The Determination of Stochastic

Loads on Horizontal Axis Wind

L. N. Freeman Turbine Blades
R. E. Wilson The FAST Code which is capable of determining structural loads of a flexible, teeter-
ing, horizontal axis wind turbine is described and comparisons of calculated loads
Department ot Mectianical Engineering, with test data are given at two wind speeds for the ESI-80. The FAST Code models
Oregon State University, a two-bladed HAWT with degrees-of-freedom for blade bending, teeter, drive train
Corvallis, OR 97331 flexibility, yaw, and windwise and crosswind tower motion. The code allows blade
dimensions, stiffnesses, and weights to differ and the code models tower shadow,
wind shear, and turbulence. Additionally, dynamic stall is included as are delta-3
and an underslung rotor. Load comparisons are made with ESI-80 test data in the
form of power spectral density, rainflow counting, occurrence histograms, and azi-
muth averaged bin plots. It is concluded that agreement between the FAST Code and
test results is good.

Introduction A disadvantage of the Veers model is that it is computationally

expensive, both in computer time and computer memory. Win-
Wind turbines are expensive to build and the capital cost
kelaar (1991) suggested a faster method for the decomposition
must be amortized over the life of the machine. Thus, costs will
of the spectral matrix. Studies using turbulence simulation for
be reduced by longer machine life, which is primarily limited
wind turbine load analysis codes has been done by Powles and
by fatigue damage. Current wind machines undergo 10* to 10'
Anderson (1984), Holley (1985), Garrad and Hassan (1986),
rotations of the low-speed shaft during a 20 year period of
Madsen (1986), Homicz (1987, 1988), Wright and Butterfield
service, while the high-speed shaft will undergo from 10 to 100
(1992), and Malcolm and Wright (1994).
times as many rotations. Thus, the determination of cyclic loads
is fundamental in making wind energy economical. Stochastic loads caused by wind turbulence are presently
universally accepted as a major consideration in designing for
The earliest documented work on wind turbine cyclic loads
long-life cost effective wind turbines. The rotor blades are of
was done on the Smith-Putnam machine and the subsequent
particular concern, not only because the blades are subject to
analyses of the War Production Board (Putnam, 1948). The
the turbulence, but also because the blades are one of the unique
cyclic loads investigated under the Smith-Putnam project were
items on the wind turbine which are usually designed especially
the deterministic loads caused by wind shear, yaw, gravity, and
for a particular wind turbine rather than being a catalog or
inertia.
modified off-the-shelf component. Extreme loads as evidenced
During the resurgence of activity in wind energy in the early by the work of Kelley (1993) and of Sutheriand (1993) are
1970s, cyclic load interest was focused on the above mentioned also believed to be of stochastic origin.
deterministic loads and cyclic loading due to tower shadow was
added. The role of wind turbulence was treated, if at all, by Since cyclic loads are the key to long machine life, an esti-
consideration of a discrete gust. mate of the stochastic loads is an integral part of the design
process. Along with several U.S. teetered rotors currently under
Despite numerous wind turbine failures, the role of turbulence
development are a similar number of structural dynamics codes
as a design driver was not recognized in the 1970s and the
which are intended for the design, design analysis, and analysis
subject of stochastic loads for wind turbines remained unre-
of teetered rotors. Additionally, there are a number of codes,
ported until the early 1980s. Discussion of turbulence-induced
both U.S. and European, which have already been developed,
cyclic loads was given at the Wind Turbine Dynamics Work-
including the works of Thresher and Hershberg (1985), Wright
shop in Cleveland in February 1981. Sundar and Sullivan
and Butterfield (1992), and Lindenburg (1993). Garrad and
(1981) of Purdue University reported on a turbulence simula-
Hassan (1986) outlined procedures for the determination of
tion of the power output of various sizes of wind turbines,
fatigue loads and their firm continues to play a significant role
while Thresher, Holley, and Lin (1981) developed a simplified
in the determination and analysis of stochastic loads. The key
treatment of the turbulence. The so-called Holley model, to-
ingredients for determination of stochastic loads are a structural
gether with linearized aerodynamics, was used by Holley,
dynamics model, a model for aerodynamic loads which are
Thresher, and Jafarey (1981) to determine the wind response
coupled to the structure motion, and turbulence input.
characteristics of horizontal-axis wind turbines.
Another recent approach has been the use of a specialized
Analytical description of the turbulence experienced by the
software package, ADAMS, to model the behavior of wind
wind turbine started with Rosenbrock (1955) who used a simple
turbines under stochastic loads (Malcolm and Wright, 1994).
argument to show that a rotating blade would experience higher
Although this method has seen limited success, it requires many
frequency turbulence than a nonrotating blade. Rotational turbu-
degrees-of-freedom and a significant amount of time to model
lence models were developed by Kristensen and Frandsen
even the simplest flexible body.
(1982), Anderson (1982), and Connell (1981, 1982). Full-
field turbulence models were developed by Veers (1984,1988).
Scope
A current topic of considerable interest related to the improve-
Contributed by the Solar Energy Division of THE AMERICAN SOCIETY OF ME-
CHANICAL ENGINEERS for publication in the ASME JOURNAL OF SOLAR ENERGY
ment of the accuracy and the reduction of time and effort needed
ENGINEERING. Manuscript received by the ASME Solar Energy Division, Jan. to determine stochastic loads is, "how simple or complex must
1998; final revision, Mar. 1996. Associate Technical Editor; P. S. Veers. the structural dynamics model be?" This study compares calcu-

Journal of Solar Energy Engineering Copyright © 1998 by ASIVIE MAY 1998, Vol. 120 / 115

Downloaded From: http://solarenergyengineering.asmedigitalcollection.asme.org/ on 04/12/2018 Terms of Use: http://www.asme.org/about-asme/terms-of-use

 Table 1 ESI-80 turbine specifications Data Code
Mean 26.34 22.78
Rated Power 250 kW StdDev 6.27 6.26
Rated Wind Speed 20.3 ms (45 mph)
Rotor Diameter 24.2 m (80 feet)
Rotor Type Teetered — Underslung
Rotor Orientation Downwind
Blade Construction Wood-Epoxy
Rotor Airfoil NASA LS(1) 04xx o
Tip Speed 77.9 m/s (173 mph)
Cut-in Wind Speed 5.9 m/s (13 mph)
Rotor rpm 60 rpm
Generator Type 300 kW, Induction
Gearbox Planetary, 30:1
Hub Height 24.9 m (81.5 feet)
Bin Magnitude of Flap Moment, kNm
Tower Open — Truss
Fig. 1 Occurrence histogram of blade flap moment at blade root for
Pitch Fixed
ESI-80 machine at 36 mph wind speed
Yaw Passive
Overspeed Control Tip Vanes
Total System Weight 9750 kg (21,500 lb)
Coning 7° flexible bodies. There are 14 degrees-of-freedom in the system.
The model accounts for blade flexibility, tower flexibility, yaw
Natural Frequencies
motion of the nacelle, variations in both rotor and generator
Teeter IHz speed, blade teetering, and blade bending. By selecting various
Tower 1.31Hz physical constants, a variety of different configurations may
First Flapwise 2.05 Hz
Second Flapwise 6.91 Hz be simulated, including generator axis tilt, preconed blades,
Edgewise 7.70 Hz teetering with selected hinge locations, "delta-3" orientation,
various restrictions on the teeter angle, selected drive-train flex-
ibility and damping, and tower flexibility parameters.
lated loads to measured loads for a contemporary lightweight The first four degrees-of-freedom arise from flapwise blade
teetered wind turbine using a structural model that has been motion of each of the two blades. The model allows for full or
incorporated into a computer code, FAST (Fatigue, Aerodynam- partial blade pitch. The blade torsional degree-of-freedom is
ics, Structure, Turbulence). The FAST code is described in Wil- not modeled in this study.
son et al. (1994) and Freeman (1996). Three simplifications The fifth degree-of-freedom accounts for teeter motion of the
have been made in the modeling: First the code was run with two blades about a pin located on the turbine hub. The intersec-
constant rotor rpm; and second, the turbulence was modeled as tion of the blades principal moment of inertia axes can be dis-
the turbulence that exists at a single point on each blade. The placed by the teeter axis by an undersling length. Additionally,
location of the point is at 80 percent of the rotor radius. Third, the model allows for blade precone and a delta-3 angle. A
the second tower modes were not modeled because of the lack lumped hub mass can be included in the code at a specified
of information available to us on the second tower modes of the distance from the teeter pin. Teeter motion can be unrestricted,
ESI-80. The FAST code results have been compared to test data restricted by teeter dampers or teeter springs, or a combination
from a horizontal axis wind turbine, the ESI-80. of both.
The ESI-80 test results (Musial, 1985) represent a valuable The sixth degree-of-freedom accounts for variations in rotor
data set based on the current existence of both the data tapes speed. This degree-of-freedom can model a motor for startup,
and the original test machine. Additionally, personnel associated a brake for shutdown, an induction generator with slip, or a
with the tests are still active in the wind energy field. The variable-speed generator.
original machine was at the University of Massachusetts during The seventh degree-of-freedom models the drive train flexi-
much of 1992 and 1993 where measurements were made on bility between the generator and the rotor. This flexibility was
the rotor to determine the actual parameters of the test machine modeled using a lumped drive train torsional spring and a
(Bywaters, 1992). By using the ESI-80 test data, the study damper.
relates most closely with ESI-80-like machines. The ESI-80 has The eighth degree-of-freedom accounts for yaw motion of
a significant amount of excitation in the range from six per the nacelle and rotor. Yaw motion can be free or fixed with a
revolution to eight per revolution. Rigid hub wind turbines do torsional yaw spring. A yaw tracking control model can be
not exhibit large excitation energy at high frequencies. implemented with the fixed yaw version. The rotor can be either
upwind or downwind. Aerodynamic nacelle loads are not cur-
Fast Code rently modeled.
The dynamic response of a horizontal-axis wind turbine The ninth and tenth degrees-of-freedom are first mode tower
(HAWT) has been modeled using five rigid bodies and three motions. The ninth and tenth degrees-of-freedom are perpendic-

Table 2 Measured parameters for the ESI-80 test turbine Data Code
4.49 4.11
Channel Description 1.37 1.34

1 Wind Speed @ 31.5 m (120 ft)

2 Wind Direction @ 31.5 m (120 ft)
3 Wind Speed @ 24.5 m (80 ft)
4 wind Direction ® 24.5 m (80 ft)
5 Wind Speed @ 12.2 m (40 ft)
6 Wind Direction @ 12.2 m (40 ft)
7 Rotor Azimuth Position
S Teeter Angle
9 Yaw Angle (£ -2 0 2 4 6 8 10 12
10 Blade Root Flap Bending Bin Magnitude of Flap Moment, kNm
11 Blade Flap Bending @ 60% R
12 Low-Speed Shaft Torque Fig. 2 Occurrence histogram of blade flap moment at 60 percent blade
station for ESI-80 machine at 36 mph wind speed

116 / Vol. 120, MAY 1998 Transactions of the ASME

Downloaded From: http://solarenergyengineering.asmedigitalcollection.asme.org/ on 04/12/2018 Terms of Use: http://www.asme.org/about-asme/terms-of-use

 to obtain the component of the blade force in the direction of
the momentum flux.
The momentum flux through a segment of the rotor disk is
obtained using Glauert's Momentum Equation. Whereas the
I blade force involves the flow relative to the blade, the momen-
I tum flux is determined in an inertial reference frame. The in-
duced velocity appears both explicitly and implicitly in the
momentum flux as well as in the blade force so that the induction
20 30 40
must be salved for using iteration. A significant amount of
Wind Speed, mph computing time is used to determine the local induction at each
time step.
Fig. 3 Mean blade flap moment versus mean wind speed for the ESl-
80 machine at 36 mph wind speed The iteration process neglects the effects of the tangential
component of the induced velocity, as well as the effects of
turbulence. The effects of turbulence are ignored during the
Data Code iteration because it is assumed that turbulence does not have a
Mean 0.24 -0.00 fully developed wake and, therefore, does not contribute sig-
StdDev . 1 « , 1.01 nificantly to the induced velocity. Once the iteration process is
completed, turbulence is used in determining the final aerody-
namic coefficients.
The aerodynamic loads are calculated in the blade deformed
position. The dynamic equations are then formulated using
Kane's method. The resulting nonlinear equations are solved in

- 6 - 4 - 2 0 2 4
Bin Magnitude of Teeter Angle, degrees

Fig. 4 Occurrence tiistogram of teeter angle for ESl-80 machine at 36

mph wind speed
Data

I Code

I Data

C 20 Code 0 1 2 3 4 5 6 7 9 10
Frequency, Hz

60 120 180 240 300 360 Fig. 6 Power spectral density of blade flap moment at blade root for
Azimuth Location, degrees ESl-80 machine at 36 mph wind speed

Fig. 5 Azimuth binning of blade flap moment at blade root for ESl-80
machine at 36 mph wind speed 1E3
4J 1E2
S*-E
ular to each other so that tower whirl can be modeled. The G
Data
eleventh and twelfth degrees-of-freedom are the second mode
^ lEl •"•B«O.

tower motions. The eleventh and twelfth tower degrees-of-free- K Code

8 lEO
dom are in the same direction as the ninth and tenth, respec-
tively. Aerodynamic tower loads are not included. The last two 6- lE-1
degrees-of-freedom, 13 and 14, are edgewise motion of the 0 10 20 30 40 50 60
Bin Magnitude of Flap Moment, kNm
blades.
The aerodynamic loading on the blades is determined using Fig. 7 Rainflow cycle counting for the blade root flap moment for ESl-
modified strip theory with nonlinear lift and drag characteristics. 80 machine at 36 mph wind speed
The aerodynamics is driven by a wind model that consists of a
deterministic portion made up of mean wind, shear, and tower
interference and a stochastic portion consisting of an atmo- Data Code
spheric turbulence model including time varying wind direction. Mean 5.67 5.73
StdDev 3.80 4.59
The major loading on the wind turbine blades is due to the
aerodynamic forces of lift and drag. The local relative wind
speed contains contributions from the local wind, the rigid-body
motion of the blade due to rotation about the drive shaft, teeter Data
and yaw axes, the flexible body motion of the blades and tower,
*
and a contribution due to induction from the wake. The induced Code
velocity is determined using strip theory wherein the local force
on the blades due to lift is equated to the momentum flux. The
blade force is based on the flow relative to the blade and contains l£ -15 -10 -5 0 5 10 15 20 25
the induced velocity explicitly in the velocity squared term and Bin Magnitude of Flap Moment, kNm
also contains the induced velocity implicitly in the lift coeffi- Fig. 8 Occurrence histogram of blade flap moment at blade root for
cient and in the various trigonometric functions that are used ESl-80 machine at 23 mph wind speed

Journal of Solar Energy Engineering MAY 1998, Vol. 1 2 0 / 1 1 7

Downloaded From: http://solarenergyengineering.asmedigitalcollection.asme.org/ on 04/12/2018 Terms of Use: http://www.asme.org/about-asme/terms-of-use

 for the FAST Code matched the turbulence magnitude and spec-
I trum of the test data. Since the test data included only the
longitudinal component of turbulence, only the longitudinal
component of turbulence was included in the comparisons.
The FAST Code was developed at Oregon State University
under contract to the Wind Technology Branch of the National
Renewable Energy Laboratory (NREL) (Wilson et al, 1994).
60 120 180 240 300
•* Azimuth Location, degrees The ESI-80 Wind Turbine
m The ESI-80 wind turbine was tested extensively (Musial et
Fig. 9 Azimuth binning of blade flap moment at blade root for ESI-80 al., 1985) and has been selected to compare calculated results
machine at 23 mph wind speed from the FAST Code to field data. The wind turbine, which has
two 40-foot (12.19 m) teetering blades, is a fixed pitch, free
yaw, downwind machine with wood epoxy composite blades.
The rotor blades employ the NASA LS (1) airfoil section. The
specifications and calculated frequencies for the ESI-80 when
operating are summarized in Table 1.

Data
Field Measurements
The ESI-80 test turbine was located in the Altamont Pass
Code
near Tracy, CA. A 120 ft (37 m) meteorological tower was
located 160 ft (50 m) to the west of the wind turbine in the
prevailing wind direction.
Table 2 lists the items that were measured during the test
program and subsequently digitized at 50 Hz by the Solar En-
3 4 5 6 7 10 ergy Research Institute (now NREL).
Frequency, Hz
FAST Results
Fig. 10 Power spectral density of blade flap moment at blade root for
ESI-80 machine at 23 mph wind speed Turbulence induced loads on the ESI-80 were examined using
ten-minute records of wind conditions and loads measurements
as reported by Wright and Butterfield (1992). The mean wind
IE3 speed for Case 1 was 36.14 mph and turbulence intensity was
12.1 percent. For Case 2, the mean wind speed was 22.6 mph
« 1E2 and the turbulence intensity was 9.7 percent.
s
c
s lEl r^. The code developed by Veers (Veers, 1984) was used for
turbulent wind simulation. This code simulates the longitudinal
81
S5 lEO component of the turbulence perpendicular to the rotor disk in
'•i nonyawed flow. A full three-component field of turbulence was
G lE-1
0 5 10 15 20 25 30 35
not calculated.
Bin Magnitude of Flap Moment, IcNm The simulation method determines the "rotationally sam-
pled" wind speed, although nonrotating wind speed can also
Fig. 11 Rainflow cycle count of blade flap moment at blade root for be obtained from the model with minor modifications. The ap-
ESI-80 machine at 23 mph wind speed proach of this method is to simulate wind speed time series in
a plane perpendicular to the mean wind direction and to propa-
gate the time series in the mean wind direction at the mean wind
speed. These signals are then rotationally sampled to prepare an
input time series for the FAST Code.
In order to facilitate the calculation of blade loads, the FAST
Data
Code was run at constant rotor angular velocity. Further, the
Code tower motion was limited to the first tower mode. Thus, ten
degrees-of-freedom were employed; six degrees-of-freedom for
the blade, teeter, yaw, and tower motion in two directions.
0 1 2 3 4 5 6 7 Data on the configuration of the ESI-80 used for the tests was
Bin Magnitude of Flap Moment, kNm

Fig. 12 Rainflow cycle count of blade flap moment at 60 percent blade

station for ESI-80 machine at 23 mph wind speed 100

the time domain using a predictor-corrector method. Tower and 10

blade loads are determined by integration along the blade.
Code
Turbulence in the wind was accounted for by use of a turbu-
lence model, the Sandia Three-Dimensional Wind Simulation
(Veers, 1984). This gives a rotationally sampled longitudinal
turbulence component for each blade at one point on the blade. 0.1
Each value represents the change in wind velocity due to turbu- 10 20 30 40 50 60
lence. These values are superimposed on the steady component Bin Magnitude of Edgewise Moment, kNm
of the wind which already includes the effects of tower shadow Fig. 13 Calculated rainflow cycle count for the blade root edgewise
and wind shear. The turbulence magnitude and spectrum used bending moment at 36 mph

118 / Vol. 120, MAY 1998 Transactions of the ASME

Downloaded From: http://solarenergyengineering.asmedigitalcollection.asme.org/ on 04/12/2018 Terms of Use: http://www.asme.org/about-asme/terms-of-use

 0.50
• 23 mph 36 mph
,0.40 ' : Mean 6.01 10.83
/ . StdDev 0.90 1.90
# - Minimum 1.42 2.78
i>0.30 r • Maximum 9.12 20.78
O

g0.20

•0.10

0.00
10 15 20 25
Bin Magnitude of Angle of Attack, degr

Fig. 14 Calculated frequency distribution of the angle-of-attack near the blade tip for the
ESI .80

facilitated by measurements made at the University of Massa- rence histogram similar to the distribution that would be ob-
chusetts. Of particular note is the presence of both teeter springs tained from a harmonic oscillator. Further improvement to the
and teeter dampers. shape of the teeter histogram was obtained by allowing the rotor
speed to vary.
Comparisons, 36.1 M P H Azimuth averaged load plots are the second form of compari-
son between test data and FAST calculations. Figure 5 shows
Histograms of test data and code calculations are shown be- the azimuth binned blade root flatwise bending moment at 36.1
low. Figure 1 shows a histogram for the 36.1 mph case for the mph. Note that the load scale covers the range from 10 to 40
blade root flatwise bending moment. Agreement between test kNm. Agreement between FAST calculations is good as all
data and code is good with a similar shape to both distributions. fluctuations shown by the data are present in the calculations.
The test data mean was 26.34 kNm, while the FAST Code The magnitude of the calculated moment between 90 deg and
mean was 3.6 kNm lower. Figure 2 shows the flapwise bending 135 deg (post tower shadow region) and between 270 deg and
moment histogram at a station 60 percent of the rotor radius. 315 deg has a maximum difference of 9 kNm below the test
Again the data is higher than the code results, the mean for the data.
data being 4.49 kNm and the mean for the code was 0.4 kNm
lower. Since the mean acceleration of the blade in the flatwise Power Spectral Density of the root flatwise bending moment
direction is zero, the difference between test data and code is shown in Fig. 6 for a wind speed of 36.1 mph. Agreement
must be from the mean aerodynamic loads, the mean centrifugal between code and test data is good including the broadening in
loads, or due to the data. Calculation of the mean blade root the region of 2 Hertz. The code failed to predict the broad
bending moment and comparison to test data shown in Fig. 3 plateau between 2 and 3 Hertz that appears in the test until the
suggest that the calibration of the strain gages drift from test edgewise degrees-of-freedom were incorporated into the code.
run to test run so that the code results shown in Figs. 1 and 2 Rainflow cycle counting is shown in Fig. 7 for the 36.1 mph
are felt to be within the range of experiment test error. case. Agreement between FAST calculations and test data is
good over the entire range. Code calculations shown in Figs. 1
Figure 4 shows the teeter occurrence histogram at 36.1 mph. through 7 were made without dynamic stall. Calculations made
Several items may be mentioned concerning the data. First, the with dynamic stall produced similar results to those produced
mean teeter angle from the test data is not zero being 0.24 deg. without dynamic stall except for the low magnitude cycles.
Second, the effects of the teeter springs/dampers can be seen
in the data; the plateau above +2 deg and a similar plateau
at about - 1 deg. While the FAST Code results also exhibit Comparisons, 22.6 MPH
"plateaus" in the region of ±2 deg, the code has a mean teeter A histogram of the blade root flatwise bending moment is
angle of zero and the calculations are more or less symmetrical shown in Fig. 8. Agreement between FAST2 calculation is very
about the origin. Third, the shape of the teeter angle distribution good as the mean, standard deviation, and distribution are all
was found to be a result of including the yaw degree-of-free- very close. The azimuth averaged flatwise blade bending mo-
dom. McCoy (1992) had modeled the ESI-80 using a code ment shown in Fig. 9 also shows good agreement between test
without a yaw degree-of-freedom and obtained a teeter occur-

140
0.016
Mean
Code
119
y^'^ 130
0.012 StdDev 24
V
8 Min 31 • * HJ
120
0.008 Max 200
• .&
o Code
%
^ 110
0.004

0.000 y 1
< 100
0 50 100 150 200 1 60 120 180 240
Azimuth Location, degrees
300 360
Bin Magnitude of Tip Location, cm
Fig. 15 Calculated frequency distribution of the blade tip flatwise de- Fig. 16 Calculated azimuth-binned blade tip flatwise deflection of the
flection of the ESI-80 operating at 36 mph ESI-80 operating at 36 mph

Journal of Solar Energy Engineering MAY 1998, Vol. 1 2 0 / 1 1 9

Downloaded From: http://solarenergyengineering.asmedigitalcollection.asme.org/ on 04/12/2018 Terms of Use: http://www.asme.org/about-asme/terms-of-use

 data and calculations in magnitude, phase angle, and representa- Bywaters, G., 1992, personal communication, Oct.
Connell, J. R., 1981, "The Spectrum of Wind Speed Fluctuations Encountered
tion of major fluctuations. The power spectral density of the by a Rotating Blade of a WECS," PNL 4083, Pacific Northwest Laboratories,
root flap moment is illustrated in Fig. 10. While agreement Richland, WA.
between the test data and code is good, there appears to be a Connell, J. R., 1982, "The Spectrum of Wind Speed Fluctuations Encountered
scale shift in the frequency, the data peaks occurring at slightly by a Rotating Blade of a WECS," Solar Energy, Vol. 29.
lower than integer values of the rotor angular velocity while Freeman, L. N., 1996, "Stochastic Loads on Horizontal Axis Wind Turbine
Blades," Ph.D. thesis, Oregon State University, CorvalUs, OR, May.
the code peaks occur at values slightly above integer values of Garrad, A. D., and Hassan, U., 1986, "The Dynamic Response of Wind Tur-
the rotor angular velocity. With the rotor angular velocity of bines for Fatigue Life and Extreme Load Prediction," European Wind Energy
1.005 Hertz, the differences are believed to be associated with Conference (EWEC).
the digitization of the data from the analog tape (Wright, 1991). HoUey, W. E., 1985, "An Atmospheric Turbulence Model for Wind Turbine
Dynamic Simulation; Its Verification and Implementation into the FLAP Code,''
Rainflow cycle counting is shown in Fig. 11 where root flap Oregon State University, unpublished OSU report.
cyclic moment count is shown at 22.6 mph. The results shown Holley, W. E., Thresher, R. W., and Jafarey, N., 1981, "Wind Response Char-
in this figure show as good agreement with the test data as the acteristics of HAWTs," Proceedings, Wind Turbine Dynamics Workshop, Cleve-
previous figures. The 60 percent blade station cyclic count is land, OH, Oregon State University, CorvalUs, OR.
shown in Fig. 12. Agreement between the code and test data is Homicz, G. F., 1987, "VAWT Aerodynamic Modeling with Stochastic
Winds," 7th Annual VAWT Aerodynamics Seminar, Bushland, TX, Sandia Na-
again good. tional Laboratory, Albuquerque, NM.
Homicz, G. F., 1988, "VAWT Stochastic Loads Using a Full 3-D Simulation,"
8th Annual VAWT Aerodynamics Seminar, Bushland, TX.
Other Output Kelley, N. D., 1993, "Full Vector (3-D) Inflow Simulation in Natural and Wind
Farm Environments Using an Expanded Version of the SNLWIND (VEERS)
In addition to the quantities previously illustrated, there are Turbulence Code," Wind Energy~1993, S. M. Hock, ed., ASME SED-Vol. 14,
a number of variables of interest for which ESI-80 test data is ASME, New York.
not available. Paramount of these quantities is the blade edge- Kristensen, L., and Frandsen, S., 1982, "Model for Power Spectra of the Blade
wise bending moment. Figure 13 shows the blade root edgewise of a Wind Turbine Measured from the Moving Frame of Reference," / of Wind
Engr. and Ind. Aero., Vol. 10.
bending moment at 36.1 mph. Shown in Fig. 13 are the rainflow Lindenburg, C. and Snel, H., 1993, "PHATAS-11 Program for Horizontal Axis
cycle count for fixed speed operation. The code was run using Wind Turbine Analysis and Simulation Version 11," Wind Energy—1993, S. M.
both fixed and variable speed operation and the difference in Hock, ed., ASME SED-Vol. 14, ASME, New York.
calculated loads was found to be minor. The rainflow cycle Madsen, P. H., 1986, "DAP-1, A Data Analysis Package for Spectral Analysis,
Version 1.2," Solar Energy Research Institute, Golden, CO, SERI internal unpub-
count shows the characteristic behavior of a bi-modal distribu- lished report.
tion, the large number of low-amplitude cycles being due to the Malcolm, D. J., and Wright, A. D., 1994, "The Use of ADAMS to Model the
gravity loads that occur once per rotor revolution. Figure 14 AWT-26 Prototype," Wind Energy—1994, W. D. Musial, ed., ASME SED-Vol.
shows the distribution of the calculated angle-of-attack near the 15, ASME, New York.
blade tip for both wind speeds. While comparison data is not McCoy, T. J., January 1992, ' 'Development of Stochastic Loads for the Design
of an Advanced Wind Turbine," Eleventh ASME Wind Energy Symposium Pro-
available, such plots may be useful in determining the magni- ceedings, P. S. Veers and S, M. Hock, eds., ASME, New York.
tude and the frequency of large angle-of-attack excursions. Musial, W. D., Butterfleld, C. P., and Handman, D., 1985, "ESI-80/EPRI Test
Finally, Figs. 15 and 16 show the calculated blade tip deflec- Program," EPRI Research Project RP1996-A, final draft report.
Powles, S. J. R., and Anderson, M. B., 1984, The Effect of Stochastic and
tion, including the effects of blade teeter and elastic flatwise Deterministic Loading on Fatigue Damage in a Large HAWT, Sir Robert McAl-
bending, at a wind speed of 36.1 mph. Again, while test data pine & Sons Ltd., London.
is not available for the tip deflection, such calculations would Putnam, P. C , 1948, Final Report on the Smith-Pumam Wind Turbine, Engi-
be of use to a wind turbine designer. Figure 15 shows the neering Research Division, New York University, War Production Board.
Rosenbrock, H. H., 1955, "Vibration and Stability Problems in Large Wind
occurrence distribution of the tip deflection while Fig. 16 shows Turbines Having Hinged Blades," Electrical Research Association Report, C/T
the azimuth-binned distribution of tip deflection. 113, London.
Sundar, R. M., and SulUvan, J. P., 1981, "Performance of Wind Turbines in a
Turbulent Atmosphere," Proceedings, Wind Trubine Dynamics, NASA CP 2185,
Conclusions Purdue University, West Lafayette, IN.
Sutherland, H. J., 1993, "Effect of Flap and Edgewise Bending Moment Phase
Comparison of ESI-80 test results for the rootflapwisebend- Relationships on the Fatigue Loads of a Typical HAWT Blade," Wind Energy—
ing moment have been made with calculations from the FAST 1993, edited by S. M. Hock, ed., ASME SED-Vol. 14, ASME, New York.
Code. The comparisons have been made at mean wind speeds Thresher, R. W., and Hershberg, E. L., 1985, "Development of an Analytical
of 22.6 and 36.1 mph and cover occurrence histograms, azimuth Model and Code for the Flapping Response of a HAWT Rotor Blade," SERI/
STR-217-2629, Oregon State University, CorvalUs, OR.
averaged bin plots, power spectral density distributions, and Thresher, R. W., Holley, W. E., and Lin, S. R., 1981, "The Response Sensitiv-
rainflow counting. Based on the results shown in Figs. 1 through ity of Wind Turbines to Atmospheric Turbulence," Proceedings, Wind Turbine
12 it is our opinion that the FAST Code is capable of good Dynamics Workshop, Cleveland, OH, Oregon State University, CorvalUs, OR.
accuracy in the determination of stochastic blade bending loads Veers, P. S., 1984, "Modeling Stochastic Wind Loads on Vertical Axis Wind
Turbines," SAND83-1909, Sandia National Laboratory, Albuquerque, NM.
on the ESI-80 wind turbine. Calculations have been made using Veers, P. S., 1988, "Three-Dimensional Wind Simulation," SAND88-0152,
bothfixedand variable rotor angular velocity and it is concluded Sandia National Laboratory, Albuquerque, NM.
that the blade flatwise loads and teeter motion are adequately Wilson, R. E., Walker, S. N., and Freeman, L. N., July 1994, "Advanced Dy-
determined for the ESI-80 using a constant rotor angular veloc- namics Code Development," OSU/NREL Report 94-01, Phase III Annual Techni-
cal Progress Report, Mechanical Engineering Department, Oregon State Univer-
ity. sity, CorvaUis, OR.
Winkelaar, D., 1991, "Fast Three-Dimensional Wind Simulation and the Pre-
diction of Stochastic Blade Loads," Tenth ASME Wind Energy Symposium,
References Houston, TX.
Anderson, M. B., 1982, "The Interaction of Turbulence with a HAWT," Pro- Wright, A. D., 1991, Digitized ESI-80 Test Data, (unpublished), SERI, Aug.
ceedings, Fourth British Wind Energy Association (BWEA) Wind Energy Confer- Wright, A. D., and Butterfield, C. P., 1992, "The NREL Teetering Hub Rotor
ence, Cavendish Laboratory, University of Cambridge, Cambridge, UK, pp. 104- Code: Final Test Results and Conclusion,'' Eleventh ASME Wind Energy Sympo-
118. sium Proceedings, P. S. Veers and S. M. Hock, eds. ASME, New York.

120 / Vol. 120, MAY 1998 Transactions of the ASME

Downloaded From: http://solarenergyengineering.asmedigitalcollection.asme.org/ on 04/12/2018 Terms of Use: http://www.asme.org/about-asme/terms-of-use