This paper presents validation and code-to-code verification of the latest version of the U.S. Department of Energy, National Renewable Energy Laboratory wind turbine aeroelastic engineering simulation tool, FAST v8. A set of 1141 test cases, for which experimental data from a Siemens 2.3 MW machine have been made available and were in accordance with the International Electrotechnical Commission 61400-13 guidelines, were identified. These conditions were simulated using FAST as well as the Siemens in-house aeroelastic code, BHawC. This paper presents a detailed analysis comparing results from FAST with those from BHawC as well as experimental measurements, using statistics including the means and the standard deviations along with the power spectral densities of select turbine parameters and loads. Results indicate good agreement among the predictions using FAST, BHawC, and experimental measurements. These agreements are discussed in detail in this paper, along with some comments regarding the differences seen in these comparisons relative to the inherent uncertainties in such a model-based analysis.
It is well known that the wind energy industry is growing rapidly worldwide,
and increasingly larger blades are being employed by state-of-the-art wind
turbine technology. With the advent of large, flexible blades, the tools
required to design and analyze these turbines also need to be updated
according to the current and future state of affairs. A wind turbine
aeroelastic simulation code consists of several modules that model different
aspects of wind turbine physics (aerodynamics, structures, pitch/torque
control modules)
International Electrotechnical Commission (IEC) matrix populated with data bins defined by mean wind speed and turbulence intensity. The numbers in the table represent the number of 10 min data sets within each bin, for a total of 1141 cases.
Electrical power.
Rotor speed.
Rotor thrust force.
Blade in-plane tip deflections.
Blade out-of-plane tip deflections.
Blade-root in-plane bending moments.
Blade-root out-of-rotor-plane bending moments.
Main-shaft bending moment – yaw (nonrotating coordinate system).
Main-shaft bending moment – tilt (nonrotating coordinate system).
Tower-bottom side–side bending moments.
Tower-top torsion moment.
PSDs of the electrical power. The three sets of graphs represent the
ensemble-averaged PSDs from the inflow velocity bins at 8 m s
PSDs of the rotor speed. The three sets of graphs represent the
ensemble-averaged PSDs from the inflow velocity bins at 8 m s
PSDs of the rotor thrust force. The three sets of graphs represent
the ensemble-averaged PSDs from the inflow velocity bins at 8 m s
PSDs of the in-plane blade-tip deflection. The three sets of graphs
represent the ensemble-averaged PSDs from the inflow velocity bins at
8 m s
PSDs of the out-of-plane blade-tip deflection. The three sets of
graphs represent the ensemble-averaged PSDs from the inflow velocity bins at
8 m s
Although the availability of models that can capture physics relevant to
state-of-the-art wind turbines is one challenge, the other equally important
challenge is validation of such codes. Historically, there have been several
international industrial as well as academic efforts to carry out
code-to-code verification and validation of wind turbine models against
experiments; most notable, among others, are the MEXICO
Published works on the validation of high-fidelity aeroelastic codes using
larger, flexible turbines, such as
FAST is an open-source multi-physics simulation tool that was created for
design and analysis of advanced land-based and offshore wind technology.
Underpinning FAST is a modularization framework that enables coupling of
various modules, each representing different physics domains of the wind
system
In the newest release of FAST, BeamDyn
PSDs of the blade-root in-plane bending moments. The three sets of
graphs represent the ensemble-averaged PSDs from the inflow velocity bins at
8 m s
PSDs of the blade-root out-of-rotor-plane bending moments. The three
sets of graphs represent the ensemble-averaged PSDs from the inflow velocity
bins at 8 m s
Finally, in the latest FAST v8, AeroDyn has been overhauled to (1) fix
underlying problems with the original theoretical treatments, (2) introduce
improved skewed-wake and unsteady aerodynamics models, (3) enable the modeling of
highly flexible and non-straight blades, and (4) support the unique features
of the FAST modularization framework
The National Wind Technology Center (NWTC) at NREL is home to several
megawatt-scale test turbines. One of these is a Siemens 2.3 MW machine
(SWT-2.3-108) that operates in an upwind configuration and is equipped with a
three-bladed, 108 m diameter rotor
This paper presents a detailed validation and code-to-code verification of
the latest version of the U.S. Department of Energy's open-source wind
turbine aeroelastic simulation tool FAST v8, which is supported by NREL. A
previous paper
The analysis presented in this paper is centered on the Siemens 2.3 MW
turbine with a 108 m rotor that is installed at NREL's NWTC. The turbine is
heavily instrumented, and measurement data have been made available through collaborative research between NREL and Siemens Energy. Data were collected
over a period of several months under normal operating conditions for a range
of wind speeds and turbulence intensities, and these data were used in this
validation exercise. Validation was conducted following the guidelines
stipulated in the IEC wind turbine load-measurement standard
A large amount of test data were recorded on the Siemens 2.3 MW wind turbine
and the NWTC 135 m met tower. The met tower is located approximately 2.5
rotor diameters upstream of the turbine and is instrumented with several
sensors along its length that measure the wind speed, wind direction, and
atmospheric pressure. The inflow wind speed data used in this analysis were
recorded at 80 m, which is close to the turbine hub height. Each 10 min
data set contains surface-pressure measurement data, strain-gage data,
turbine-operation data, and the inflow data from the met tower. In total,
several months of data have been recorded, of which this report presents data
from 1141 10 min data sets. The data samples were selected based on their
mean hub-height wind speed and turbulence intensities in order to populate
the recommended test matrix of the IEC 61400-13 (see Table
Each bin in the test matrix (Table rotor speed electrical power blade-root bending moments (in and out of the rotor plane) main-shaft bending moments (yaw and tilt directions, in a nonrotating coordinate system) tower-top torsional moment tower-bottom bending moments (parallel and perpendicular to the mean wind
direction) blade-tip deflections (in and out of the rotor plane).
For brevity, only select PSDs are presented in this paper at three different
operating conditions for each QOI: one at below-rated, one at rated, and one at
above-rated hub-height inflow wind speed.
BHawC is the Siemens in-house aeroelastic simulation tool used to study the dynamic response of wind turbines. The model consists of substructures for foundation, tower, nacelle, drivetrain, gearbox, hub, and blades. The structure is modeled primarily with finite beam elements, and the aerodynamics is modeled using blade element momentum theory. The code is coupled to a controller identical to that on the real turbine.
The structural model of BHawC employs a co-rotational beam formulation, which is a combined multi-body and linear finite-element representation allowing for geometric nonlinearities through a series of multiple bodies, each composed of a linear finite element. The BHawC model of the SWT-2.3-108 blade used in the current study was initially curved in space and discretized into 16 linear elements. In other parts of the turbine where bearings are present, special elements are introduced and the drivetrain consists purely of torsional elements.
The aerodynamic force in BHawC is calculated at a given number of points
(which varies depending on the turbine) on the blades positioned
independently of the structural nodes. Blade element momentum theory is
applied to determine the tangentially and axially induced velocities at these
aerodynamic calculation points, and Prandtl's tip loss correction as well as
a correction for thrust at high induction values are implemented. The
blade element implementation in BHawC also allows for unsteady and skewed
inflow. The aerodynamic force is based on 3-D-corrected coefficients for
stationary airfoil data, and a Beddoes–Leishman-type model for
unsteady/dynamic events. In addition, BHawC contains a model for tower
shadow, and it also calculates the aerodynamic forces on the nacelle and
tower. For further details on BHawC, see
The FAST model of the SWT-2.3-108 machine was created based on the
information obtained from Siemens as well as the inflow data from the
experimental database for each of the cases simulated. This three-bladed
upwind turbine has a fixed coning angle of
Details of the FAST v8 model have been described in a previous paper
In the BeamDyn, it is possible to model a blade defined by many
cross-sectional property stations with relatively few node points for
integration while capturing all of the provided material properties
The two-dimensional airfoil aerodynamic properties for the blade were
obtained from Siemens. The blade was modeled in AeroDyn using 20 nodes. The
lift-coefficient data were processed to include rotational augmentation
effects in the same way as BHawC, according to a variation of the
The measured yaw error in most test cases was very small, with a few
exceptions of yaw errors higher than
FAST used the same turbulent inflow input files as BHawC, which used HawC-style turbulence boxes. For each case simulated, the turbulence box is scaled according to the mean wind speed and the turbulence intensity at the turbine hub height, obtained from the 80 m wind speed and direction measurement from the met tower.
A controller in a bladed-style dynamic-link library (DLL) form obtained from Siemens was used for pitch and torque control in the FAST simulations. It is known that this DLL controller does not employ the exact same control logic as the physical turbine or the BHawC simulations, but it is similar.
The following are improvements to the current FAST model over the model used
in the previous study In the FAST simulations in this paper, wind shear was modeled assuming
a power-law profile. In the previous paper, a power-law exponent of 0.2 was
assumed for all 1141 simulations, whereas in this work, the shear exponent
was derived using experimental measurements. The met tower collected wind
speed and direction information at the heights of 3, 10, 26, 80, 88, and
134 m. These data were used to estimate a power-law exponent by the
least-squares best-fit method. For the few cases where a best-fit
approximation was deemed unsuitable (e.g., low-level jet boundary layer
profile), a power-law coefficient of 0.2 was assumed. The BHawC simulations
used atmospheric shear estimates based on lidar measurements, and so there
may be small differences among the inflow conditions of the test turbine,
FAST simulations, and the BHawC simulations, adding to the uncertainty in
this exercise. It was found that the rotor mass imbalance was modeled too high
in the previous paper and has been corrected in the current model. A part of the experimental measurements of the main-shaft bending
moments was found to contain corrupt data, and those data have been filtered out. A controller in a bladed-style DLL form obtained from Siemens was used
for pitch and torque control in the FAST simulations. In the previous study,
it was observed that the FAST simulations with the Siemens DLL controller
gave rise to some resonance effects in the coupled drivetrain and tower
side–side modes, which in the previous study were suppressed by artificially
increasing the drivetrain damping. It was later found that the FAST-DLL
controller interface requires a low-pass filter to filter out high-frequency
content on the generator speed in order for the FAST model to represent the
control mechanism as close to the BHawC model/experimental turbine as possible.
This filter was added to FAST in the simulations shown in this paper, resolving
the resonance issue highlighted earlier. The drivetrain properties used in the
current simulations are those that were prescribed by Siemens (i.e., the damping
was reduced to its original value in the new results); no further artificial changes have been made. The previous study did not include unsteady aerodynamics effects,
which are included in this work
The simulations shown in this work were generated using FAST v8.15.01a-bjj compiled in double precision with
ElastoDyn v1.03.02a-bjj, BeamDyn v1.01.01, AeroDyn v15.00.01a-bjj, InflowWind v3.02.00a-adp, and ServoDyn v1.04.00a-bjj.
The FAST simulations with BeamDyn used a time increment of 0.0005 s, whereas those with ElastoDyn used 0.005 s. All FAST simulations were carried out for 12 min, where the first 2 min were ignored to remove initial-transient effects. BHawC simulations were 11 min long with a 0.02 s integration time increment, and the first minute of transience was ignored. Note that the structural blade in ElastoDyn is straight and only includes bending DOFs, whereas the structural blade in BeamDyn is curved (as is the blade model in BHawC) and includes the DOFs bending, torsion, shear, and extension, with composite coupling terms.
The plots presented herein show the mean and the standard deviation values of
several QOIs for each of the 1141 10 min sample cases simulated, along with
the PSDs of select bins at below-rated, at-rated, and above-rated operating
conditions. The
In the previous paper, each 10 min test case was simulated using three
different FAST configurations, to be able to analyze the relative
improvements in FAST among the capabilities in its previous public releases
(FAST v8.10 and earlier) and the current developments with BeamDyn and
AeroDyn. One observation from the previous study has been that the effect of
the aerodynamic blade curvature on the FAST results is negligible. Therefore,
this study presents comparisons of FAST simulations with straight ElastoDyn
blades and with curved BeamDyn blades, where both used curved aerodynamic
blades. In the figures that follow, the different data identified in the
legends are as follows:
Data – experimental measurements BHawC – results from Siemens' BHawC simulations FAST (ED) – FAST simulations with ElastoDyn FAST (BD) – FAST simulations with the new BeamDyn.
Note that ElastoDyn and BeamDyn designate structural modules used to
model the blades only; the structural dynamics of the remainder of the
turbine (drivetrain, nacelle, and tower) were always modeled in ElastoDyn.
For Figs.
Figures
The current results consist of analyses in the time and frequency domains. The results indicate that the predictions from the latest version of FAST compare consistently well with those of BHawC as well as the experimental data. A comparison between the results from various FAST simulations indicate an improvement of the FAST simulations using BeamDyn over those using ElastoDyn.
The BHawC data shown in this paper are the same as those shown in
The agreement among the results from the improved FAST model, BHawC, and experimental measurements remains very good. Below is a discussion regarding the observations from the comparisons presented in the figures and improvements to the results published in the previous study as a result of the newer FAST model.
Figures
Figure
Figures
Figures
The main-shaft yaw and tilt bending moments shown in Figs.
Figure
Figure
Figures
Figures
Similarly, peaks A and D in Fig.
PSDs of the rotor thrust force plotted in Fig.
The tools FAST (BD), FAST (ED), and BHawC ranked according to how well their results compare to the experimental measurements in the mean and standard deviation for each QOI. No value indicates that no discernible difference could be seen among the different tools for that QOI.
The tools FAST (BD), FAST (ED), and BHawC ranked according to how well their PSD results compare to the experimental measurements for each QOI. No value indicates that no discernible difference could be seen among the different tools for that QOI.
The PSDs of the blade-tip deflections, shown in Figs.
The main difference between the two blade models used in the FAST
simulations, ElastoDyn and BeamDyn, can most clearly be observed in the
blade-root bending moments. The PSD of the blade-root in-plane bending
moments shown in Fig.
The PSDs of the blade-root out-of-plane bending moments are shown in
Fig.
Figures
The PSDs of the tower-bottom side–side bending moments plotted in
Fig.
PSDs of the main-shaft bending moment – yaw (nonrotating coordinate
system). The three sets of graphs represent the ensemble-averaged PSDs from
the inflow velocity bins at 8 m s
PSDs of the main-shaft bending moment – tilt (nonrotating coordinate
system). The three sets of graphs represent the ensemble-averaged PSDs from
the inflow velocity bins at 8 m s
PSDs of the tower-bottom side–side bending moments. The three sets
of graphs represent the ensemble-averaged PSDs from the inflow velocity bins
at 8 m s
PSDs of the tower-top torsion moment. The three sets of graphs
represent the ensemble-averaged PSDs from the inflow velocity bins at
8 m s
In this paper, the latest results from FAST v8 with the new BeamDyn and AeroDyn modules have been presented. The release of BeamDyn and the AeroDyn overhaul within FAST v8 opens up new possibilities in modeling and designing advanced aeroelastically tailored blades. Following the IEC-61400-13 standard as a guideline, a comparison has been presented between FAST with and without its latest improvements, Siemens' in-house code BHawC, and the experimental data that were acquired from a series of field-test measurements from the Siemens 2.3 MW wind turbine at the NWTC.
Tables
Based on the presented results overall, FAST v8 has been sufficiently validated against field measurements, along with a code-to-code verification with BHawC. FAST v8 has been demonstrated to be valid for aeroelastic load analyses of wind turbines, even with blades with significant aeroelastic tailoring.
The results did not include uncertainties that are inherent to the experimental data-acquisition process as well as the modeling and simulation process. For example, there may be errors due to the differences in the controller used in the simulation compared to the actual machine; there may be uncertainties associated with the structural properties of the blades, tower, drivetrain, etc.; there may be uncertainties due to the limitations in the information available about the inflow wind and shear conditions measured that were used to recreate the turbulence boxes; and there may be a measurement error that has not been quantified in this paper. Analysis of all such uncertainties and their impact on the results requires a detailed, separate work, which will be a part of future work. In addition, validation in conditions with elevated yaw errors and wind speed variations, both associated with extreme conditions, correspond to disparate physics and numerics and require special considerations, which will also be a part of future work.
The software package FAST v8, the source code, its submodules,
and related information can be accessed for free at
Srinivas Guntur, Jason Jonkman, Michael A. Sprague, Scott Schreck, and Qi Wang are affiliated with NREL, which produced the FAST tool. Ryan Sievers is employed by Siemens, who produced the turbines from which data were obtained and who produced BHawC.
This work is the outcome of a collaborative research and development agreement between NREL and Siemens Wind Power, CRADA no. CRD-08-303. The submitted paper has been offered by employees of the Alliance for Sustainable Energy, LLC (Alliance), a contractor of the U.S. Government under Contract No. DE-AC36-08GO28308. The U.S. Government retains a nonexclusive, paid-up, irrevocable, worldwide license to publish or reproduce the published form of this work, or allow others to do so, for U.S. Government purposes, and the publisher, by accepting the article for publication, acknowledges this. Edited by: Gerard J. W. van Bussel Reviewed by: two anonymous referees