See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/255657650

Aerodynamic Loads on Solar Panels

Conference Paper · April 2013

DOI: 10.1061/9780784412848.137

CITATIONS READS

0 707

1 author:

Aly Mousaad Aly

Louisiana State University
72 PUBLICATIONS   667 CITATIONS   

SEE PROFILE

Some of the authors of this publication are also working on these related projects:

Risk, Reliability, and Uncertainty Quantification of Structural Systems Subjected to Shock and Vibration View project

Wind tunnel testing/procedure View project

All content following this page was uploaded by Aly Mousaad Aly on 19 September 2018.

The user has requested enhancement of the downloaded file.

 Solar Energy 135 (2016) 423–434

Contents lists available at ScienceDirect

Solar Energy
journal homepage: www.elsevier.com/locate/solener

On the evaluation of wind loads on solar panels: The scale issue

Aly Mousaad Aly ⇑
Louisiana State University, Baton Rouge, LA 70803, USA

a r t i c l e i n f o a b s t r a c t

Article history: Solar power can improve the quality of life and reduce dependency on traditional energies that are a sig-
Received 28 January 2015 nificant source of pollution and global warming. Solar panels are common devices used for collecting
Received in revised form 4 June 2016 solar energy. To balance between sustainability and resilience, it is essential to provide an accurate esti-
Accepted 7 June 2016
mate of the design wind loads for the solar panels. Traditionally design wind loads for buildings and other
Available online 15 June 2016
structures are obtained using building codes and standards. The solar panels represent a relatively recent
technology and indeed there is no complete guidance ready for codification of wind loads on these types
Keywords:
of structures. Available wind tunnel data show discrepancies in wind loads on solar panels, owing to
Solar energy
Solar panels
inconsistent model scales and test flows, among other factors. To eliminate such discrepancies in the test
CFD results and to allow for accurate wind load estimation, the current paper investigates the geometric scale
LES and the inflow turbulence characteristics as potential causes of high uncertainties. Computational fluid
Multi-scale dynamics (CFD) simulations are employed and results are compared with available wind tunnel data,
Wind tunnel testing as a complementary tool with a potential to simulate wind loads at full-scale. The results show that
Turbulence the geometric scale is a primary reason for the discrepancies in peak wind loads, which can be avoided
Wind loads by adapting the inflow turbulence and using a proper testing protocol. The results show an evidence of
Peak pressures
the correctness of a hypothesis that the lack of large-scale turbulence can dramatically affect peak wind
Wind velocity spectra
loads on test objects. Consequently, recommendations are articulated regarding the best usage of the
available wind load estimation tools. This is expected to lead to consistent and accurate results from wind
tunnel testing and CFD simulations, a crucial step toward codification of wind loads on solar panels.
Ó 2016 Elsevier Ltd. All rights reserved.

1. Introduction
1.1. Background
The importance of the solar power as a source of eco-friendly
energy was documented early in 1911 (Shuman, 1911). With
worldwide concerns regarding the impact that combustible fuels
have on the increase in greenhouse gas emissions and climate
change, sustainable development policies supporting the integra-
tion of renewable energies have started. Photovoltaic (PV) or solar
panel systems are common devices used for collecting solar energy
(Singh, 2013). Probably technology will lead to ‘Covering the Planet
with Solar Panels’ (Webb, 2007). For large-scale solar energy pro-
duction, PV panels are usually mounted on ground. The main con-
cern regarding solar panels installation is their vulnerability to
highly turbulent and gusty wind. Developing design standards is
still in an incipient stage, and building guidelines are not clearly
defined. Indeed there are no code guidelines available yet, that
⇑ Address: Department of Civil and Environmental Engineering, Louisiana State
University, 3316T Patrick Frank Taylor Hall, Baton Rouge, LA 70803, USA.
E-mail address: aly@LSU.edu
can support the evaluation of design wind loads for solar panels
(Stathopoulos et al., 2012). This means that the designer may
roughly estimate the loads with the consequences of either overes-
timation, which put obstacles on a technology that is supposed to
be economic, or underestimation, which may lead to failure. For
both scenarios the design is not feasible. It is worthy to mention
that in the past decades, the solar cell industry has grown signifi-
cantly, however, because of its relatively high cost, solar power
accounts for a little portion of the U.S. electricity (Robert, 2003).
According to Morton (2006), sunlight is a ubiquitous form of
energy, but not as yet an economic one. This highlights the impor-
tance of building wind resistant solar panels in an economic way,
which can be achieved by accurate estimation of the design loads.
For large structures, e.g. high-rise buildings, wind loads are easier
to be predicted by wind tunnel testing (at geometric scales ranging
from 1:100 to 1:500) as the flow depicts all necessary turbulence
content; however, for small structures, for example solar panels,
testing in wind tunnels can be challenging. Aerodynamic testing
of small structures, for instance ground-mounted solar panels, is
typically carried out at scales larger than those usually used in
boundary-layer wind tunnels (e.g., 1:30 scale). This is to alleviate
potential technical problems; especially those related to pressure

http://dx.doi.org/10.1016/j.solener.2016.06.018
0038-092X/Ó 2016 Elsevier Ltd. All rights reserved.
424 A.M. Aly / Solar Energy 135 (2016) 423–434
Nomenclature
A tributary area
an, bn, cn, an Fourier coefficients
CNF normal force coefficient
CNF,peak peak force coefficient
CP(t) time history of pressure coefficient
CP,bottom(t) pressure coefficient (bottom surface)
CP,tot(t) total pressure coefficient
CP,top(t) pressure coefficient (top surface)
cE turbulent kinetic energy constant
E(j) energy of wave number j
f(x) synthesized inlet fluctuations
k turbulent kinetic energy
P(t) instantaneous surface pressure
U mean wind speed
ustd root-mean square speed
v 01 , v 02 , v 03 inlet fluctuating velocity fields
resolution, interference effects from measuring instruments and
falls in the uncertain flow regimes close to the tunnel floor. The
problem is that, at such geometric scales, the flow lacks essential
turbulence content.
1.2. Review
In order to obtain realistic wind loads by wind tunnel testing,
fundamental laws of similitude should be employed. These are
mainly the similarity of the test model and the approaching flow
to full-scale (Plate, 1982). Wind loads on solar panels depend on
wind speed, terrain characteristics, shape, which includes inclina-
tion angle, and installation type, among other factors. The majority
of solar panels installations are either roof-mounted or ground-
mounted. Although the literature has a number of studies concern-
ing wind loads on roof mounted solar panels (Radu et al., 1986;
Radu and Axinte, 1989; Wood et al., 2001; Kopp et al., 2002;
Chung et al., 2008; Meroney and Neff, 2010; Banks, 2013;
Browne et al., 2013; Cao et al., 2013; Aly and Bitsuamlak, 2014;
Stathopoulos et al., 2014; Warsido et al., 2014), a paucity of papers
address ground mounted solar panels (Bitsuamlak et al., 2010;
Kopp et al., 2012; Abiola-Ogedengbe, 2013; Aly and Bitsuamlak,
2013; Stathopoulos et al., 2014; Warsido et al., 2014; Shademan
et al., 2014a,b; Jubayer and Hangan, 2014). Pfahl et al. (2011) pre-
sents wind loads on heliostats and photovoltaic trackers of various
aspect ratios, with a model scale of 1:20. The study suggests that
higher aspect ratios are advantageous for the dimensioning of
the foundation, and the pylon and the elevation drive but disad-
vantageous for the azimuth drive. Blackmon (2014) investigates
the effects of wind load on projected fatigue life and safety factor
for heliostat drive unit. The study shows that having heliostats
fully stowed at bottom wind speeds than legacy specifications of
50 mph increase life and reduce safety factors required.
Complimentary to the wind tunnel studies (Kopp et al., 2002,
2012; Aly and Bitsuamlak, 2012, 2013), a limited computational
fluid dynamics (CFD) studies on wind loads for the solar panels
are reported in the literature. Shademan and Hangan (2009)
carried out CFD simulations to estimate the wind loads on stand-
alone and arrayed solar panels. They observed that at a specific
distance between two sets of panels the drag coefficient for the
downstream panels reaches a minimum. Shademan et al. (2014a,
b) carried out Detached Eddy Simulations (DES) to analyze the
influence of ground clearance on solar panels. The study shows
that an increase in the clearance produces stronger vortex
shedding fluctuations, and larger mean and unsteady wind loading.
v0 ðxÞ turbulent velocity field
W number of windows
bn velocity field calculation figure
e dissipation rate
je unscaled wave number
jg Kolmogorov wave number
jn wave number
^n
u amplitude of Fourier mode
q air density
rn direction of Fourier mode
t molecular viscosity
wn phase of Fourier mode
N number of samples per window
x spatial coordinate
n Fourier mode
However, at a relatively small clearance, the vortex shedding
effects are minimized which suggests a reference value for
minimizing the unsteady wind loading on ground-mounted solar
panels. Bitsuamlak et al. (2010) carried out full-scale testing and
CFD simulations on stand-alone solar panels. The results show
similar patterns of pressure coefficient distribution when
compared to full-scale measurements, but the magnitude of
the pressure coefficients was generally underestimated by the
numerical calculations when compared to the experimental
results. Aly and Bitsuamlak (2013) presented wind tunnel test data
along with CFD simulation results on ground-mounted solar panels
at different scales and under different wind flow characteristics.
They concluded that the mean wind loads on the solar panels
obtained from CFD simulations are in a good agreement with the
wind tunnel results at different geometric scales.
Warsido et al. (2014) carried out a boundary-layer wind tunnel
tests to investigate the effects of different spacing parameters on
the wind loading of ground-mounted solar arrays. The study shows
that the magnitude of force and moment coefficients on the ground
mounted array decreased across panel rows as a result of the shel-
tering effect from the neighboring upwind panels. Meroney and
Neff (2010) presented the results of a hybrid numerical and phys-
ical model program to predict wind loads on PV collector arrays.
The hybrid numerical and physical model was developed to per-
form rapid and cost-effective comparisons of alternate PV design
configurations. Some of the advantages of the CFD simulations
include generation of continuous wind load information over the
structural elements as well as whole-flow field at the possibility
of executing full-scale simulation studies without geometric scale
constraints. However, one primary shortcoming of a CFD based
approach is being computationally costly to predict peak loads
on structures under turbulent flows. When a relatively accurate
turbulence closure, for example Large Eddy Simulations (LES), is
incorporated into the numerical modeling at high Reynolds num-
bers, CFD simulations require a very high performance computing
capabilities. This coupled with the costly commercial CFD license
for parallel computing is limiting CFD for design wind load evalu-
ation in practice. Wind tunnel experiments still remain an eco-
nomic choice compared to CFD simulations, for wind load
applications. However, concurrent CFD studies provide additional
opportunities to explain/augment wind tunnel studies. In any case,
Stathopoulos et al. (2012) carried out a review on wind loads on
solar collectors, in which they conclude that results show signifi-
cant differences among different studies, some of which corre-
spond to similar configurations. The discrepancies in the results
 A.M. Aly / Solar Energy 135 (2016) 423–434
are hindering an attempt toward codification. The current paper
seeks reasons and puts practical solutions to these discrepancies.
1.3. Lack of hybrid multi-scale experimental/computational testing
methods
Capabilities of existing hurricane testing facilities do not allow
full- or large-scale turbulence simulations and testing of effects
of coupled stressors (Aly, 2014). Boundary-layer wind tunnel
(BLWT) testing of structures is an industry wide accepted proce-
dure and is considered the main source of information for wind
load calculations and codification. However, the majority of
boundary-layer wind tunnels are built for testing large structure
models with geometric scales ranging between 1:500 and 1:100.
Accordingly, producing representative aerodynamic models of
the solar panels at such scales makes the models too small, result-
ing in technical problems: resolution of pressure data on such
small models becomes low, and the uncertainty in the flow
measured speed will be very high. To alleviate these problems,
development of a standardized and practical modeling hybrid
multi-scale experimental/computational procedure that accounts
for different time and geometric scales is important. Scale
effect studies are required for the solar panels as they have
relatively small sizes. CFD simulations with LES require
appropriate inflow simulation to permit testing full-scale solar
panels. The current study is trying to overcome this barrier by
integrating available testing facilities and carrying out hybrid
experimental/computational research that is scalable and will
complement concurrent wind load evaluation studies on solar panels.
1.4. Focus of the current study
The current paper presents CFD investigations of ground-
mounted solar panels, to understand the model scale effects on
the pressure distribution. CFD studies require experimental valida-
tion as the error of the computational analysis can be defined with
a reference to a full-scale or a wind tunnel-study. Due to scarcity of
full-scale data, CFD studies carried out on full-scale objects are
usually validated by small-scale wind tunnel studies. This may lead
to inconsistent physics related to boundary conditions, geometric
scales and test flows. To alleviate these issues, a virtual boundary
layer wind tunnel is modeled in CFD, by exactly simulating all floor
roughness elements in the physical test section. The virtual wind
tunnel (CFD) has same boundary conditions as the physical one
(experiment). This allows for ‘apple to apple’ comparison between
the CFD and the experimental results. In addition, the current
study investigates aerodynamic models representative of a
ground-mounted solar panel at different scales, tested in different
flows, to examine a hypothesis that the lack of large-scale turbu-
lence can dramatically affect peak wind loads on test objects. The
results presented validate this hypothesis. Finally, recommenda-
tions are formulated regarding the best use of the available wind
load estimation tools.
2. Methodology
2.1. CFD simulations
A physical wind tunnel was geometrically modeled with a scale
of 1:1; all the roughness features physically exist in the tunnel
were modeled and presented in a virtual wind tunnel. This repre-
sents the computational domain for the scaled CFD simulations
(see Figs. 1 and 2). The roughness features including spires and
ground blocks were presented in the computational domain
(Fig. 2) to simulate an open country atmospheric boundary layer,
425
Fig. 1. 3D geometric model of the virtual wind tunnel used in the current study:
dimensions are 26 m length, 2.4 m width and height ranging from 1.55 m (at spires)
to 2.15 m (at the turn table).
proposed for testing the ground-mounted solar panels. The virtual
tunnel has a length of approximately 26 m, a width of 2.4 m and a
varying height from the inlet to the exit (1.55–2.15 m). Different
scales of a ground-mounted solar panel were introduced on the
turntable of the virtual tunnel for CFD simulations (Fig. 3). The geo-
metric scales used in the virtual tunnel are 1:50, 1:20 and 1:10. An
additional 3-D computational domain (CD) following AIJ
(Tominaga et al., 2008), and COST (Franke et al., 2007) guidelines
was also considered to allow CFD simulations on a solar panel with
scale 1:1. The top boundary of the 3-D CD was set at 7.5H (H is the
maximum height of the full-scale ground mounted solar panel).
Lateral boundaries were set at a distance of 3.3W (W is the cross
wind dimension of the solar panel). This gives a blockage ratio of
1.75%, which is less than the 3% mentioned in Franke et al.
(2007). The inflow boundary was set at a distance of 6.8H, and
the outflow boundary was at 13.6H downstream of the panel to
permit wake flow redevelopment. The solar panel at full-scale is
similar to a rectangular plate with 1.336 m  9.144 m and a thick-
ness of 0.11 m. The arrangement of the module is: tilt angle (a) of
40° and leg height (Hleg) of 24 in. (0.61 m) at full-scale. The geom-
etry modeled in AutoCAD was exported to SolidWorks (to create
surfaces and to examine the geometry, i.e., to ensure that the ‘air
tightness’ is achieved), and finally exported as an IGES file readable
by a meshing software. Once the mesh has been created it was
exported as unstructured mesh (.msh) file that can be read by a
CFD solver. To allow for a grid independence study, the computa-
tional domain and the domain immediately around the solar panel
were modeled with different grid sizes (small, medium and large)
as shown in Fig. 3.
2.1.1. Mean wind loads
Mean wind loads on three different scaled models (1:50, 1:20
and 1:10) of a ground-mounted solar panel were obtained by run-
ning CFD simulations in the virtual wind tunnel. In addition, mean
wind loads on a panel scaled 1:1 were obtained using the 3-D CD
described earlier. The solver used is the Reynolds Stress Model
(RSM). In RSM, the eddy viscosity approach has been discarded
and the individual Reynolds stresses are directly computed, using
differential transport equations. The individual Reynolds stresses
are then used to obtain closure of the Reynolds-averaged momen-
tum equation.
2.1.2. Peak wind loads
In order to capture the peak wind loads on the solar panels, the
LES turbulence closure was used. LES for two different scales (i.e.
1:1 and 1:10) were conducted. For peak loads on a solar panel
scaled 1:1, the transient inflow wind was essential. The approach
requires a generated inflow that simulates the natural wind with
both time and space correlations. The method discussed in
Davidson (2007, 2008) was followed in this study to generate the
inflow characteristics in a time history. Synthesized inlet fluctua-
tions were generated using a Fourier series as follows:
426 A.M. Aly / Solar Energy 135 (2016) 423–434

Fig. 2. Minimum mesh for a solar panel located in the virtual wind tunnel.

(a) (b) (c)

Fig. 3. Meshes used in the grid independence study: (a) small, (b) medium and (c) large.

X
1
A general form for a turbulent velocity field can thus be
f ðxÞ ¼ U þ ðan cosðnxÞ þ bn sinðnxÞÞ
written as
n¼1
X
1 X
N
¼cþ cn cosðnx  an Þ ð1Þ v0 ðxÞ ¼ 2 ^ n cosðjn  x þ wn Þrn
u ð6Þ
n¼1 n¼1

in which U is a constant that corresponds to the mean wind speed at where u^ n ; wn and rn are the amplitude, phase and direction of the
a certain height, and x is a spacial coordinate. The Fourier coeffi- Fourier mode n. The synthesized turbulence at one time step can
cients are given by be generated once these parameters are estimated (Davidson,
Z p 2011). A modified von Karman spectrum was considered and the
1
an ¼ f ðxÞ cosðnxÞdx ð2Þ amplitude of each mode in Eq. (6) is obtained from
p p pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
^n ¼
u EðjÞDj ð7Þ
Z p
1
bn ¼ f ðxÞ sinðnxÞdx ð3Þ u2std ðj=je Þ4
p p EðjÞ ¼ cE  17=6 e
2ðj=jg Þ2
ð8Þ
je 1 þ ðj=je Þ 2
qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
2
cn ¼ a2n þ bn ð4Þ
pffiffiffiffiffiffiffiffiffi
  j ¼ ji ji ð9Þ
b
an ¼ tan1 n : ð5Þ
an jg ¼ e0:25 t0:75 ð10Þ
 A.M. Aly / Solar Energy 135 (2016) 423–434
where ûn is the amplitude of Fourier mode n, E(j) is the energy of
wave number j, cE is the turbulence kinetic energy constant, ustd
is the standard deviation of the fluctuating velocity, jg is the Kol-
mogorov wave number, e is the dissipation rate (energy per unit
time and unit mass), and t is the molecular viscosity. The constant
cE is obtained by integrating the energy spectrum over all
wavenumbers to get the turbulent kinetic energy, which yield the
value of cE to be about 1.453. The expression in Eq. (6) can be com-
puted to obtain the inlet fluctuating velocity fields (v1, v2, v3)
X
N
v 0
1 ¼2 u^ n cosðbn Þr1 ð11Þ
n¼1
X
N
v 02 ¼ 2 ^ n cosðbn Þr2
u ð12Þ
n¼1
X
N
v 03 ¼ 2 ^ n cosðbn Þr3
u ð13Þ
n¼1
and
n n n
bn ¼ k1 x1 þ k2 x2 þ k3 x3 þ wn ð14Þ
rn represents the direction of Fourier mode n, bn is the velocity field
calculation figure, k is the turbulent kinetic energy, wn is the phase
of Fourier mode n. Following the generation of fluctuating velocity
components, time correlations were introduced and with anisotro-
pic synthetic turbulent fluctuations (Davidson, 2011; Huang et al.,
2010; Zhang et al., 2015). Once the input wind is defined, the gen-
erated time series are introduced to the CFD solver through a user
defined function that permits reading the input wind velocity corre-
sponding to each time step at different location points for the 1:1
scale CFD model.
2.2. Experimental validation
CFD simulations require experimental validation as the error of
the computational analysis can be defined with a reference to a
full-scale or a wind tunnel study. The wind tunnel has the clean
and fully developed turbulence but it can be challenging to simu-
late wind flow at large-scales. Four sizes of the ground-mounted
solar panel used in the current study were tested in a physical
boundary-layer wind tunnel with scales 1:50, 1:30, 1:20, and
1:10 (Aly and Bitsuamlak, 2013). Fig. 4 shows photographs of
two instrumented model scales of a ground-mounted solar panel
in a physical boundary-layer test section. The available wind flow
entailed to simulate the atmospheric flow over open terrain expo-
sures and designed for testing models of length scales of about
1:400. Such scales are widely used in many boundary-layer wind
tunnel testing facilities. Wind pressure data over the top and bot-
tom surfaces of the test models were collected for time periods
corresponding to 1 h at full-scale (according to the geometric scale
and the mean wind velocity used). The data were collected at a
(a)
Fig. 4. Photograph of a solar panel model scaled 1:10 (a) and 1:50 (b, c) tested in a physical BLWT.
427
sampling rate of 400 Hz. Pressures are usually measured with
respect to the mean static pressure in the wind tunnel test section,
corresponding to the mean static pressure in full-scale wind flow
(BLWTL, 2007; Ho et al., 2005). At the location of each pressure
tap, the time history of the pressure coefficient, CP(t), is obtained
from the time history of the instantaneous surface pressure (mea-
sured with respect to the mean static reference pressure in the
wind tunnel), P(t), as
C P ðtÞ ¼ PðtÞ=ð0:5  qU 2 Þ ð15Þ
where q is the air density at the time of the test and U is the mean
wind speed measured at the mean height of the solar panel. In
designing the solar panels, it is necessary to determine the net pres-
sure on the individual modules. The net (total) pressure coefficient
at any location, CP,tot(t), is the simultaneous difference between the
pressure coefficient at the top surface, CP,top(t), and the pressure
coefficient at the bottom surface, CP,bottom(t), at the same location
C P;tot ðtÞ ¼ C P;top ðtÞ  C P;bottom ðtÞ: ð16Þ
The coefficient of the normal force (CNF) acting on a test model
is defined as follows:
!, !
X
m X
m
C NF ðtÞ ¼ C P;net;i ðtÞAi Ai ð17Þ
i¼1 i¼1
where Ai is the tributary area (Aly, 2013) of tap number i, i = 1,
2, 3 . . ., m (m represents total number of taps). The tap layout is
shown in Fig. 5.
The peak (or mean extreme over 1 h of equivalent full-scale
time) pressure and force coefficient values (CP,peak and CNF,peak)
may be obtained from the measured pressure time histories.
However, to reduce the uncertainties inherent in the randomness
of the peaks, a probabilistic analysis was performed using an
automated procedure developed by Sadek and Simiu (2002) for
obtaining statistics of the pressure peaks from the pressure
time histories. Because estimates obtained from this approach
are based on the entire information contained in the time series,
they are more stable than estimates based on observed peaks. In
the data analysis, 95 percentile peak pressure/force coefficients
were used.
3. Results
3.1. Average pressures
Simulated CFD wind speed and turbulence intensity profiles
compared with measured physical boundary layer wind tunnel
profiles for an open country terrain are shown in Fig. 6. The figure
shows that CFD mean velocity profiles (for the virtual wind tunnel
simulation case) at the entrance to the test section are matching
the data obtained from the physical experiment. In addition, the
turbulence intensity is sufficiently represented in the virtual wind
tunnel; this is attributed to the realistic representation of the
(b) (c)
428 A.M. Aly / Solar Energy 135 (2016) 423–434

Fig. 5. Tap layout for a test model scaled 1:10. Dimensions are in in. (1 in. = 2.54 cm).

Fig. 6. Simulated CFD wind speed profiles and comparison with measured BLWT profiles for open terrain: (a) mean velocity profile and (b) turbulence intensity profile. Note
that vBLWT designates virtual boundary layer wind tunnel (see Fig. 1).

roughness elements in the virtual wind tunnel, with a high quality
mesh. As a grid independence study, Fig. 7 shows mean pressure
coefficients on a centerline going around the top and bottom sur-
faces of solar panels scaled 1:50, 1:20, 1:10, and 1:1. Three differ-
ent types of meshes were used: minimum, average and maximum
intensity grids. The average and maximum intensity grids show
similar results; accordingly, the solutions from the maximum
intensity grid were considered to be grid independent.
Contour plots of net (total) mean pressure coefficients for a 1:10
scale solar panel modeled computationally and experimentally are
presented in Figs. 8 and 9, respectively. The figures show that the
mean net pressure distribution obtained by the CFD simulations
in a virtual wind tunnel is very similar to that obtained experimen-
tally. This attests that CFD simulations can be used for the estima-
tion of mean pressures on solar panels when an accurate
turbulence closure, for instance the RSM, is used. Fig. 10 shows that
the mean pressure distribution on a full-scale solar panel obtained
by the CFD simulations is very similar to the mean pressure distri-
bution obtained both computationally and experimentally on a
1:10 model (see Figs. 8 and 9).

Fig. 7. Grid independency study: mean pressure coefficients on a centerline going around the top and bottom surfaces for solar panels at different geometric scales: (a) 1:50,
(b) 1:20, (c) 1:10 and (d) 1:1.
 A.M. Aly / Solar Energy 135 (2016) 423–434
Fig. 8. CFD mean net pressure coefficients’ distribution for a panel scaled 1:10.
Fig. 9. Experimentally obtained mean net pressure coefficients’ distribution for a
panel scaled 1:10.
Fig. 10. CFD mean net pressure coefficients’ distribution for a panel scaled 1:1.
Fig. 11 shows mean pressure coefficients for a full-scale solar
panel modeled computationally along with experimental results
on different scale models. It was not possible to test the full-
scale panel in the virtual tunnel as per size constraints. However,
the 3-D CD was used and the anisotropic wind speed was modeled
at the entrance boundary to represent a typical open country pro-
file (Section 2). There is no experimental data available for the full-
scale panel; however the physical wind tunnel data obtained for
model scales 1:50, 1:20 and 1:10 are presented for comparison.
It is also shown that the CFD mean pressures are in a good agree-
ment with the results obtained experimentally. In addition, the
results show agreement between mean pressures at different
scales, which indicates the less sensitivity for a scale distortion.
Hence, the possibility of testing the solar panels in physical wind
429
Fig. 11. CFD mean pressure coefficients on a centerline going around the top and
bottom surfaces of a solar panel with physical BLWT results at different scales.
tunnels at relatively large-scales when the objective is to estimate
mean pressure values. This may lead to confidence in the measured
mean wind loads as well as better resolution of the mean
pressures.
3.2. Peak pressures
Fig. 12 shows the spectral characteristics of the wind flow gen-
erated as an inlet velocity boundary condition for both scaled and
full-scale CFD studies. The approach discussed in Davidson (2007,
2008, 2011) to generate flow fluctuations from a target spectrum
was followed (Section 2). Two types of wind profiles were synthe-
sized: the first flow (represented as ‘Gen. WT’ in Fig. 12) was gen-
erated using the characteristics of the wind flow in the physical
wind tunnel and is used for CFD simulations on scaled models.
The second flow (designated by ‘Gen. F/S’ in Fig. 12) was created
using the target characteristics of the flow in nature and was used
for full-scale CFD simulations. The reference spectra are von
Karman (Harris, 1968 as cited in Holmes, 2007) and Eurocode 1
(2004). It is shown that the wind tunnel is missing the low-
frequency turbulence (i.e., missing the large eddies required for
testing, for instance, a 1:10 scale model).
Fig. 13 shows CFD time histories of pressures on models scaled
1:1 and 1:10 (the 1:10 scale was tested in virtual wind tunnel, see
Fig. 1). The simulation time step was chosen to yield a courant
number less than 0.2 for both cases. In addition, 40 iterations per
time step were set and the mean values of pressure coefficients
were compared with RSM results (both LES and RSM mean results
were very similar to each other and to those obtained from
Fig. 12. Spectrum of the input wind flow generated for the full-scale CFD
simulations (Gen. F/S) with the physical wind tunnel spectrum (Gen. WT). The
reference spectra are von Karman (Harris, 1968 as cited in Holmes, 2007) and
Eurocode 1.
430 A.M. Aly / Solar Energy 135 (2016) 423–434

Fig. 13. CFD with LES pressure coefficient time histories for four taps located on: (a) top and (b) bottom surfaces of the 1:1 scale model; and (c) top and (d) bottom surfaces of
the 1:10 scale model. These results were obtained with a courant number less than 0.2.

physical wind tunnel testing). A contour plot of CFD (with LES)
velocity magnitude (m/s) is shown in Fig. 14. Fig. 15 shows peak
pressure coefficients obtained by CFD simulations on 1:1 and
1:10 scale models, with LES. The peak pressure coefficients
(Sadek and Simiu, 2002) derived from the physical wind tunnel
testing (experimentally obtained for a wide range of scales) are sig-
nificantly low compared to LES peak pressure coefficients (for the
full-scale model). This may be attributed to the lack of large eddies
in the physical wind tunnel. However, the CFD peak pressures
obtained on 1:10 scale model simulated in the virtual wind tunnel
are not far from those obtained from the physical wind tunnel
experiment. This means that CFD with a proper turbulence closure
like LES can predict both mean and peak pressures on the solar
panels. The problem, however, is that the physical wind tunnel is
missing the large-scale turbulence necessary for testing the solar
panels at a reasonable scale (from 1:10 to 1:50). To alleviate this
issue, the peak loads over three seconds approach is followed for
the analysis.

Fig. 14. CFD (with LES) velocity magnitude (m/s) around a solar panel scaled 1:1.
 A.M. Aly / Solar Energy 135 (2016) 423–434
Fig. 15. Peak pressure coefficients on a centerline going around the top and bottom
surfaces of a solar panel (CFD 1:10 and CFD 1:1) with BLWT model scale results (WT
1:50, WT 1:20 and WT 1:10).
3.3. Peak loads over three seconds
Peak loads are usually obtained from the time history data that
corresponds to 1 h or 10 min at full-scale. A new approach termed
the three seconds (3-s) method, permits the calculation of 3-s peak
loads by dividing the time history into several windows (W), each
one has a size of 3-s at full-scale (Aly and Bitsuamlak, 2013; Aly,
2014). The mean value of these peak values can be presented as
a robust estimate of the 3-s peak value. Accordingly, the 3-s peak
value of the normal force coefficient, CNF, can be expressed as
!,
X
W
C NF;peak;3-s ¼ max jC NF j W:
N¼1
The 3-s peak pressures (Banks, 2011) can be correctly simulated
in wind tunnel laboratories, provided that the high frequency part
of the wind velocity spectra is matching with that at full-scale
(Melbourne, 1979; Tieleman et al., 1996). Tieleman et al. (1996)
explained that the similarity of the high frequency part of the spec-
tra with the flow in nature should provide similarity in the mean
values of pressures, which also is valid in the current study. The
spectra in Fig. 12 are presented similar to the way suggested by
Richards et al. (2007), normalizing by the mean wind speed to
allow comparing the turbulence energy spectrum in the tunnel
to that expected at full-scale. This shows a better picture of the tur-
bulence content and any mismatch with the target spectra in nat-
ure rather than using the integral length scale. The figure shows
that the laboratory flow, when used to test large models (scaled
1:10), is missing the low frequency turbulence. However, when
the objective is to estimate the 3-s peak values of pressures, the
spectrum produced in the wind tunnel can be matching with the
spectrum in nature. In any case, for the current study, the best
matching at high frequency is achieved with a reduced turbulence
wind flow, for the range of model scales used. A similar approach
was suggested by Dyrbye and Hansen (1997) to test a structure
at relatively large-scale in a traditional BLWT, which required the
use of model scales larger than integral length scales.
Fig. 16 shows the 3-s peak normal force coefficients for a range
of scaled solar panels. All model scales used (1:30, 1:20, 1:10, and
1:1) have similar 3-s peak values. On the other hand, the peak and
standard deviation (STD) values calculated from the whole time
history, which is the typical approach used in wind tunnel data
analysis, are not consistent for different geometric scales. The
mean values of the total normal force coefficients are consistent
except for the scale 1:5, which may be attributed to blockage at
such large-scale. The results indicate that aerodynamic testing of
larger models of isolated ground-mounted solar panels can be
achieved by using low turbulence wind flows that have same high
frequency content as the flow in nature, when the target is to
431
Fig. 16. 3-s peak normal force coefficients for solar panel models investigated
computationally (CFD 1:1) and experimentally (1:5, 1:10, 1:20, and 1:30), in
comparison with mean, standard deviation (STD) and peak values obtained from
the total time histories.
estimate mean and 3-s peak values of loads. Accordingly, it is
possible to test larger models in artificial wind with relatively
low turbulence. This reveals the importance of such data analysis
approach as it can provide consistent peak wind load results for
solar panels, a crucial step toward codification.
It is worth noting that research carried out by the author in Fu
et al. (2012) shows that testing of low-rise residential homes might
by achieved in a low turbulence flow, when the objective is to esti-
mate the peak values of loading (on structural components). Such
conclusion remains in force for testing rigid models where the
flow-structure interaction (aeroelasticity) is not an issue and the
ð18Þ
test models are relatively small in nature where the surface pres-
sures are highly correlated. These correlations are relatively high
over typical residential homes (Fu et al., 2012). However, since
the ground-mounted isolated solar panels are believed to be smal-
ler in dimension than the typical residential homes, it is believed
that testing in a low turbulence flow is robust for estimating the
3-s peak loads.
4. Discussion
4.1. Wind tunnel testing: advantages and challenges with small-size
structures
Over the past decades, wind tunnels have been used to predict
wind loads on buildings, especially high-rise buildings and towers,
where the integral length scale of turbulence in the wind tunnel is
proportional to that in nature at a similar ratio as the geometric
scale. The advantages of wind tunnel testing (Table 1), compared
with recent open-jet testing, for instance, include having a long
test section that permits the generation of a fully developed turbu-
lent flow that can be used to test large size objects (high-rise build-
ings). The wind tunnel can also be used for typical aeroelastic tests
in a low turbulence flow. The settling room and the honeycomb
allow the generation of a very smooth flow that can be managed
for a target terrain profile. The problem, however, is related to
the testing of low-rise buildings and small-size structures, where
it is not possible to produce turbulence in the wind tunnel at a rea-
sonably large-scale. According to Richards et al. (2007), at rela-
tively large-scale of wind tunnel modeling of civil engineering
structures, it is very difficult to model the full turbulence and only
the high-frequency end of the spectrum is matched (that was also
reflected in the current study as indicated by Fig. 12 showing the
wind tunnel flow used for testing a 1:10 solar panel).
Bienkiewicz et al. (2009) carried out a comparative study of test
flows and wind pressures on low-rise buildings using wind
tunnel data generated at six wind engineering laboratories. They
432 A.M. Aly / Solar Energy 135 (2016) 423–434

Table 1 when testing bluff bodies with sharp edges. Even if the Reynolds
Benefits and challenges associated with experimental and computational simulations number similitude in wind tunnel tests is not satisfied, the aerody-
of wind loads on structures.
namic force coefficients on the model and the prototype will prob-
Technique (a) Physical BLWT (b) Virtual BLWT (c) CFD with ably be the same (Prandtl and Tietjens, 1934). Given Prandtl and
proper inflow Tietjens (1934) assumption, and the difficulty of doing tests at a
Benefits Fully developed It can provide CFD can allow Reynolds number equal to full-scale, the Reynolds number simili-
turbulence is ‘apple to apple’ testing any tude was many times ‘relaxed’ in the past (Stathopoulos and
available, which comparison with structure with any
can be used to test physical wind size at any wind
Surry, 1983). However, recent full-scale measurements, data
large size objects tunnel results. The speed, provided obtained from large wind tunnels and a series of studies from
(e.g., high-rise interference with that proper inflow Simiu and Scanlan (1996) have disputed the assumption made by
buildings) and the modeling is Prandtl and Tietjens (1934). Moreover, full-scale studies at Silsoe
aeroelastic instrumentation is available. In
cast doubt on the basic assumption that bluff-body pressures are
models. This is not a concern, in addition, the cost
attributed to the addition, it can and the human insensitive to Reynolds number (Hoxey et al., 1998).
long test section provide effort required can
with roughness continuous be less than that 4.2. CFD simulations: advantages and current challenges
elements, which pressure and flow for a typical
has smooth flow data physical
The use of a virtual wind tunnel in the current study has the fol-
at the entrance, experiment
owing to the lowing advantages:
settling room and
the honeycomb.  There is no need for an inflow that is depicting a real boundary
Such smooth flow
layer wind profile. A uniform input to the virtual wind tunnel is
can be managed
for creating a very similar to the real situation in a physical wind tunnel.
target wind  The virtual wind tunnel is more realistic with respect to simu-
profile with lating a control volume of the atmospheric boundary layer sur-
desired rounding a real structure. This is because there is a possibility of
turbulence
unrealistic results if the computational domain is expanded
characteristics
without representation of exposure effects (Yoshie et al.,
Current Testing small The Estimation of peak
2006). However, a small computational domain causes blockage
challenges structures at computational wind loads
relatively large domain becomes requires powerful and unrealistic inflow velocity profile.
geometric scales complex, which computational  The proposed virtual wind tunnel approach has very similar
can be difficult, may lead to capabilities and boundary conditions to the scenario in a physical wind tunnel.
especially for long-lasting lengthy
This permits ‘apple to apple’ comparison, which makes the
fixable test simulations that simulations. Also,
objects (for requires the need for LES
CFD focus to be basically the grid size and the turbulence clo-
instance, energy significant time turbulence sure (the solver). The success of the computational simulations
infrastructure, and closure makes the in the current study reveals that CFD is a potential tool for wind
architectural computational simulation load evaluation on civil engineering structures, provided that
features, small resource. extremely time
the boundary conditions and the surrounding objects are prop-
components, etc, Validation of the consuming.
where the integral computational Furthermore, CFD erly simulated. This help focus future research on the creation
length scale of data with wind requires of proper boundary conditions for the CFD simulations, to
turbulence can be tunnel results is validation with mimic the true physics in nature.
far different from necessary experimental data
 The created virtual wind tunnel can be used to study several
the geometric or full-scale
scale) measurements,
wind engineering problems, the same way as physical boundary
which can be a layer wind tunnels are used.
challenge, as per
limited However, the challenge with this approach is the increased
capabilities of
requirements for time and computational resources. This may
experimental
facilities, and the make the use of inflow that is statistically generated as an input
lack of reliable for a smaller computational domain more preferred, especially
full-scale data for transient simulations for the purpose of peak pressure estima-
tion. Also, another shortcoming of a CFD based approach is being
computationally costly to predict peak loads on structures under
investigated the variability in the laboratory wind loading (Fritz turbulent flows. When a relatively accurate turbulence closure,
et al., 2008; Simiu, 2009). The comparative results (as presented for example LES, is incorporated into the numerical modeling at
in Fritz et al., 2008) show that ratios of maximum to minimum high Reynolds number, the CFD simulations require high perfor-
negative pressures at a corner tap varied between 1.2 and 3.0 for mance computing capabilities. This challenge coupled with the
open terrain. This variability was primarily attributed to differ- costly commercial CFD license is limiting CFD for the estimation
ences in the approach flows employed in the physical modeling of design wind loads in practice. Wind tunnel testing still remains
of wind pressures on tested buildings. The variability in the an economic choice compared to CFD simulations. Nevertheless,
approach flows resulted, to a large extent, from the differences in concurrent CFD studies provide additional opportunities to
the along-wind turbulence intensity implied by different empirical explain/augment wind tunnel testing.
models (Simiu, 2009). This is more complicated when dealing with Table 1 summarizes the advantages of the CFD simulations in
a scale different from the typical scales a BLWT was built for. comparison with BLWTs. The CFD is potential for full-scale
Another challenge with small-size wind tunnels is the fact that simulations and challenging with computational resource. CFD
the Reynolds number can be far different from the one in nature. simulations require experimental validation as the error of
For many decades, it has been a common belief that Reynolds the computational analysis can be defined with a reference to a
number has minor effects on the wind load evaluation, especially full-scale or a wind tunnel-study. The wind tunnel has the clean
 A.M. Aly / Solar Energy 135 (2016) 423–434
and fully developed turbulence but it is challenging to simulate
wind flow at large-scales, for peak loads prediction over a long per-
iod of time (e.g., 1 h or 10 min). The CFD is a newcomer that should
be welcomed and additional research work is needed in different
disciplines including mechanical engineering, structural engineer-
ing, physics, and computer science, to allow building the more effi-
cient simulation capabilities. There is a need for parallelized
efficient simulations and more capable hardware (GPU) to permit
feasible CFD simulations at full-scale of larger engineering
problems.
4.3. Future work
Although an approach is presented to deal with the scale issues
in wind tunnel testing of ground-mounted solar panels, more
research on roof-mounted solar panels and residential homes is
needed. Future research should include both area-averaged pres-
sures and total loads on larger structural members and compo-
nents. The 3-s approach needs further investigations. It is worthy
to mention that while both top and bottom CFD localized peak
pressures at full-scale are far from measured wind tunnel pres-
sures (Fig. 15), the overall net normal force is within the range of
the measured values (Fig. 16). This said, such conclusion may not
be generalized to include low-rise buildings with closed envelope,
for instance, and perhaps CFD simulations with proper inflow pro-
duce peak roof pressures far different from those obtained by wind
tunnel testing. However, the 3-s peak loads are shown to be consis-
tent for both wind tunnel at different scales and CFD at full-scale
(with LES simulations utilizing proper inflow). These findings
should be further investigated for low-rise buildings and other
small structures in future research.
5. Conclusions
The paper presents intensive CFD investigations on ground-
mounted solar panels, aimed at understanding the model scale
and the inflow effects on the pressure distribution. Mean pressure
coefficients are not significantly affected by the model size while
peak pressures are alerted by both geometric scale and inflow tur-
bulence. Concerning mean pressure coefficients, there is an excel-
lent agreement between the CFD simulations and the experimental
results. The LES results on models tested in a virtual boundary-
layer wind tunnel are very similar to those obtained in a physical
wind tunnel, regarding both mean and peak pressures. This shows
the capability of CFD turbulence closures, for example LES, to cap-
ture both mean and peak pressure fluctuations. However, peak
pressures derived from the BLWT testing (for a wide range of
scales) are low compared to the LES generated peak pressures
(for the full-scale model with proper inflow turbulence). This
may be attributed to the lack of large scale turbulence in the phys-
ical wind tunnel. The CFD is a newcomer that is expected to reduce
the experimental effort, especially with the rapid development in
computing technology, as well as investigating wind engineering
problems at full-scale. To minimize discrepancies in peak load esti-
mation, the 3-s data analysis approach was presented. The method
shows consistent results between CFD and wind tunnel peak loads,
at different geometric scales. This reveals the importance of such
data analysis approach, as well as CFD simulations with proper
inflow turbulence, for peak wind load evaluations on solar panels,
a crucial step toward codification.
Acknowledgements
The data used for CFD validation in the current paper comes
from an experiment carried out by the author at the boundary
433
layer wind tunnel of Western University. Many thanks to Dr. Girma
Bitsuamlak and the wind tunnel team for their support.
References
Abiola-Ogedengbe, A., 2013. Experimental investigation of wind effect on solar
panels M.S. Thesis. University of Western Ontario, London, ON, Canada.
Aly, A., Bitsuamlak, G., 2014. Wind induced pressures on solar panels mounted on
residential homes. J. Archit. Eng., ASCE 20 (1), 04013003.
Aly, A.M., 2014. Atmospheric boundary-layer simulation for the built environment:
past, present and future. Build. Environ. 75, 206–221.
Aly, A.M., Bitsuamlak, G., 2012. Aerodynamic pressures on ground mounted solar
panels: test model scale effects. In: Proceedings of the 2012 Int’l Conference on
Advances in Wind and Structures (AWAS’12), Seoul, South Korea. August.
Aly, A.M., Bitsuamlak, G., 2013. Aerodynamics of ground-mounted solar panels: test
model scale effects. J. Wind Eng. Ind. Aerodyn. 123, 250–260.
Aly, A.M., 2013. Proposed approach for determination of tributary areas for
scattered pressure taps. Wind Struct. 16 (6), 617–627.
Banks, D., 2013. The role of corner vortices in dictating peak wind loads on tilted flat
solar panels mounted on large, flat roofs. J. Wind Eng. Ind. Aerodyn. 123, 192–
201.
Banks, D., 2011. Measuring peak wind loads on solar power assemblies. In: The 13th
International Conference on Wind Engineering, Amsterdam, Netherlands, July.
Bienkiewicz, B., Endo, M., Main, J.A., 2009. Comparative inter-laboratory study of
wind loading on low-rise industrial buildings. In: ASCE/SEI Structural Congress.
American Society of Civil Engineers, Austin, Texas. April–May.
Bitsuamlak, G., Dagnew, A., Erwin, J., 2010. Evaluation of wind loads on solar
panel modules using CFD. In: The Fifth International Symposium on
Computational Wind Engineering (CWE2010), Chapel Hill, North Carolina,
USA, May 23–27.
Blackmon, J.B., 2014. Heliostat drive unit design considerations–Site wind load
effects on projected fatigue life and safety factor. Sol. Energy 105, 170–180.
BLWTL, Boundary Layer Wind Tunnel Laboratory, 2007. Wind tunnel testing: a
general outline. University of Western Ontario, London, Ontario, Canada, p.
2007, May.
Browne, M.T.L., Gibbons, M.P.M., Gamble, S., Galsworthy, J., 2013. Wind loading on
tilted roof-top solar arrays: the parapet effect. J. Wind Eng. Ind. Aerodyn. 123,
202–213.
Cao, J., Yoshida, A., Saha, P., Tamura, Y., 2013. Wind loading characteristics of solar
arrays mounted on flat roofs. J. Wind Eng. Ind. Aerodyn. 123, 214–225.
Chung, K., Chang, K., Liu, Y., 2008. Reduction of wind uplift of a solar collector
model. J. Wind Eng. Ind. Aerodyn. 96, 1294–1306.
Davidson, L., 2007. Using isotropic synthetic fluctuations as inlet boundary
conditions for unsteady simulations. Adv. Appl. Fluid Mech. 1 (1), 1–35.
Davidson, L., 2008. HYBRID LES-RANS: inlet boundary conditions for flows with
recirculation. Advances in Hybrid RANS-LES Modelling, Notes on Numerical
Fluid Mechanics and Multidisciplinary Design, vol. 97. Springer Verlag, pp. 55–
66.
Davidson, L., 2011. Fluid mechanics, turbulent flow and turbulence modeling. Div.
of Fluid Dynamics, Dep. of Applied Mechanics, Chalmers University of
Technology, Göteborg, Sweden.
Dyrbye, C., Hansen, S., 1997. Wind Loads on Structures. Wiley & Sons, Chichester,
England.
Eurocode 1, 2004. Actions on structures—part 1–4: general actions—wind actions.
European Standard prEN 1991-1-4.
Franke, J., Hellsten, A., Schlünzen, H. and Carissimo, B., 2007. Best practice guideline
for the CFD simulation of flows in the urban environment. COST Action 732;
quality assurance and improvement of microscale meteorological models.
Fritz, W.P., Bienkiewicz, B., Cui, B., Flamand, O., Ho, T.C.E., Kikitsu, H., Letchford, C.
W., Simiu, E., 2008. International comparison of wind tunnel estimates of wind
effects on low-rise buildings: test-related uncertainties. J. Struct. Eng. ASCE 134,
1887–1890.
Fu, T.C., Aly, A.M., Chowdhury, A.G., Bitsuamlak, G., Yeo, D., Simiu, E., 2012. A
proposed technique for determining aerodynamic pressures on residential
homes. Wind Struct. 15 (1), 27.
Harris, R.I., 1968. On the spectrum and auto-correlation function of gustiness in
high winds. Electrical Research Association. Report number 5273.
Ho, T.C.E., Surry, D., Morrish, D., Kopp, G.A., 2005. The UWO contribution to the NIST
aerodynamic database for wind loads on low buildings: part 1. Archiving format
and basic aerodynamic data. J. Wind Eng. Ind. Aerodyn. 93, 1–30.
Holmes, D.J., 2007. Wind Loading of Structures. Taylor and Francis, New York, NY.
Hoxey, R.P., Reynolds, A.M., Richardson, G.M., Robertson, A.P., Short, J.L., 1998.
Observations of Reynolds number sensitivity in the separated flow region on a
bluff body. J. Wind Eng. Ind. Aerodyn. 73, 231–249.
Huang, S.H., Li, Q., Wu, J.R., 2010. A general inflow turbulence generator for large
eddy simulation. J. Wind Eng. Ind. Aerodyn. 98, 600–617.
Jubayer, C.M., Hangan, H., 2014. Numerical simulation of wind effects on a stand-
alone ground mounted photovoltaic (PV) system. J. Wind Eng. Ind. Aerodyn.
134, 56–64.
Kopp, G.A., Surry, D., Chen, K., 2002. Wind loads on a solar array. Wind Struct. 5 (5),
393–406.
Kopp, G.A., Farquhar, S., Morrison, M.J., 2012. Aerodynamic mechanisms for wind
loads on tilted, roof-mounted, solar arrays. J. Wind Eng. Ind. Aerodyn. 111, 40–
52.
434 A.M. Aly / Solar Energy 135 (2016) 423–434
Melbourne, W.H., 1979. Turbulence effects on maximum surface pressures.
Proceedings of the 5th International Conference on Wind Engineering, Fort
Collins, Colorado, vol. 1, pp. 541–551, July.
Meroney, R.N., Neff, D.E., 2010. Wind effects on roof-mounted solar photovoltaic
arrays: CFD and wind-tunnel evaluation. In: The Fifth International Symposium
on omputational Wind Engineering (CWE2010), Chapel Hill, North Carolina,
USA, May.
Morton, O., 2006. Solar energy: a new day dawning?: Silicon Valley sunrise. Nature
443 (7107), 19–22.
Pfahl, A., Buselmeier, M., Zaschke, M., 2011. Wind loads on heliostats and
photovoltaic trackers of various aspect ratios. Sol. Energy 85, 2185–2201.
Plate, E.J., 1982. Wind tunnel modeling of wind effects in engineering. In: Plate, E.J.
(Ed.), Engineering Meteorology, Studies in Wind Engineering and Industrial
Aerodynamics, vol. 1. Elsevier, Amsterdam, Oxford, New York, pp. 573–639.
Prandtl, G., Tietjens, O., 1934. Applied Hydro and Aeromechanics. McGraw-Hill,
New York.
Radu, A., Axinte, E., 1989. Wind forces on structures supporting solar collectors. J.
Wind Eng. Ind. Aerodyn. 32, 93–100.
Radu, A., Axinte, E., Theohari, C., 1986. Steady wind pressures on solar collectors on
flat-roofed buildings. J. Wind Eng. Ind. Aerodyn. 23, 249–258.
Richards, P.J., Hoxey, R.P., Connell, B.D., Lander, D.P., 2007. Wind-tunnel modelling
of the Silsoe Cube. J. Wind Eng. Ind. Aerodyn. 95 (9–11), 1384–1399.
Robert, F., 2003. Tricks for beating the heat help panels see the light. Science 300
(5623), 1219.
Sadek, F., Simiu, E., 2002. Peak non-Gaussian wind effects for database-assisted low
rise building design. J. Eng. Mech. 128 (5), 530–539.
Shademan, M., Hangan, H., 2009. Wind loading on solar panels at different
inclination angles. In: 11th American Conference on Wind Engineering, San
Juan, Puerto Rico, June.
Shademan, M., Balachandar, R., Barron, R.M., 2014a. Detached eddy simulation of
flow past an isolated inclined solar panel. J. Fluids Struct.
Shademan, M., Barron, R.M., Balachandar, R., Hangan, H., 2014b. Numerical
simulation of wind loading on ground-mounted solar panels at different flow
configurations. Can. J. Civ. Eng. 41, 728–738.
View publication stats
Shuman, F., 1911. Power from Sunshine. Sci. Am. 105, 291–292.
Simiu, E., 2009. Toward a standard on the wind tunnel method. NIST Technical Note
1655.
Simiu, E., Scanlan, R., 1996. Wind Effects on Structures. John Wiley & Sons, New
York.
Singh, G.K., 2013. Solar power generation by PV (photovoltaic) technology: a
review. Energy 53, 1–13.
Stathopoulos, T., Surry, D., 1983. Scale effects in wind tunnel testing of low
buildings. J. Wind Eng. Ind. Aerodyn. 13 (1–3), 313–326.
Stathopoulos, T., Zisis, I., Xypnitou, E., 2014. Local and overall wind pressure and
force coefficients for solar panels. J. Wind Eng. Ind. Aerodyn. 125, 195–206.
Stathopoulos, T., Zisis, I., Xypnitou, E., 2012. Wind loads on solar collectors: a
review. In: Proceedings of Structures Congress 2012, Chicago, Illinois, pp. 1169–
1179, March.
Tieleman, H.W., Surry, D., Mehta, K.C., 1996. Full/model scale comparison of surface
pressured on the Texas Tech experimental building. J. Wind Eng. Ind. Aerodyn.
61, 1–23.
Tominaga, Y., Mochida, A., Yoshie, R., Kataoka, H., Nozu, T., Yoshikawa, M.,
Shirasawa, T., 2008. AIJ guidelines for practical application of CFD to
pedestrian wind environment around buildings. J. Wind Eng. Ind. Aerodyn.
96, 1749–1761.
Warsido, W.P., Bitsuamlak, G.T., Barata, J., Gan Chowdhury, A., 2014. Influence of
spacing parameters on the wind loading of solar array. J. Fluids Struct. 48, 295–
315.
Webb, S.A., 2007. Covering the Planet with Solar Panels. Science 315 (5813), 869.
Wood, G.S., Denoon, R.O., Kwok, K.C.S., 2001. Wind loads on industrial solar panel
arrays and supporting roof structure. Wind Struct. 4, 481–494.
Yoshie, R., Mochida, A., Tominaga, Y., 2006. CFD prediction of wind environment
around a high-rise building located in an urban area. In: The Fourth
International Symposium on Computational Wind Engineering, July,
Yokohama, Japan.
Zhang, Y., Habashi, W.G., Khurram, R.A., 2015. Predicting wind-induced vibrations
of high-rise buildings using unsteady CFD and modal analysis. J. Wind Eng. Ind.
Aerodyn. 136, 165–179.