You are on page 1of 22

642082

research-article2016
WIE0010.1177/0309524X16642082Wind EngineeringMohamed and Wood

Original Article

Wind Engineering

Computational modeling of wind flow 2016, Vol. 40(3) 228249


The Author(s) 2016
Reprints and permissions:
over the University of Calgary campus sagepub.co.uk/journalsPermissions.nav
DOI: 10.1177/0309524X16642082
wie.sagepub.com

MA Mohamed and DH Wood

Abstract
This paper describes computer simulations of the wind flow over the University of Calgary campus undertaken to assess the likely
regions for siting wind turbines and photovoltaic modules. The importation and cleaning of the building geometries is discussed along
with the requirements for modeling adjacent buildings on the accurate simulation over one building in particular. The effect of the
terrain and the many trees on the campus is shown to be significant for the wind resource over the roof of a six-storey building. It is
shown that the choice of turbulence model is not critical if the purpose of the simulation is to identify regions for further exploration
via wind speed measurements.

Keywords
Wind resource assessment, urban wind energy, computational fluid dynamics, terrain, trees, turbulence models

Introduction
Deploying wind turbines in a built-up environment is a complex task when compared to siting them in a rural area unless
the latter has significant variations in topography and roughness. The reasons relate to the complexity of flow over large
buildings, which can include significant regions of stagnating and recirculating flow and their interaction with other build-
ings and any topography. Some of the difficulties are highlighted in the Warwick report on small-scale wind generation
in UK cities (Encraft, 2008). With the majority of earths population now living in cities, it is essential that opportuni-
ties for distributed, often small-scale, electricity generation on rooftops, near industrial areas and so on, are aggressively
exploited. Whereas photovoltaics (PV) currently is the most cost-effective technology for small-scale distributed genera-
tion when the resource availability is roughly equal, there are natural synergies with wind energy. PV generally requires
low wind locations to minimize the extreme wind loads (see e.g. Stathopoulos etal., 2014), whereas wind turbines clearly
perform best in high wind speeds and low turbulence. To find suitable locations for both technologies requires detailed
wind resource assessment.
The main problem with wind resource assessment for small energy projects is the cost relative to the power produced.
In many cases, computational fluid dynamics (CFD) offers a cheaper and more useful alternative than a measurement
campaign. In any case, CFD is often a good first choice because it gives the whole wind flow field. This is especially
useful in highlighting areas of high wind speed for potential wind turbine installation and low wind areas for PV. The
highlighted areas can then be surveyed effectively by cup or other anemometers.
At a scale large compared to building height, the wind flow over a city experiences varying roughness superposed
on a possibly varying topography. If the wind flows over a step change in roughness, an internal boundary layer forms
and grows with streamwise distance. The flow above it is not affected by the step, apart, possibly, from a displacement
required by continuity. Unfortunately, however, most turbines and PV modules will be installed on or around the buildings
which represent the roughness elements, so there is no alternative to modeling much of the complex geometry with high

University of Calgary, Calgary, AB, Canada

Corresponding author:
DH Wood, University of Calgary, 2500 University Drive Northwest, Calgary, AB T2N 1N4, Canada.
Email: dhwood@ucalgary.ca
Mohamed and Wood 229

Figure 1. The EEEL building. Picture taken from the University of Calgary website.

fidelity. This has the disadvantage of requiring models for a whole city that can only be solved using massive computa-
tional resources (Franke etal., 2011). In addressing the problem of using CFD and wind speed measurements to map the
renewable energy possibilities for the city of Calgary with a population of just over 1,100,000 living in an area of
approximately 848 km2, we decided to begin with the University of Calgary campus with an area of approximately 4.13
km2. The reasons were:

the building geometries and topography were available in a form suitable for meshing for CFD analysis;
the University is keen to extend its current renewable energy generation of about 7.4 kWp of PV;
the campus is reasonably unsheltered locally to the west, the predominant wind direction, and so is approximately
independent of the remaining city topography.

The major problem we faced was that no useful wind speed data has been collected over the campus. Therefore, we
decided to undertake CFD wind flow assessments prior to a measurement campaign. The Energy Environment Experiential
Learning (EEEL) building on the campus, shown in Figure 1, was selected as the main focus because it has a flat roof
suitable for installation of PV and small wind turbines. The commercial CFD code, CFX, part of ANSYS 15, was chosen
for the analyses.

Inlet boundary conditions


The wind rose for the campus was obtained from the Canadian Wind Atlas (CWA) (http://www.windatlas.ca) and is dis-
played in Figure 2, showing the prevailing wind directions are west, north-west, and southwest. The annual wind speed
is 5.21 m/s at a 50 m height. The shape factor from the Weibull distribution given by CWA is 1.74 and the scale factor is
5.85 m/s.
Based on the above information, the velocity at the inlet for the campus simulations was fitted to a log law, and the
turbulent kinetic energy, k, and its dissipation rate, , follow the usual formulation for neutral conditions

u z + z0
U= ln( ) (1)
z0

u2
k= (2)
C

u3
= (3)
( z + z0 )

where u is the friction velocity, z is the vertical displacement, z0 is the aerodynamic roughness length, and is the von
Krmn constant with a value of 0.412. C is a model constant set at 0.09. The value for u (0.301 m/s) was determined
via the estimate of z0 = 0.04 m from the CWA website and the annual wind speed at 50 m. The shear stress transport
(SST) k was chosen as the primary turbulence model as Tarokh etal. (2014) found it was the most accurate of six
turbulence models when compared with wind tunnel measurements of the flow over a rectangular building.
230 Wind Engineering 40(3)

Figure 2. Annual wind rose from CWA within the campus region.

Figure 3. CAD model of the University of Calgary in AutoCad Civil 3D, when processing the shape files. The arrow shows the
direction of a westerly wind and points to the EEEL building in the middle of the figure near the left edge.

Generating the campus geometry


The building geometry data was obtained from the Spatial & Numeric Data Services from the Universitys Taylor Family
Digital Library. The data was processed using ArcGIS software (http://www.esri.com/software/arcgis) for analyzing geo-
graphic information systems. It is able to process billions of cloud points taken from LiDAR and convert them into the
much more useful digital data structure TIN (Triangulated Irregular Network). The data was converted to shape files
before exporting them as a CAD model. The extrusions of the buildings to their actual height were performed in AutoCad
Civil 3D. This is where simplifications were made to the buildings.
The ideal computational grid would include every aspect of the physical geometry of the urban setting at least to the
scale of z0 . This is unrealistic for a manageable computation. In order to reduce the computational burden, buildings
surrounding the building of interest should be modeled only to the accuracy necessary for the accurate simulation of the
former. The reduced models should capture the main shape (Blocken etal., 2012). Figure 3 shows the resulting campus
model and Figure 4 shows the detailed and simplified model for Craigie Hall, one of the buildings surrounding EEEL
which is the main focus.
Mohamed and Wood 231

Figure 4. Explicit and implicit modeling of building geometries.

Figure 5. Surface contour generated in Matlab.

Terrain effects
The log law, equation (1), is a consequence of local equilibrium between the production of turbulence and its dissipation
rate which occurs only when the terrain is flat and homogeneous and z0 is constant. As shown by Mohamed and Wood
(2013), the other terms in the turbulent kinetic energy equation, turbulent diffusion and advection, become important for
the case of a flow over complex terrain (which may be moderate hills or other changes in elevation). There have been
many calculations of wind flow over urban terrain which do not take into account the effects of terrain (see Blocken
etal., 2012; Tabrizi etal., 2014). That is to say, the buildings were placed on a flat surface, not the real terrain. The present
calculations used both flat ground and the actual terrain represented as described in the next section.

Representing the terrain


Terrain data in the form of spatial co-ordinates was provided by the Spatial and Numeric Data Services (University of
Calgary). The points were imported into Matlab where they were parsed via the Delaunay function into triangulated
networks (Figure 5). The surface function was then used before converting the output to a stereolithography (STL) format.
Meshing was done in ANSYS ICEM. Figure 6 shows the meshed geometry of the campus on the generated terrain.
Mesh lines are removed for clarity. The mesh was computed using the Octree meshing algorithm, which produces tet-
rahedral elements around the object surfaces. Four prismatic layers were grown from the terrain while five layers were
extruded from the roofs of buildings.
The computational domain for the simulations with a westerly wind is shown in Figure 7. The domain size is based on
the recommendations of Franke etal. (2004) who recommended a maximum blockage ratio of 3%. The blockage ratio in
CFD is defined as the projected windward area of the buildings to the cross-section of the computational domain. This is to
prevent artificial acceleration of the flow over the tallest building. The advection and turbulence numerics were solved via
232 Wind Engineering 40(3)

Figure 6. The meshed campus with meshing lines hidden for clarity. Refer to compass for the layout of the campus buildings. The
x-axis points west.

Figure 7. The computational domain for the campus simulations.

a high resolution scheme which is essentially a blend of first order and second order based on spatial gradient, described
in detail by Barth and Jespersen (1989). The convergence criteria, the normalized residuals of the momentum and so on,
were set to 105 as recommended by Franke etal. (2004). Details of the mesh size are given in the next section.

Mesh independence
Prior to the actual simulations, grid independence studies were performed at the five locations depicted in Figure 8. The
results are shown in Figure 9. Meshes 1, 2 and 3 consisted of 10,748,228 nodes, 13,435,285 nodes, and 16,659,753 nodes
respectively.
The grid convergence index (GCI) is often used to report mesh independence studies without the need for grid-dou-
bling (Roache, 1997). It is an error estimator (factor of safety included) between the discrete solutions of two grids based
on a generalized Richardson extrapolation. We used this methodology to compute the GCI at the five different locations
for mean wind speed profiles (we are primarily interested in the mean wind speed) from the ground to the top of the com-
putational domain. The procedure is described in detail by Roache etal. (1986). GCI12 and GCI 23 , where GCI12 is the
index for meshes 1 and 2 and so on, for Location A defined in Figure 8, are 1.103% and 0.108% respectively. For Location
B, these values are 3.2% and 2.2249%. At Location C, there is a slightly larger error band between mesh 3 and mesh 2.
GCI 23 is 10.4408% while GCI12 is 3.8658%. At Location D, GCI12 and GCI 23 are 0.2751% and 0.0728% respectively.
Finally, at Location E, GCI12 and GCI 23 are 1.5653% and 0.4435% respectively. Although there is some discrepancy at
Mohamed and Wood 233

Figure 8. Mesh independence taken at five marked locations.

Location C, the error bands were considered acceptable. Nevertheless, the most refined grid, mesh 3, was chosen for the
simulations.

Results and discussion


Figure 10 shows the difference between simulations with and without the terrain for the campus with respect to wind
power density. The contours are shown for a plane 4 m above the EEEL building. Taller buildings protrude through the
contour. Wind power density per unit area is computed via

1 3
Pw = A3(1 + ) (4)
2 ks

where is the gamma function, A is

U
A= (5)
(1 + 1/ k s )

and U is the mean wind speed (Romani etal., 2015). Note that equation (4) is used only for the mean wind speed at 4
m above the roof of the EEEL building for each wind direction. All six simulations imply good wind power potential on
the west side 4 m above the roof of the EEEL building (this is a typical height for a building-mounted wind turbine). The
height of the EEEL building, zb , is 32 m.
The immediate observation from the simulations for a west wind is that there is more Pw for the flat ground case than
for the one with actual terrain on the west side 4 m above the roof of EEEL. The wind coming from this direction to the
EEEL building is affected mostly by terrain and not buildings (see Figure 6(b)) which underlines the importance of terrain
and roughness for wind resource on the roofs of buildings.
In order to quantify the differences in Pw between terrain and flat ground cases, an area integral of Pw was computed
for the plane 4 m above the roof of the EEEL building (shown in Figure 11). The results are displayed in Table 1.
As shown, flat ground cases give higher Pw compared to the actual terrain. This is an important result because the dif-
ferences clearly highlight terrain effects which are not even mentioned in Franke etal.s (2004) recommendations on the
use of CFD in wind engineering.
For further study of terrain effects, the vertical profiles on the line shown in Figure 12 for normalized U and k were
extracted. Both properties are normalized by U e and U e2 respectively where U e is the wind speed at 50 m at the inlet,
denoted in this paper as the reference wind speed. Figure 13 displays the results.
As with the conclusion earlier that flat-ground simulations give higher Pw , the general trend is similar for mean
wind speed profiles taken along Projection A, apart from the north-west wind. For this case, the wind profiles for both
234 Wind Engineering 40(3)

Figure 9. Velocity profiles for different meshes at four locations on campus. The height of the tallest building on campus ztb is
66 m and freestream velocity Ue is 5.21 m/s. The origin of Location E is the midpoint of the leading edge of the roof of the EEEL
building. (a) Location A. (b) Location B. (c) Location C. (d) Location D. (e) Location E.

cases appear to coincide for z / zb 1 . Note that the typical height for a building-mounted wind turbine falls in the range
z / zb 0.2 . The flat-ground simulations give higher wind speeds (see Figure 14(a) and (c)) for z / zb 1 , indicating that
it is not always the case that the inclusion of terrain in simulations results in reduced velocities.
Mohamed and Wood 235

Figure 10. Distribution of Pw over the campus for the actual terrain and flat ground. Wind is simulated at the inlet from the
west, north-west and south-west. The height of the plane is 4 m above the roof of the EEEL building. Refer to compass for wind
directions. Units for Pw are in W/m2. (a) Wind from west: actual terrain. (b) Wind from west: flat ground. (c) Wind from north-
west: actual terrain. (d) Wind from north-west: flat ground. (e) Wind from south-west: actual terrain. (f) Wind from south-west: flat
ground.

The vertical profiles for k shown in Figure 13(d) to (f) show more turbulence for the flat ground case near the roof. In
general, more turbulence occurs when the wind is from the west. For the flat ground case, the profile of k relaxes toward
the inlet value of k / U e2 = 0.012 from equation (2) slower when the wind is from the south-west than when it is from the
west or north-west. This could be due to interference from the buildings upstream for the south-west case.
236 Wind Engineering 40(3)

Figure 11. A cut-plane (highlighted in gray) taken 4 m above the roof of the EEEL building for the analysis of area integral of Pw .

Table 1. Area integral of Pw comparisons between flat ground and terrain calculations 4 m above the roof of EEEL based on the
cut plane shown in Figure 11.

Wind direction and terrain type Pw (W)

*West, flat ground 366, 874


*West, actual terrain 213, 869
*North-west, flat ground 666, 667
*North-west, actual terrain 641, 877
*South-west, flat ground 266, 695
*South-west, actual terrain 234, 990

Figure 12. Projection of a line from the roof of the EEEL building to assist in analyzing the velocity and turbulent profiles.

The plots for Pw profiles along Projection A are shown in Figure 14. The west and south-west profiles have the same
trend for flat and actual terrain but there appears to be a shift in magnitudes.

Details of the flow over the roof of the designated building


When the wind hits the edge of a building, the flow separates and high turbulence is usually generated. The flow in this
region above the roof is also characterized by a re-circulation region. It is well known that turbine performance drops in
the presence of low mean wind speed and high turbulence. Hence it important to know not only the wind energy potential
but also the regions and the size of the re-circulation zone. Figures 15, 16, and 17 depict the wind speeds at different planes
parallel to the x- and y-axes for the three cases. These planes are re-oriented in the direction of the incoming velocity at
Mohamed and Wood 237

Figure 13. Mean wind speed and turbulent kinetic energy profiles taken along Projection A for flat ground (green lines) and actual
terrain (red lines) calculations. The top row shows the mean wind speed in the direction indicated and the bottom row shows k.
Freestream velocity Ue = 5.21 m/s and the height of the building zb is 32 m. The origin is at the EEEL roof. (a) Wind from west. (b)
Wind from north-west. (c) Wind from south-west. (d) Wind from west. (e) Wind from north-west. (f) Wind from south-west (see
colour version of this figure online).

Figure 14. Pw taken along Projection A for flat ground (green lines) and actual terrain (red lines) calculations. The height of the
building, zb, is 32 m. The origin is the EEEL roof. (a) Wind from west. (b) Wind from north-west. (c) Wind from south-west (see
colour version of this figure online).
238 Wind Engineering 40(3)

Figure 15. Wind speeds over the EEEL building with streamlines depicting areas of re-circulation. Wind is from the west. Wind
speed is in m/s. (a) Plane 1. (b) Plane 2. (c) Plane 3. (d) Plane 4. (e) Plane 5.

the inlet and show the contours of velocity magnitude and streamlines around the EEEL building. The primary purpose of
the planes is to show the re-circulation zones. It is important to know the wind energy potential but also the regions and
the size of the re-circulation zone. Figures 15, 16, and 17 depict the wind speeds in different planes parallel to the x- and
y-axes for the three cases. These planes are re-oriented in the direction of the incoming velocity at the inlet and show the
contours of velocity magnitude and streamlines around the EEEL building. The primary purpose of the planes is to show
the re-circulation zones.
As shown in Figures 15, 16 and 17, the wind speed increased (for the three cases) over the western part of the roof of
the EEEL building. This region of increased speed is also devoid of re-circulation zones: the two characteristics which
are advantageous for a wind turbine. Note however that wind turbines with sufficiently large rotor diameters would not
perform well in this region (Van~Bussel and Mertens, 2005) because the large blades will traverse the re-circulation zone
below the accelerated flow. At this stage, this region near the west edge of the EEEL building is marked as a potential area
of good wind resource. The vector plot for one of the planes is shown in Figure 18.
Figure 19 depicts the re-circulation zone above the EEEL building. It can be inferred from the velocity vectors that if
a turbine were to be placed further from the edge of the building, then special attention would need to be paid to its rotor
diameter to avoid being immersed in the re-circulation region. As an indicator of the size of the region, Figure 19 shows
its depth (3.7 m) at one location.

Turbulence
Most wind turbine power curves exclude the effects of turbulence (Lubitz, 2014). It cannot be generalized that high tur-
bulence always degrades the power output of a wind turbine, however, as increased turbulence means more power in the
wind that is potentially exploitable. However, turbulence and hence gust induces sudden changes in the wind direction
which is undesirable in harnessing wind energy.
The distribution of k for the campus is shown in Figure 20. As before, the k contours are in the horizontal plane 4 m
above the roof of the EEEL building. The turbulence levels above the west side of the EEEL building, deemed as a poten-
tially good site, are low compared to the other areas within the concentrated perimeter.
Mohamed and Wood 239

Figure 16. Wind speed over the EEEL building with streamlines depicting areas of re-circulation. Wind is from the north-west.
Wind speed is in m/s. (a) Plane 1. (b) Plane 2. (c) Plane 3. (d) Plane 4. (e) Plane 5. (f) Plane 6. (g) Plane 7. (h) Plane 8. (i) Plane 9.

Effect of trees
As shown by Mohamed and Wood (2015b),trees can distort the flow over buildings as the increased turbulence and veloc-
ity defect are convected upwards, even for buildings of heights comparable to the EEEL building. We now analyze their
effect on the flow over the EEEL building. The modeling of trees was based on equations (1) to (5) from Mohamed and
Wood (2015b). Each tree was modeled as a fluid medium with a sink term introduced into the momentum equation and
source terms in the transport of k and . The k- turbulence model was used. The tree simulations were done for a flat
terrain with prismatic layers applied to the cells adjacent to the walls. The positions of the trees were obtained through
the vegetative canopy data supplied by the Spatial & Numeric Data Services from the Taylor Family Digital Library; see
Figure 21. The mesh with trees is shown in Figure 22(a) with 22(b) displaying the prismatic layers on the buildings and
trees. All trees were assumed to be 18 m high which is the approximate height of an array of trees west of the EEEL build-
ing, and identical, to simplify the simulation.
The simulations followed the previous boundary conditions with the wind from the west. The velocity and k profiles
were plotted along Line B (See Figure 23). The line was projected in a region of high wind power density.
The profiles for velocity and turbulent kinetic energy were plotted along Line B defined in Figure 23 and they are
depicted in Figures 24 and 25.
240 Wind Engineering 40(3)

Figure 17. Wind speeds over the EEEL building with streamlines depicting areas of re-circulation. Wind is from the south-west. (a)
Plane 1. (b) Plane 2. (c) Plane 3. (d) Plane 4. (e) Plane 5. (f) Plane 6. (g) Plane 7. (h) Plane 8. (i) Plane 9.

Figure 18. Vector plots showing the wind from west.

As shown, trees play a significant role in determining the wind speed over the six-storey EEEL building. The wind
speed with trees does not exhibit a region of negative U / z and is considerably larger than the treeless case close
to the roof. The difference in velocity profiles spans a region between z / zb 0.1 and z / zb 1. There is an increase in
velocity below z / zb 0.1 (the first 3 m from the roof) when trees are included in the simulations. This is an important
Mohamed and Wood 241

Figure 19. A closer look at the vectors on the west side of the EEEL rooftop.

Figure 20. Turbulent kinetic energy distribution for the three cases. The units for k are m2/s2. (a) Wind from west. (b) Wind from
west, zoomed in. (c) Wind from north-west. (d) Wind from north-west, zoomed in. (e) Wind from south-west. (f) Wind from
south-west, zoomed in.
242 Wind Engineering 40(3)

Figure 21. CAD representation of the vegetative canopy at the University of Calgary.

Figure 22. Computational mesh used for calculating wind flow over the campus with modeled trees. (a) Computational grid with
trees. (b) Zoomed-in view of grid.

Figure 23. Velocity and turbulent kinetic energy profiles taken along a line at the top of the EEEL building. The wind direction is
along the x-axis.

result because a wind turbine would most likely be installed in this range. This confirms the earlier finding about the
effects of trees on the wind flow around the roof of buildings when the tree height is smaller than, but comparable to,
the building height.
The turbulent kinetic energy profiles between the two cases show significant differences. Figure 25(a) shows a reduc-
tion of k in the range z / zb > 0.1. The energy budget in Figure 26 shows that the advection of k is greater in the presence
of trees. The extra k generated by the trees is convected over the roof. The production term is larger for the case with trees
near the roof but becomes smaller with no trees in the region z / zb = 0.01 0.1 .
Mohamed and Wood 243

Figure 24. Effects of trees on the velocity profiles taken at the top of the EEEL building along Line B in Figure 23. Freestream
velocity Ue is 5.21 m/s and building height zb is 32 m. The origin is at the roof of the EEEL building. (a) U profile in the range
z / zb = 0 2 . (b) U profile very near the roof.

Figure 25. Effects of trees on the k profiles taken at the top of the EEEL building along Line B in Figure 23. Freestream velocity Ue
is 5.21 m/s and building height zb is 32 m. The origin is the roof of the EEEL building. (a) k profile in the range z / zb = 0 2 . (b) k
profile very near the roof.

Figure 26. Turbulent kinetic energy budget at the top of EEEL along Line B. The acronyms WT and NT stand for with trees and
no trees respectively.
244 Wind Engineering 40(3)

Figure 27. Wind power density per unit area distribution for different turbulence models. The wind is from the west. (a) SST k-.
(b) RNG k-. (c) BSL RSM.

The same trend can seen in the dissipation of k. The production term is bigger for z / zb < 0.1 without trees and k is
bigger in that region as well when compared to the case with trees. Note that k only depends on the normal stresses. The
x,y-plane Reynolds shear stress gives the main production term in the presence of vertical shear. This explains why there
is an increase in k with trees for z / zb < 0.1 in Figure 24(b).

Turbulence models
One of the uncertainties in CFD is the accuracy of turbulence models. For wind resource assessment in general, the stand-
ard k - turbulence model remains a popular choice, but as discussed by Mohamed and Wood (2015a) and the references
therein, it can seriously overestimate k in built-up environments with many stagnation regions. There have been some
recommendations of the use of other models for urban wind flow.
Wright and Easom (1999) recommended the renormalization group (RNG) k- for building pressures while Blocken
etal. (2012) suggested the realizable k- for assessing wind comfort and safety for pedestrians. Franke etal. (2011)
proposed using Reynolds stress models (RSM) for the aforementioned applications. The next part of the study assesses
the performance of the SST k-, baseline (BSL) RSM and the RNG k- turbulence models. The focus is the EEEL build-
ing. The wind power density for the three turbulence models taken 4 m above the roof of the EEEL building is shown
in Figure 27.
Mohamed and Wood 245

Table 2. Area integral of Pw comparisons between different turbulence models 4 m above the roof of EEEL based on the cut plane
shown in Figure 11.

Turbulence model Pw (W)


*SST k- 213, 869
*RNG k- 299, 143
*BSL RSM 270, 418

Figure 28. k distribution for different turbulence models. The simulations were simulated with wind coming from the west at the
inlet. (a) SST k-. (b) RNG k-. (c) BSL RSM.

Although there are differences in the distributions of wind power density, all models suggest that the maximum wind
resource is located above the west facade of the EEEL building. Further, the three models would suggest similar regions in
which to measure the wind speed for confirmation. Table 2 shows the area integral for Pw for different turbulence models
taken from the plane shown in Figure 11. There is less agreement in turbulence with the RNG k- turbulence models on
the north-west side of the building.
The U / U e profiles along Projection A (see Figure 12) show that the SST k-, RNG k-, and RSM models make similar
predictions for wind speed barring a small extended re-circulation zone near the wall predicted by the RSM model. The
SST shows very little overshoot in k near the roof compared to the other two models with the RNG k- predicting high
levels of turbulence compared to the other two models.
246 Wind Engineering 40(3)

Figure 29. U and k distribution for different turbulence models taken along Projection A. The simulations were simulated with
wind coming from the west at the inlet. Freestream velocity Ue = 5.21 m/s and the height of the building zb is 32 m. The origin is
at the roof of the EEEL building. (a) Velocity profile. (b) k profile.

Wind turbine and PV module siting


Microgeneration technologies that benefit from wind resource assessment include not only wind turbines but also PV
modules, the latter because of sensitivity to wind loads due to their light weight and large frontal area. In short, PV requires
low wind speed while wind turbines need high wind speed.
The above analyses suggest that the potential wind turbine siting zone should be located along the west side of the
EEEL building. Due to the inherent nature of accelerated and hence skewed flow near the edge of the west roof of the
EEEL building, small turbines with a characteristic size of Dt / zb > O(0.1) (Mertens, 2006), where Dt is the rotor diam-
eter, should be used.
To obtain maximum power, PV modules are normally inclined at an angle slightly less than the latitude of the site.
Although they are susceptible to wind loads, there are not many guidelines or codes to assist designers with this aspect.
Most building codes, such as the Canadian National Building Code, NBC (2010), require dynamic pressure in order to
determine the extreme wind load. Figure 30 shows the dynamic pressure qh distribution over the Schulich School of
Engineering 2 m above the roof of Block D for winds coming in from the west compared to when the wind is coming from
the other two directions. This could be due to the roof of Block D being sheltered by the adjacent building to the west.
From the three cases, the maximum qh above the roof of Block D is 18.37 Pa.
On the 27 November 2011, Calgary was hit by hurricane-strength winds with wind speeds recorded at 149 km/h. Such
high wind speeds can destroy PV modules. The next part of the study is focused on extreme wind speeds. This case used the
149 km/h wind speed as the inlet condition. The SST simulations followed the previous studies of westerly winds with u
increased to 2.32 m/s to model the quoted value. The results from the extreme wind simulations are depicted in Figure 31.
Although the roof of Block D is sheltered by the Calgary Centre for Information Technology (CCIT) building in the
middle-left of Figure 31 there is a significant increase in qh . The magnification is almost 31 times that due to a mean wind
speed of 5.21 m/s at 50 m height.
To see how the results change relative to the change in dynamic pressure, we normalized the dynamic pressure for both
cases of extreme wind speed and annual wind speed with the qh based on respective velocities taken at 50 m at the inlet,
that is, 41.4 m/s for the extreme wind speed case and 5.21 m/s for the annual wind speed case. As shown in Figure 32,
there are more regions of low normalized qh (range between 0 and 0.1) for the extreme wind speed case over Block D
than for the annual wind speed case. However, the bulk flow pattern is similar for both cases which highlights the small
Reynolds number effect on the flow.

Conclusions
An analysis was done for the siting of wind turbines on the roof of the EEEL building of the University of Calgary campus
via CFD. The process included several considerations including the effects of terrain, turbulence models, trees, and wind
directions. The exclusion of terrain led to significantly higher estimates for wind power density and implies that terrain
Mohamed and Wood 247

Figure 30. Contours of dynamic pressure, qh . Plane taken 2 m from roof of the Schulich Engineering Building Block D. Units are
in Pa. The EEEL building is below the axes graphic. (a) Wind from the west. (b) Wind from north-west. (c) Wind from south-west.

Figure 31. Contours of dynamic pressure, qh , for an extreme wind speed case. Plane taken 2 m from roof of the Schulich
Engineering Building Block D. Units are in Pa.

has to be included in computational wind resource assessment for urban settings. Different turbulence models tended to
predict differences in wind power density especially above the EEEL building but they apparently give a fair indication of
what range can be expected. Trees also caused deviations in wind speed and turbulence above the roof of the 32 m high
248 Wind Engineering 40(3)

Figure 32. Contours of normalized dynamic pressure, qh, for an extreme wind speed case and annual wind speed case. Plane
taken 2 m from roof of the Schulich Engineering Building Block D. (a) Annual wind speed case. (b) Extreme wind speed case.

EEEL building. It is recommended that any computational assessment of the wind resource include them. All results were
carefully considered before choosing the potential location which eventually narrowed down to the edge of the EEEL
building on its west side. This study also looked at the wind loads on a possible PV modules installation on the roof of
Block D, Schulich School of Engineering, and considered extreme wind conditions. Dynamic pressure can be used to
calculate the extreme wind loads on the modules and it was found that under extreme wind conditions the magnification
was 31 times over the loads due to the annual average wind speed.

Declaration of conflicting interests


The authors declare that there were no conflicts of interest in this study.

Funding
This research was supported by the Natural Science and Engineering Research Council and the ENMAX Corporation under the
Industrial Research Chairs programme.

References
Barth TJ and Jespersen DC (1989) The design and application of upwind schemes on unstructured meshes. AIAA Paper 89-0366.
Blocken B, Janssen W and van Hooff T (2012) CFD simulation for pedestrian wind comfort and wind safety in urban areas: General
decision framework and case study for the Eindhoven university campus. Environmental Modelling & Software 30: 1534.
Encraft (2008) Warwick wind trials. Fourth interim report. Available at: http://www.warwickwindtrials.org.uk/resources. Accessed 18
May 2015.
Franke J, Hellsten A, Schlnzen H and Carissimo B (2011) The COST 732 best practice guideline for the CFD simulation of flows
in the urban environment: A summary. International Journal of Environment and Pollution 44(14): 419427.
Franke J, Hirsch C, Jensen A, etal. (2004) Recommendations on the use of CFD in wind engineering. In: Cost action C, volume 14.
p. C1. Available at: http://www.kuleuven.be/bwf/projects/annex41/protected/data/Recommendations%20for%20CFD%20in%20
wind%20engineering.pdf (accessed 18 May 2015).
Lubitz WD (2014) Impact of ambient turbulence on performance of a small wind turbine. Renewable Energy 61: 6973.
Mertens S (2006) Wind energy in the built environment: Concentrator effects of building. PhD Thesis, Delft University of Technology,
Delft, the Netherlands.
Mohamed M and Wood D (2013) Vertical wind speed extrapolation using the k- turbulence model. Wind Engineering 37(1): 1336.
Mohamed M and Wood D (2015a) Modifications to Reynolds-averaged NavierStokes turbulence models for the wind flow
over buildings. International Journal of Sustainable Energy. Epub online ahead of print 19 February 2015. DOI: 10.1080/
14786451.2015.1014903.
Mohamed MA and Wood DH (2015b) Computational study of the effect of trees on wind flow over a building. Renewables: Wind,
Water, and Solar 2(1): 18.
Roache PJ (1997) Quantification of uncertainty in computational fluid dynamics. Annual Review of Fluid Mechanics 29(1): 123160.
Roache PJ, Ghia KN and White FM (1986) Editorial policy statement on the control of numerical accuracy. Journal of Fluids
Engineering 108(1): 22.
Mohamed and Wood 249

Romani D, uri M, Jovii I, etal. (2015) Long-term trends of the Koshava wind during the period 19492010. International
Journal of Climatology 35(2): 288302.
Stathopoulos T, Zisis I and Xypnitou E (2014) Local and overall wind pressure and force coefficients for solar panels. Journal of Wind
Engineering and Industrial Aerodynamics 125: 195206.
Tabrizi AB, Whale J, Lyons T, etal. (2014) Performance and safety of rooftop wind turbines: Use of CFD to gain insight into inflow
conditions. Renewable Energy 67: 242251.
Tarokh A, Mohamed MA, Wood D, etal. (2014) Investigation of different turbulence models for wind flow over a simulated building.
In: Computational wind engineering conference, Hamburg, Germany.
Van Bussel G and Mertens S (2005) Small wind turbines for the built environment. In: Proceedings of the 4th European and African
conference on wind engineering.
Wright N and Easom G (1999) Comparison of several computational turbulence models with full-scale measurements of flow around
a building. Wind and Structures, an International Journal 2(4): 305323.

You might also like