Proceedings of the ASME 2011 30th International Conference on Ocean, Offshore and Arctic Engineering

OMAE2011
June
Proceedings of the 30th International Conference on 19-24,
Ocean, 2011, Rotterdam,
Offshore, and ArcticThe Netherlands
Engineering
OMAE2011
June 19-24, 2011, Rotterdam, The Netherlands

OMAE2011-49863
OMAE2011-49863

DEVELOPMENT AND VERIFICATION OF A COMPUTATIONAL FLUID DYNAMICS

MODEL OF A HORIZONTAL-AXIS TIDAL CURRENT TURBINE

Michael J. Lawson ∗ Ye Li Danny C. Sale

National Renewable Energy Laboratory National Renewable Energy Laboratory University of Washington
National Wind Technology Center National Wind Technology Center Mechanical Engineering Deaprtment
Golden, CO 80401 Golden, CO 80401 Seattle, WA 98195
Email: [email protected] Email: [email protected] Email: [email protected]

ABSTRACT INTRODUCTION
This paper describes the development of a computational The kinetic energy available within tidal currents is an un-
fluid dynamics (CFD) methodology to simulate the hydrodynam- tapped source of renewable energy. If a cost-effective method
ics of horizontal-axis tidal current turbines (HATTs). First, an of capturing this energy can be developed, tidal currents could
HATT blade was designed using the blade element momentum be harnessed to help satisfy the world’s growing energy needs.
method in conjunction with a genetic optimization algorithm. Horizontal-axis tidal current turbines (HATTs) are one promis-
Several unstructured computational grids were generated using ing technology that is being developed for this purpose [1].
this blade geometry and steady CFD simulations were used to HATTs operate using the same mechanical principles as
perform a grid resolution study. Transient simulations were then wind turbines. At their most basic, HATTs convert kinetic en-
performed to determine the effect of time-dependent flow phe- ergy within a tidal stream into electricity. Several companies
nomena and the size of the computational timestep on the numer- have developed HATTs with the intent of commercializing the
ical solution. Qualitative measures of the CFD solutions were technology. Two examples are Marine Current Turbine’s twin-
independent of the grid resolution. Conversely, quantitative com- rotor SeaGen that is undergoing testing off the coast of Northern
parisons of the results indicated that the use of coarse computa- Ireland, and Verdant Power’s single rotor turbine that has been
tional grids results in an under prediction of the hydrodynamic deployed successfully in the East River near New York City.
forces on the turbine blade in comparison to the forces predicted The ability to design efficient HATT blades and accurately
using more resolved grids. For the turbine operating conditions predict their performance is critical for the success of the tidal
considered in this study, the effect of the computational timestep current turbine industry. Traditionally, reduced order numerical
on the CFD solution was found to be minimal, and the results methods, such as the blade element momentum method and the
from steady and transient simulations were in good agreement. vortex element method, have been used to design and predict
Additionally, the CFD results were compared to corresponding the performance of wind turbine blades. Batton et al. [2, 3] and
blade element momentum method calculations and reasonable Baltazar et al. [4,5] recently demonstrated that these methods can
agreement was shown. Nevertheless, we expect that for other also be employed to characterize HATT blade performance. In
turbine operating conditions, where the flow over the blade is both cases, results from the numerical simulations were shown
separated, transient simulations will be required. to be in good agreement with experimental measurements of an
HATT made by Bahaj et al. [6, 7].
Nevertheless, it is important to note that the blade element
momentum method and the boundary element method rely on ex-
∗ Address all correspondence to this author. perimental measurements and empirical correlations to achieve

1 Copyright © 2011 by ASME

Downloaded From: http://proceedings.asmedigitalcollection.asme.org/ on 02/01/2016 Terms of Use: http://www.asme.org/about-asme/terms-of-use

 accurate results. For example, blade element momentum calcu- TABLE 1. TURBINE SPECIFICATIONS.
lations use compiled experimental data to estimate the hydrody-
namic forces on the blade, whereas boundary element methods
typically employ empirical correlation to account for far-field
wake effects. Unfortunately, neither method inherently models
the effects of viscosity, which need to be considered to achieve
the most accurate turbine performance predictions possible.
Conversely, computational fluid dynamics (CFD) simula-
tions of the Navier-Stokes equations model fluid flows starting
from first principles, and therefore inherently capture viscous ef-
fects. In the last decade, CFD models of wind turbines have been
used to accurately reproduce experimental results (e.g. [8–12])
and have been used extensively to study complex flow phenom-
ena that have not been quantifiable through the use of simplified
numerical methods. Given the successful use of CFD in the wind
industry, it is practical to utilize CFD tools to advance HATT
technology. mary shape for the turbine blade. A NACA 63-series blade was
selected because the minimum pressure coefficient is relatively
large, which makes this airfoil resistant to cavitation. Addition-
Objective
ally, the NACA 63-series blades were an attractive choice for the
The objective of this study is to develop a CFD methodology
current application because they are known to delay stall and are
to simulate HATTs and to determine the effects of computational
less sensitive to leading edge roughness than the NACA 4- and
grid resolution and computational timestep on the accuracy of
5-series airfoils.
the simulations. First, we used the blade element momentum
Given that the blade design was intended for a VSVP tur-
method in conjunction with a genetic optimization algorithm to
bine, we assumed a circular cross-section at the blade root (to
design an HATT blade. Using this blade geometry, several CFD
allow for a blade-pitching mechanism) that transitioned to the
simulations were performed using computational grids with suc-
NACA 63(1)-424 airfoil shape at 20% of the blade span. It may
cessively finer spatial resolution. Finally, transient simulations
be possible to improve the performance of the blade (i.e. in-
were performed to determine if temporal effects must be consid-
crease power generation) by using several different airfoil shapes
ered to make accurate predictions of HATT performance. CFD
to define the blade geometry. Future studies will be performed to
predictions of turbine rotor torque and blade root flap moment
explore this possibility.
were also compared to corresponding blade element momentum
calculations. It is noted that the CFD model only includes the tur-
bine blade and hub geometries. Other turbine components, such Blade optimization
as the tower and nacelle, were not considered in this preliminary We used our recently developed numerical tool, Harp Opt
study. [14], to conduct the basic turbine blade design. Harp Opt is the
combination of the blade element moment code WT Perf [15]
with a genetic algorithm to optimize the blade twist angle and
TURBINE BLADE DESIGN chord length distribution along the span. Harp Opt has been
Specifications and operating conditions widely used by industry and academia to study HATTs.
In this paper we consider a hypothetical 550 KW turbine in- The optimization objective for the blade design process was
tended for deployment in the Northern Admiralty Inlet of Puget to maximize power generation over a range of flow speeds from
Sound in Washington State. An acoustic Doppler current profiler 0.5 m/s to 3.0 m/s. Within this range of flow speeds, Harp Opt
survey at this site [13] shows the mean water velocity is approx- determined the rotor’s rotation speed and the blade’s pitch an-
imately 1 m/s, although velocities as high as 3 m/s occur during gle. The optimization algorithm bounded the turbine’s maximum
the tidal cycle. power output at 550 kW and also avoided cavitation. Specifi-
The design and operating specifications for the turbine are cally, cavitation was avoided by comparing the minimum pres-
presented in Tab. 1. The turbine considered in this study had a sure over the blade to the fluid vapor pressure and avoiding blade
20 m diameter rotor comprised of a 2 m diameter hub and two 9 designs and turbine operating conditions for which cavitation is
m blades. The blade design was optimized for a variable-speed likely. XFOIL was used to predict the NACA 63(1)-424 airfoil’s
variable-pitch (VSVP) turbine with a maximum rotation rate of lift, drag, and pressure coefficients used for the blade element
11.5 rpm. A NACA 63(1)-424 airfoil was chosen as the pri- momentum analysis.

2 Copyright © 2011 by ASME

Downloaded From: http://proceedings.asmedigitalcollection.asme.org/ on 02/01/2016 Terms of Use: http://www.asme.org/about-asme/terms-of-use

 FIGURE 1. THREE-DIMENSIONAL REPRESENTATION OF THE
TURBINE BLADE DESIGNED USING Harp Opt.

Figure 1 presents an image of the blade and Fig. 2 summa-

rizes turbine operating conditions determined by Harp Opt. The
WT Perf blade element momentum calculations show that the
angle of attack for the blade was well below the stall angle across
the entire blade span for all operating conditions. Thus the flow
over the blade was expected to remain fully attached under nor-
mal operating conditions.

CFD METHODOLOGY
CFD simulations predict the motion of a fluid within a given
domain by discretizing the governing Navier-Stokes equations
and solving the resulting set of algebraic equations. The remain- FIGURE 2. TURBINE OPERATING CONDITIONS DETER-
der of this paper describes the development and verification of a MINED USING Harp Opt.
CFD methodology to quickly and accurately simulate the perfor-
mance of an HATT.

Computational domain and boundary conditions

Assumptions
The CFD model assumed that the density of sea water was We took advantage of the 180 degree periodicity of the two-
constant and that the flow was free of cavitation (in accordance bladed rotor and only simulated one blade and half of the hub.
with the Harp Opt design). Additionally, the effects of gravity Accordingly, the computational domain consisted of a single tur-
were neglected in the CFD simulations, which affectively ren- bine blade and half of the hub placed in a large half-cylinder with
dered the blade neutrally buoyant. Future CFD simulations could a periodic boundary condition applied across the bottom of the
be easily adapted to include gravitational forces, thus allowing domain as shown in Fig. 3. The domain was separated into two
the effects of buoyancy and cavitation to be studied. subdomains to facilitate simulating the rotation of the rotor. One
subdomain contained the grid elements surrounding the rotor and
those in the wake region (referred to as the rotor domain), while
Turbulence modeling the other subdomain contained the cells in the outer region (re-
The k-ω SST (shear stress transport) Reynolds-averaged ferred to as the outer domain).
Navier-Stokes (RANS) sub-grid scale model [16] was used to A uniform, temporally steady velocity profile was assumed
simulate turbulence in the flow. The k-ω SST model was selected at the inlet of the computational domain. The authors are not
because it has been successfully used to simulate turbulent flow aware of any spatiotemporal turbulent intensity measurements in
over airfoils. Additionally, several CFD studies of wind turbines a tidal current stream. Thus we made the simplifying assumption
(e.g. [8–12, 17, 18]) have successfully used the k-ω SST model that the turbulence intensity was 5% and uniform at the inlet of
to make numerical predictions of turbine performance. A few the computational domain. A pressure outlet boundary condition
of these studies (e.g. [8–12]) have shown good agreement with was applied at the outlet. To reduce flow recirculations on the
experimental measurements [19]. far-field boundaries and to improve the stability of the numerical

3 Copyright © 2011 by ASME

Downloaded From: http://proceedings.asmedigitalcollection.asme.org/ on 02/01/2016 Terms of Use: http://www.asme.org/about-asme/terms-of-use

 TABLE 2. COMPUTATIONAL GRID STATISTICS.

spacing in this region. The height of the first prism layer was set
so that the grid elements against the blade surface were within
the logarithmic region of the boundary layer between y+ = 30
and y+ = 100, depending on the location of the grid element on
FIGURE 3. THE COMPUTATIONAL DOMAIN WITH GRID RE- the blade and on the turbine operating condition simulated. Wall
FINEMENT REGIONS SHOWN. functions derived from equilibrium turbulent boundary layer the-
ory were utilized to model turbulent effects below the first grid
cell.
solution, a slip (or symmetry) boundary condition was applied Five grids were generated for use in the grid refinement and
across the curved cylindrical surface of the outer domain. Lastly, timestep studies. Table 2 summarizes the statistics for each grid
a no-slip boundary condition was specified on all rotor surfaces. and Fig. 4 illustrates the grid resolution achieved. Henceforth,
the grids will be referred to by the grid number given in Tab. 2.
Grid generation
The fidelity of any CFD simulation depends on the type of Numerical methods
grid utilized (e.g. structured or unstructured) and the grid resolu- The commercial CFD code STAR CCM+ was used to solve
tion. Structured hexahedral grids are the most desirable because the incompressible Navier-Stokes equations using a second-
they inherently provide highly accurate numerical solutions. As order-accurate finite-volume discretization scheme via a seg-
such, they have generally been used in CFD simulations of tur- regated algebraic multi-grid iterative solver. A second-order-
bines. The downside of using structured grids is that the grid accurate backward temporal discretization was used for transient
generation process is a manually intensive and time consuming simulations.
task. In contrast, the creation of unstructured grids is a more au- Steady simulations were judged to be converged when the
tomated process that can be performed in a fraction of the time. residuals of the numerical solution stabilized and were reduced
However, when using unstructured grids, care must be taken to by approximately three orders-of-magnitude. Several global pa-
insure the grid quality (e.g. grid element skewness, grid element rameters, including rotor torque and thrust, were also monitored
aspect ratio, and grid density) does not adversely impact the ac- to ensure convergence of the solution. Unsteady simulations
curacy of the numerical solution. were deemed converged when the average rotor torque over a
In the present study, unstructured grids consisting of poly- revolution was within approximately 1% of the value from the
hedral elements having an average of 14 faces per element were previous revolution. Within each timestep, residuals were re-
generated using the STAR CCM+ (CD-Acapco; Melville, New duced by three orders-of-magnitude before the solution was ad-
York) grid generation utility. The grid resolution was defined by vanced in time.
specifying the grid size and expansion rate on all the surfaces of A rotating reference frame model was used to simulate the
the computational domain. Additionally, we found it critical to rotation of the blade and hub. This method simulates rotation
use volumetric refinement zones to increase the grid resolution without the need to physically rotate the computational grid by
around the blade, the rotor, and in the wake region as shown in forming the governing equations for the rotor domain in a refer-
Fig. 3. These refinement zones allowed for the grid resolution ence frame that rotates with the turbine, while the outer domain
to be increased in regions where high velocity gradients were remains in a stationary coordinate system.
expected, without adding additional grid elements in regions of All simulations were performed on the 153 node Redrock
the computational domain where a coarse grid was acceptable. computer cluster at the National Renewable Energy Laboratory.
Prism layer grid elements were extruded from the surfaces of the Each node is comprised of two 2.93 GHz quad-core Intel Ne-
blade and the hub to improve the grid quality and control the grid halem processor with 12 GB of shared memory. All simulations

4 Copyright © 2011 by ASME

Downloaded From: http://proceedings.asmedigitalcollection.asme.org/ on 02/01/2016 Terms of Use: http://www.asme.org/about-asme/terms-of-use

 TABLE 3. TURBINE OPERATION CONDITIONS SIMULATED
WITH STAR CCM+. NOTE THAT THE ROTOR SPEEDS AND THE
BLADE PITCH ANGLES FOR THE CORRESPONDING FLOW VE-
LOCITIES WERE DETERMINED USING THE HARP OPT DESIGN
CODE.

FIGURE 4. COMPARISON OF THE SPATIAL RESOLUTION

ACHIEVED FOR (a) GRID 1 AND (b) GRID 5. THE RESOLUTION FIGURE 5. BLADE PRESSURE DISTRIBUTION IN PASCALS
OF THE OTHER COMPUTATIONAL GRIDS FALLS BETWEEN FOR (a) GRID 1 AND (b) GRID 5 FOR OPERATING CONDITION
THESE TWO EXTREMES. C.

were performed using 64 cores. Steady simulations required be- sure distributions and velocity contours for operating condition
tween 64 and 960 cpu-hours, depending on the grid size. Tran- C are presented in Figs. 5 and 6, respectively. Although the
sient simulations were performed using grid 1 and required be- blade pressure distributions are nearly indistinguishable for the
tween 256 and 1500 cpu-hours, depending on the size of the com- different grids, a close examination of the velocity contours re-
putational timestep. veals that small structures within the wake were better resolved
using the higher resolution grid. A more diffuse wake region was
predicted using grid 1 because the coarse grid cells in the wake
RESULTS AND DISCUSSION region (see Fig. 4) were insufficient to resolve the shear layer be-
Grid resolution study tween the turbine’s wake and the free-stream flow. Note that the
The three turbine operating conditions presented in Tab. 3 CFD results for operating conditions A and B exhibited similar
were simulated under a steady flow assumption using each com- qualitative trends to those shown in Figs. 5 and 6.
putational grid. The results from these simulations were used to To quantitatively evaluate the effect of grid resolution on the
study the influence of the grid resolution on the numerical solu- numerical solution, the rotor torque and blade root flap moment
tion. For brevity, graphical representations of the CFD solution were calculated for each grid. Figures 7 and 8 present these quan-
(i.e. contour plots and iso-surfaces) will only be presented for tities plotted verses grid resolution for operating conditions A, B,
grids 1 and 5, the coarsest and finest resolution grids, respec- and C. The relative differences between rotor torque and blade
tively. root flap moment predicted by grids 1–4 with respect to the value
Qualitatively, the features of the CFD solutions were found predicted using grid 5 are presented in Tab. 4.
to be markedly similar for each grid. To illustrate the qualita- As Fig. 7 illustrates, the CFD rotor torque predictions ex-
tive similarities between the solutions, examples of blade pres- hibited monotonic convergence with increasing grid resolution.

5 Copyright © 2011 by ASME

Downloaded From: http://proceedings.asmedigitalcollection.asme.org/ on 02/01/2016 Terms of Use: http://www.asme.org/about-asme/terms-of-use

 FIGURE 6. VELOCITY CONTOURS IN A PLANE THROUGH
THE CENTER OF THE COMPUTATIONAL DOMAIN FOR (a) GRID
1 AND (b) GRID 5 FOR OPERATING CONDITION C. NOTE THAT
THE FLOW IS FROM LEFT TO RIGHT.

TABLE 4. THE RELATIVE DIFFERENCE BETWEEN ROTOR

TORQUE AND BLADE ROOT FLAP MOMENT PREDICTIONS
MADE USING GRIDS 1-4 AND PREDICTIONS FROM GRID 5.

FIGURE 7. THE EFFECT OF GRID RESOLUTION ON ROTOR

TORQUE PREDICTIONS FOR TURBINE OPERATION CONDI-
TIONS (a) A, (b) B, AND (c) C.

relative flow velocity was highest.

One prominent difference between the predicted flow fields
for grids 1 and 5 was visualized using iso-surfaces of vorticity
to elucidate the blade tip-vortices. Figure 10 shows that the tip-
Moreover, torque predictions for grids 4 and 5 were within 1% vortices predicted by grid 1 were much larger and enveloped a
of each other for all operating conditions simulated. Thus, based larger portion of the blade than the tip-vortices predicted by grid
on these torque convergence data, it was concluded that further 5. Effectively, grid 1 tip-vortices induced more downwash on the
increasing the grid resolution would have produce a negligible suction side of the blade, which decreased the torque generation
change (almost certainly less than 1%) in the rotor torque pre- near the blade tip as Fig. 9 clearly illustrates.
dicted by CFD simulations. However, the results showed that Predictions of the blade root flap moment were insensitive to
use of a very coarse grid results in a significant under prediction the grid resolution (see Fig. 8 and Tab. 4), although the conver-
of rotor torque. Indeed, as Tab. 4 shows, grid 1 consistently under gence of this quantity was not monotonic. Nevertheless, given
predicted rotor torque by 11%–12% compared to grid 5. that the maximum difference in the predicted flap moment be-
To further investigate the cause of the discrepancy in the ro- tween grids 1 and 5 was only 1.2%, this non-monotonic conver-
tor torque predictions between the different grids, we computed gence was not of significant concern. One possible reason this
the torque contribution from 40 evenly sized blade sections along convergence behavior was observed is that a steady flow assump-
the span. Figure 9 clearly shows that grid 1 significantly under tion may not be appropriate for the highly unsteady far-wake re-
predicted the rotor torque with respect to grid 5 on the outboard gion. More specifically, the near-wake region is characterized
section of the blade, which was where the majority of the rotor by a highly organized vortex wake with a structure that is ef-
torque was generated. This suggests that grid 1 had insufficient fectively time-independent with respect to the rotating reference
resolution to accurately capture the physical phenomena of the frame model. In contrast, the far-wake region is inherently un-
water flow over the outboard sections of the blade, where the steady regardless of the reference frame used.

6 Copyright © 2011 by ASME

Downloaded From: http://proceedings.asmedigitalcollection.asme.org/ on 02/01/2016 Terms of Use: http://www.asme.org/about-asme/terms-of-use

 FIGURE 9. SECTIONAL TORQUE CONTRIBUTIONS FOR
GRIDS 1, 3, AND 5.

FIGURE 8. THE EFFECT OF GRID RESOLUTION ON BLADE

ROOT FLAP MOMENT PREDICTIONS FOR TURBINE OPERA-
to transient simulations performed using the more resolved grids.
TION CONDITIONS (a) A, (b) B, AND (c) C.
The transient simulations were performed using timesteps of
0.05, 0.01, and 0.005 seconds for operating condition C. These
timesteps correspond 260, 521, and 1044 timesteps per rotor rev-
Although a large difference in the predicted rotor torque olution, respectively. To reduce the required time for transient
(11–12%) was seen between grids 1 and 5, a less pronounced dif- simulations to reach a quasi-steady state, the simulations were
ference (approximately 3%) was observed between grids 3 and 5. initialized using the steady flow solution. For all transient sim-
Thus, if the purpose of CFD simulations is to accurately predict ulations, the torque stabilized within approximately three simu-
global turbine performance characteristics (e.g. rotor torque, flap lated rotor revolutions. Significant oscillations in the rotor torque
moment, and hydrodynamic forces) a grid with a resolution sim- were not observed and were not expected as the angle of attack
ilar to grid 3 appears sufficient. However, if fine details in the of the blade over the entire span was well below the stall angle.
flow (e.g. the structure of tip vortices and details of the blade Visualization of limiting streamlines on the blade confirmed the
boundary layer and wake) must be resolved, a finer grid is re- flow was fully attached.
quired.
The size of the computational timestep was found to have
very little effect on the magnitude of torque generated by the tur-
Timestep study bine as Tab. 5 shows. Further, the difference between the torque
Transient simulations were performed for two reasons: (1) predicted by the steady and unsteady simulations was within
to determine if time-accurate CFD simulations provide signif- 2.5%. This suggests that for attached flow conditions, steady
icantly different results from corresponding steady simulations; CFD simulations sufficiently captured the physical phenomena
and (2) to evaluate the effect of the computational timestep on the that determine the magnitude of hydrodynamic loading on the
numerical solution. Grid 1 was used for all transient simulations blade. This conclusion must, however, be considered carefully
to reduce the computational resources required to perform the given the set of assumptions used for the present simulations. It
timestep study. It is generally acceptable to use separate grid and was assumed the inflow conditions were spatiotemporally uni-
timestep convergence studies to determine the effect of spatial form. If a time-varying nonuniform velocity profile containing
and temporal resolution on CFD solutions [20]. Thus we expect realistic turbulent structures is used, time-accurate simulations
the results from the timestep study using grid 1 to be applicable are required.

7 Copyright © 2011 by ASME

Downloaded From: http://proceedings.asmedigitalcollection.asme.org/ on 02/01/2016 Terms of Use: http://www.asme.org/about-asme/terms-of-use

 FIGURE 10. VISUALIZATION OF BLADE TIP-VORTICES FOR
(a) GRID 1 AND (b) GRID 5. NOTE THAT THE SAME VORTEX
THRESHOLD IS USED IN BOTH IMAGES.
FIGURE 11. COMPARISON OF ROTOR TORQUE AND BLADE
ROOT FLAP MOMENT PREDICTIONS FROM STAR CCM+ (US-
TABLE 5. RESULTS FROM THE TIMESTEP STUDY. FOR COM- ING GRID 5) AND WT Perf FOR OPERATING CONDITION C.
PARISON, THE CORRESPONDING RESULT FROM THE STEADY
SIMULATIONS IS ALSO SHOWN. NOTE THAT ALL RESULTS
ARE FOR GRID 1.
namic forces were under predicted across the entire blade span
by WT Perf in comparison to the CFD solution. Unfortunately,
there is no experimental data available for the current blade de-
sign, making it impossible to assess the accuracy of either numer-
ical method. Nevertheless, the reasonable agreement between
the CFD results and the blade element momentum method cal-
culations provides confidence in the results obtained with both
numerical methods.

CONCLUSIONS AND FUTURE WORK

A RANS CFD methodology to simulate HATTs was devel-
Comparison of the CFD results with blade element mo- oped and the numerical results were verified via a grid refinement
mentum method predictions study and a timestep study. Steady simulations were performed
The rotor torque and blade root flap moment predictions using successively more resolved computational grids to deter-
from steady CFD simulations performed using grid 5 were com- mine the effect of grid resolution on CFD predictions. Quali-
pared with the corresponding blade element momentum predic- tatively, the CFD solutions calculated using the different grids
tions from WT Perf. Figure 11 presents this comparison. It is were markedly similar. However, the use of a coarse compu-
clear that the blade element momentum method significantly un- tational grid resulted in a significant under prediction of rotor
derestimated the hydrodynamic forces on the blade compared to torque compared to results obtained with higher resolution grids.
the CFD solution. Indeed, plotting the sectional torque contribu- Transient simulations were performed using successively
tions to the total rotor torque (Fig. 12) revealed that the hydrody- smaller computational timesteps to determine if temporal effects

8 Copyright © 2011 by ASME

Downloaded From: http://proceedings.asmedigitalcollection.asme.org/ on 02/01/2016 Terms of Use: http://www.asme.org/about-asme/terms-of-use

 A: Journal of Power and Energy, 216(1), pp. 1–14.
[2] Batten, W. M. J., Bahaj, A. S., Molland, A. F., and Chap-
lin, J. R., 2008. “The prediction of the hydrodynamic per-
formance of marine current turbines”. Renewable Energy,
33(5), pp. 1085–1096.
[3] Batten, W. M. J., Bahaj, A. S., Molland, A. F., Chaplin,
J. R., and Sustainable Energy Res, G., 2007. “Experimen-
tally validated numerical method for the hydrodynamic de-
sign of horizontal axis tidal turbines”. Ocean Engineering,
34(7), pp. 1013–1020.
[4] Baltazar, J., and de Campos, J., 2008. “Hydrodynamic
Analysis of Horizontal Axis Marine Current Turbine with
a Boundary Element Method”. Proceedings of the 27th In-
ternational Conference on Offshore Mechanics and Arctic
Engineering.
[5] Baltazar, J., and de Campos, J. A. C. F., 2009. “Un-
steady analysis of a horizontal axis marine current turbine
in yawed inflow conditions with a panel method”. First In-
ternational Symposium on Marine Propulsors.
[6] Bahaj, A., Batten, W., and McCann, G., 2007. “Exper-
FIGURE 12. CONTRIBUTIONS TO ROTOR TORQUE PRE- imental verifications of numerical predictions for the hy-
DICTED USING CFD AND WT Perf FOR OPERATING CONDI- drodynamic performance of horizontal axis marine current
TION C. turbines”. Renewable Energy, 32(15), pp. 2479–2490.
[7] Bahaj, A., Molland, A., Chaplin, J., and Batten, W., 2007.
“Power and thrust measurements of marine current turbines
must be considered to make accurate predictions of turbine per- under various hydrodynamic flow conditions in a cavita-
formance. For the turbine operating conditions considered, the tion tunnel and a towing tank”. Renewable Energy, 32(3),
computational timestep was shown to have little effect on the nu- pp. 407–426.
merical solution. Additionally, the results from steady and tran- [8] Pape, A., and Lecanu, J., 2004. “3D Navier–Stokes com-
sient simulation were shown to be in general agreement. putations of a stall-regulated wind turbine”. Wind Energy,
It is important to note that the flow over the turbine blade was 7(4), pp. 309–324.
fully attached for all the operating conditions simulated herein. [9] Sorensen, N., Michelsen, J., and Schreck, S., 2002.
The consideration of more extreme operating conditions, where “Navier-stokes predictions of the NREL phase vi rotor in
the flow over the blades is separated and inherently unsteady, will the NASA Ames 80 ft by 120 ft wind tunnel”. Wind En-
likely require the use of transient CFD simulations to achieve ergy, 5(2-3), pp. 151–169.
accurate numerical results. Moreover, when the flow over the [10] Johansen, J., Sorensen, N., Michelsen, J., and Schreck, S.,
blades is separated, predictions of transition, separation, and flow 2002. “Detached-eddy simulation of flow around the NREL
reattachment have a significant influence on the accuracy of CFD Phase VI blade”. Wind Energy, 5(2-3), pp. 185–197.
solutions [8]. Thus, in these cases, advanced RANS models ca- [11] Duque, E., Burklund, M., and Johnson, W., 2003. “Navier-
pable of predicting these complex physical phenomena may be Stokes and comprehensive analysis performance predic-
required. tions of the NREL phase VI experiment”. Journal of Solar
This work represents the first step in the development of a Energy Engineering, 125, p. 457.
CFD methodology to characterize HATT performance. We plan [12] Tongchitpakdee, C., Benjanirat, S., and Sankar, L., 2005.
to further develop this model by considering realistic inlet ve- “Numerical simulation of the aerodynamics of horizontal
locity profiles and turbulence levels. Additionally, we hope to axis wind turbines under yawed flow conditions”. Journal
include structural elements of the turbine, such as the tower and of solar energy engineering, 127, p. 464.
the nacelle, in future simulations. [13] Polagye, B., and Thomson, J., 2010. Tidal Energy Refer-
ence Model Inflow Conditions. Tech. Rep. Unpublished,
University of Washington.
REFERENCES [14] Sale, D. C., and Li, Y., 2010. “Preliminary results from
[1] Fraenkel, P., 2002. “Power from marine currents”. Pro- a design methodology and optimization code for horizon-
ceedings of the Institution of Mechanical Engineers, Part tal axis wind and hydrokinetic turbines.”. The 29th Inter-

9 Copyright © 2011 by ASME

Downloaded From: http://proceedings.asmedigitalcollection.asme.org/ on 02/01/2016 Terms of Use: http://www.asme.org/about-asme/terms-of-use

 national Conference on Ocean, Offshore And Arctic Engi-
neering.
[15] http://wind.nrel.gov/designcodes/simulators/wtperf/.
NWTC Design Codes - WT Perf, Accessed January 15,
2011.
[16] Menter, F., 1994. “Two-equation eddy-viscosity turbulence
models for engineering applications”. AIAA journal, 32(8),
pp. 1598–1605.
[17] Laursen, J., Enevoldsen, P., and Hjort, S., 2007. “3D CFD
quantification of the performance of a multi-megawatt wind
turbine”. In Journal of Physics: Conference Series, Vol. 75,
IOP Publishing, p. 012007.
[18] Johansen, J., Madsen, H., Gaunaa, M., Bak, C., and
Sørensen, N., 2009. “Design of a wind turbine rotor for
maximum aerodynamic efficiency”. Wind Energy, 12(3),
pp. 261–273.
[19] Simms, D., 2001. NREL unsteady aerodynamics exper-
iment in the NASA-Ames wind tunnel: a comparison of
predictions to measurements. Tech. Rep. NREL/TP-500-
29494, NREL.
[20] Roache, P. J., 1998. Verification and Validation in Compu-
tational Science and Engineering. Hermosa, Albuquerque,
NM.

10 Copyright © 2011 by ASME

Downloaded From: http://proceedings.asmedigitalcollection.asme.org/ on 02/01/2016 Terms of Use: http://www.asme.org/about-asme/terms-of-use