You are on page 1of 14

An Analysis of Wind Turbine Farm Arrangements

and its Affect on Power Output

Kevin Winstead

The University of North Carolina at Charlotte


William States Lee College of Engineering
Applied Computational Fluid Dynamics, Fall 2014

Table of Contents
Objective: ................................................................................................................................................. 3
Abstract: ................................................................................................................................................... 3
Methodology: .......................................................................................................................................... 3
Geometry: ................................................................................................................................................ 4
Mesh: ......................................................................................................................................................... 5
Physics: ..................................................................................................................................................... 7
Plots of Interest: .................................................................................................................................... 8
Flow Visualization: ............................................................................................................................... 9
Numerical Results: ............................................................................................................................. 13
Conclusion: ........................................................................................................................................... 14
References ............................................................................................................................................ 14

Objective:
The objective of this project is to analyze a wind turbine farm using CFD and determine how the
arrangement of turbines inside that farm affects the power output of each turbine and the farm as
a whole. Along with the power analysis, flow visualizations will be analyzed and described.
Abstract:
The purpose of this project is to analyze the arrangement of turbines in a wind turbine farm to
determine which will result in max power from the farm and turbines individually. This analysis
was completed using Star-CCM and each turbine was modeled as a porous disk with a porosity
of 55%. The simulated farm was made up of 25 disks arranged in a 5 x 5 grid. In total there were
three simulations completed. One with a normal grid, one with the rows staggered side-to-side,
and one with the rows staggered up and down. In the end, the results showed that staggering the
rows in either direction resulted in a marked increase in power output. In some cases over a
200% increase in power was calculated. The turbines on the first row were the least affected by
staggering the rows and the second row was affected the most. The third, fourth, and fifth rows
were each affected about the same with a gain in power of slightly less than 200%.
Methodology:
The methodology that I used in analyzing the different wind farm arrangements was taken from
numerous scholarly papers and research that I completed prior to starting the simulations. From
this research I determined that wind turbine farms have been reliably simulated in the past using
a porous disk to represent the wind turbine blades. The reason for using the porous disk is simple,
computational cost. To model and mesh an actual wind turbine with the blades down to a small
enough scale would require an enormous number of cells for just one wind turbine. In this
analysis, 25 turbines would be used. This made it completely unfeasible to model the details of
each wind turbine and its components.
The porous disk used in these simulations was modeled after the GE 1.5xle wind turbine. This
wind turbine is one of GEs most popular and is the most widely used in its class. It has a
maximum output of 1.5MW and reaches this max power at a wind speed of about 10-12 m/s.
For each one of the simulations, probes were inserted into the flow to measure the velocity. A
probe point was placed 50m in front of each disk and velocity was measured in the direction of
flow. A report was then created for each one of these points and then monitors and a plot were
created to visualize the velocity as the iterations were completed. In total, 75 probe points were
inserted into the flow over the three simulations. Once the simulations had finished iterating,
these velocities were recorded and input into a spreadsheet.
1
(1)
P = AV 3
2
Equation 1, above, was used to calculate the representative power for each disk using the
measured velocity, the known density, and the known area. These power values were then
analyzed to see what effect the manipulation of the disks had on the power output.

Geometry:
As described above, the geometry in these simulations was relatively simple. It was simple
enough in fact, to enable me to create all of the geometrical features within Star-CCM. This
greatly reduced the complications associated with trying to import a mesh from an outside
program.

10x dia.

30x dia.

Figure 1: General Arrangement of Disks in Tunnel


Figure 1, above, shows the general arrangement of the blades within the wind tunnel that was
created in Star-CCM. The distance from the inlet to the first row of disks was chosen based on
the research that was completed prior to the start of the simulations and is representative of
enough space for the flow to fully develop before reaching the first row of disks. The distance
between the last row and the outlet of the wind tunnel was chosen to be longer because of the
large wake that will be developed as a result of the flow passing all of the disks. These two
distances are similar to that chosen by other researchers performing similar research.
For all of the simulations, the disks were set at a distance of four times the diameter away from
each other side-to-side. A distance of eight times the diameter separated each of the rows. These
numbers were chosen based on the best results form others who have completed similar research
and from the research that was completed prior to the beginning of the project.

Figure 2: Geometry of Normal Grid Simulation

For the side-to-side staggered simulation the second and fourth rows were offset by half the
distance between two disks on the same row. This means that the disks on the second row fall
directly in-between the two disks of the row in front of it.

Figure 3: Geometry of Side-to-Side Staggered


For the up-and-down staggered simulation, the second and fourth rows were raised up to a hub
height of 200m. This hub height is double what the original hub height was and allows the
second and fourth rows to be high enough to not impact flow over the lower rows.

Figure 4: Geometry of Up-Down Staggered


Mesh:
The mesh of any CFD simulation is of critical importance. The mesh is what defines how well
your result will be. The finer the mesh, the finer the result will be. For this case, determining the
appropriate mesh size is a game of trial and error. For all of the simulations that were a part of
this project, each simulation ended up with about 2.1 million cells total.
Models

Polyhedral Mesher
Prism Layer Mesher
Surface Remesher

Base Size 82.5m (1 dia.)


Prism Layer Parameters
8 Prism Layers
1.5 Prism Layer Stretching
20% Prism Layer Thickness
Surface Size
15% Relative Minimum
30% Relative Target
Volumetric Control
15% Relative Size

Figure 5: Mesh Scene

Figure 6: Up Close Mesh Scene with Disk


Figures 5 and 6 above show the mesh that was developed for the project. A relatively coarse
mesh was chosen for the outer area of the tunnel that was away from the disks. A volume source
was created and allowed for a finer mesh around the disks. Figure 6 shows the mesh along a cut
place going through the middle of the tunnel. For the physics models that were used in this
project, it is important to try and have a fine of a mesh as possible, especially in the wake region,
in order for the solution to come out correctly.

Physics:
The physics was chosen for this project to allow for a solution to easily converge.
Models
Three Dimensional
Steady
Ideal Gas
Coupled Flow
K-Epsilon Turbulence
Boundary Conditions
Velocity Inlet - 10m/s
Pressure Outlet
Top and Sides - Symmetry
Ground - Wall

Plots of Interest:

Figure 7: Simulation Residuals


Figure 7 shows the residuals for the normal grid setup. This plot is almost identical to the
residual plot for both of the staggered simulations. Overall, the residuals are a little higher than
would be ideal. Even though the residuals are high, they have all completely flat-lined and show
that the simulation can be stopped.

Figure 8: Side-to-Side Staggered Velocity vs. Iteration


Figure 8, above, is probably the most important depiction of the results of the simulation. This
shows each of the 25 velocity measurements vs the number of iterations that were performed.
Overall, these values have mostly flat-lined and can be taken and used in the data analysis. It is
interesting how the group divides itself into two after about 100 iterations. The lower and much
steadier group is the velocities of the first row where the flow is calm and not very turbulent.
Once the flow passes over the first row and gets to the later rows, it is much more turbulent and
varies a lot more over the iterations. Although the top group still has large variations, overall
they are not diverging and probably wont get any better with further iterations.

Both Figure 7 and Figure 8 are from one of the simulations but ar (S. Aubrun n.d.) (Konstantinos
Rados n.d.) (Giorgio Crasto n.d.) (Melheim 2011) (General Electric n.d.)e overall representative
of the results from all of the simulations.
Flow Visualization:

Figure 9: Velocity Scalar Scene Normal Grid


The velocity scene shown in Figure 9 clearly shows the effect the wake of one turbine has onto
another. The wake of each turbine extends all the way into the interaction with the turbine behind
it.

Figure 10: Pressure Scalar Scene Normal Grid

From Figure 10 above, one can see how big of an affect the arrangement of the turbines has on
the pressure in front of each of the disks.

Figure 11: Vorticity Scalar Scene Normal Grid


The vorticity scene above also shows how the vorticity of the front turbines propagates through
all of the turbines and can clearly be seen.

Figure 12: Velocity Scalar Scene Side-to-Side Staggered


Figure 10 shows the scalar velocity in the wind turbine farm. It is clear from this flow
visualization that the disks in the back are affected by the wake from the disks that are at the
front. It can also be seen that the farm as a whole has a huge wake that extends past the boundary
of my simulation.

Figure 13: Pressure Scalar Scene Side-to-Side


The pressure scalar scene shown above also clearly shows that the pressure on the front of the
disks is much higher on the first row than it is on the last row.

Figure 14: Vorticity Scalar Scene Side-to-Side

Figure 15: Wall Y+ Scalar Scene Side-to-Side


Figure 13 shows a Y+ scalar scene for the side-to-side staggered simulation. The max Y+ value
is 0.679 which is within the necessary below 1 value.

Figure 16: Streamline Scene Side-to-Side


The streamline scene shown above in Figure 14 shows how the flow coming from one cylinder
can travel and impact the flow around the other cylinders that are behind it.

Numerical Results:
30.00

Power (MW)

25.00

23.88

22.17

20.00
15.00

11.14

10.00
5.00
0.00
Normal

Side Staggered

Up Down Staggered

Total Farm Power (MW)

Figure 17: Total Wind Farm Power Output


Figure 17 shows the total power for each farm that was calculated using the measured velocities
in front of each disk. These numbers represent the total power that could be produced by the
farm with that arrangement. As can be seen, there is quite an improvement by staggering the
arrangement of the turbines instead of leaving them in a straight line.
1.60
1.40

256%

19%

Power (MW)

1.20
1.00

195%

0.80

193%

229%

0.60
0.40
0.20
0.00
Row 1

Row 2

Regular Simulation

Row 3

Side Staggered

Row 4

Row 5

Up Down Staggered

Figure 18: Power Produced Per Row


Figure 18 shows how the power improvements look per row in each of the farms. As would be
expected, the first row showed the lowest gain by staggering. Because it is already at the front of
the flow, staggering it wouldnt make that much of a difference in the power output. The
staggering made the biggest difference in the second row, where staggering the disks up and

down caused the power output to increase by 256%. The last three rows also show a good
improvement in power output with each being around 200%.
Conclusion:
In conclusion, it can be said that staggering the turbines in a wind farm produces far greater
power than having all of the turbines lined up. To get the absolute best performance, the turbines
should be staggered in the top and bottom configuration because this setup provided a slightly
better performance over the side-to-side staggered pattern. In fact, staggering the wind turbines
can almost double their power output. In reality, the up-down staggered is probably even better
than my data shows that it is. This is due to the ability of wind turbines at different heights to be
able to play off of the boundary layer effect of the air flowing over the ground. All in all, this
was a great project to help learn how CFD can be applied to research situations and has truly
showed how powerful the software can be.
References
General Electric. "1.5 MW Wind Turbine." GE Energy.
http://geosci.uchicago.edu/~moyer/GEOS24705/Readings/GEA14954C15-MW-Broch.pdf
(accessed 12 4, 2014).
Giorgio Crasto, Arne Reidar Gravdahl. "CFD Wake Modeling Using Porous Disk."
Konstantinos Rados, Savas Mosfilis, Nikolaos Stergiannis. "CFD Modeling Approaches of Wind
Turbine Single and Multiple Wakes." (Technological Educational Insitute of Western
Macedonia).
Melheim, Jens A. "Wind and Wake Modeling Using CFD." Trondheim, 2011.
S. Aubrun, Ph. Devinant, G. Espana. "Physical Modeling of the far wake from wind turbines."
L.M.E.

You might also like