLA Value
Load range = 5.183 kNm
Cycle Count = 10.4e9
Short-term (10 minute) fore-aft tower-base bending moment DELs versus the mean wind speed for an S-N curve with m = 10
Fore-aft tower-base bending moment cumulative fatigue spectra comparison of field-measured loads and FAST predictions
• The SLA DLC A yields a single fatigue load range value of 5.183 kNm, with a cycle count of 10.4e9.
• Although similar, FAST’s over-prediction of high-range loads compared to the Field spectra are demonstrated.
• This outcome is consistent with the ultimate loads comparison.
10
This study focused on tower-base bending moments of a horizontal-axis windturbine to provide an indication of the overall performance of the loadderivation methods. Notable outcomes include the following:
• The SLA DELs are different than those calculated from the Field and FASTdata;
o High dependence on the assumed S-N curve slope and the number of cycles.
o SLA DEL more conservative for steel (m = 3 to 6) and less conservative forcomposites (m =10).
• Better agreement between the field and FAST predictions for side-to-sidethan for fore-aft bending at high wind speeds;
o Likely caused by a lower fidelity turbine control model used in the FASTsimulations.
• The fatigue spectra reveals FAST’s over-prediction of large cycles.
• SLA does not comprehensively address fatigue;
o Conservative and coarse → one load range at one cycle count.
• We recommend that this study be repeated with wind turbines of differentconfigurations and sizes, and that field load extrapolation be conducted.
Final Remarks
11
• The Authors of this work would like to sincerely thank Paul Migliore of AnemErgonics for allowing his tower and turbine to be used for the field measurement campaign of this study.
• The Presenter would like to thank Rick and Jeroen for their excellent mentorship and guidance throughout the project.
Acknowledgements
Thank you!
Questions?