Structures Congress 2013, pp. 1555-1564. doi: 10.1061/9780784412848.

137

Aerodynamic Loads on Solar Panels

Aly Mousaad Aly1 and Girma Bitsuamlak2
1

Research Fellow, 2Associate Professor, Associate Director WindEEE Research

Institute, Department of Civil and Environmental Engineering, Western
University , Boundary Layer Wind Tunnel Laboratory Rm. 105, London, ON,
Canada, N6A 5B9. Emails: 1aalysaye@UWO.ca, 2gbitsuam@UWO.ca

ABSTRACT
The existing literature has limited aerodynamic data for the evaluation of design
wind loads for solar panels. Furthermore, there are no provisions in building codes
and standards to guide the design of these types of structures for wind. This paper
presents a systematic wind tunnel study to evaluate wind loads on solar panels
mounted on low-rise gable buildings. A preliminary geometric scale effect study
using a simple isolated solar panel was carried out to permit design appropriate
wind tunnel experiments. Following the scale effect study, wind loads on solar
panels mounted on different critical zones of low-rise residential roof are
systematically investigated. The results of the current paper provide useful
information for the design of the solar panels.
INTRODUCTION
There is a renewed interest in using solar panels (photovoltaic/thermal) as a
renewable source of energy (Lubitz, 2011; Chung et al., 2011; Kopp et al., 2012a,
2012b). However, a cost-effective and safe design that will make such energy
generation alternative competitive with traditional energy resources remains a
challenge. A safe, yet economical, design of the solar panels for wind requires
accurate information on the estimation of the wind loads, among other factors,
which can affect significantly the construction cost. Since these structures are
relatively new, wind loads on solar panels are hardly covered by current design
standards. The existing literature has less comprehensive data for evaluating wind
loads on solar panels mounted on residential buildings. Furthermore, there is no
provision in building codes and standards to guide the design of these types of
structures for wind. As a result, structural engineers and practitioners adopt
provisions that were established for low-rise buildings, resulting in either
conservative designs that overestimate the wind loads or unsafe structures
(Barkaszi and OBrien, 2010).
Boundary-layer wind tunnel (BLWT) testing of structures is an industry wide
accepted procedure and is considered the main source of information for wind
load calculations and codification. Most BLWTs were built for testing models of
large civil engineering structures that have geometric scales ranging from 1:500 to
1:100. However, producing typical aerodynamic models of the solar panel
modules at such scales makes the panels too small, resulting in technical
problems: the resolution of pressure data on such small-scale models becomes low
and also the interference of the instrumentation becomes significant. A research
work carried out by Holdo et al. (1982) suggests that length scales are very
1

important scaling parameters and should be maintained in testing of low-rise

buildings. Stathopoulos and Surry (1983) compared wind loads from low building
models with different geometric scales. They concluded that small relaxation
factor in the scale (up to a factor of two) may lead to an error of the order of 10%
in both local and area loads acting on the roof while the error is in the order of
25% in the peak pressure coefficients acting on the building walls. Zhao et al.
(1996) carried out a pressure distribution study on a bluff body in wind in which
they concluded that, as long as the roughness height is correctly modeled, accurate
results of the pressure coefficient on the building surface can be obtained, even if
the scale ratio of the body size is not correctly selected. According to Richards et
al. (2007), at relatively large-scale wind-tunnel modeling of civil engineering
structures, it is very difficult to model the full turbulence spectra and so only the
high-frequency end of the spectrum is matched. Bienkewicz et al. (2007) carried
out a comparative study of approach flow and wind pressures on low buildings
using wind tunnel data generated at six wind engineering laboratories and
investigated the variability in the laboratory wind loading. This variability was
primarily attributed to differences in the approach flows employed in physical
modeling of wind pressures on tested buildings. According to Bienkewicz et al.,
(2007) the variability in the data collected on low-rise buildings from different six
wind tunnels was primarily attributed to differences in the approach flows
employed in physical modeling of wind pressures on tested buildings, carried out
by the participating laboratories. The variability in the approach flows resulted in
a large measure from the differences in the along-wind turbulence intensity
implied by different empirical models, defining the target wind exposures used by
the laboratories. This indicates the sensitivity of the aerodynamic data to accurate
inflow characteristics. It is warranted to target objectively defined wind flow
characteristics representing different exposures. This is further complicated when
dealing with a scale different from the typical scales BLWTs were built for.
A number of studies have been developed for the estimation of the wind loads
on roof-mounted solar panels (Geurts and Van Bentum, 2007; Shademan and
Hangan, 2009; Kopp, et al. 2002; Ginger et al., 2011; Chung et al., 2011). Radu et
al. (1986) conducted wind tunnel tests on an array of solar panels mounted on top
of a five-story residential building, and showed the significance of the sheltering
effect from the building and the first row of solar collectors. Peterka et al. (1987)
conducted a series of wind tests on an array of heliostats, and investigated the
wind load reductions obtained from utilizing protective fences, as well as the
sheltering effect from neighboring panels. Wood et al. (2001) conducted windtunnel pressure tests on a scaled model of a large industrial building with solar
panels mounted parallel to a flat roof. The study concluded that the orientation of
a solar panel with respect to wind-flow and its proximity to the leading edge of
the building had a significant effect when designing a solar panel system. Kopp et
al. (2002) conducted wind-tunnel tests to study aerodynamically induced torque
loads on a solar panel array system. Based on the observations gathered, a large
peak torque resulted from a strong vortex shedding, in addition to turbulence in
the incoming wind-flow.
The current paper investigates systematically the sensitivity of the
aerodynamic pressure data when experimentally testing small-size structures, such
as, ground mounted solar panels at different test model sizes, in different wind
profiles. The paper also presents wind load investigations on solar panel modules
2

mounted on low-rise buildings with gable roofs that have two distinct slopes.
Wind loads on the solar panels mounted on several zones of the roofs were
systematically investigated in a boundary-layer wind tunnel for different wind
directions. The aim of this study is to produce useful information for the
estimation of wind loads on the solar panels.
CHOICE OF AN APPROPRIATE GEOMETRIC SCALE
Five sizes of solar panels were considered in a boundary-layer wind tunnel study
with scales 1:50, 1:30, 1:20, 1:10 and 1:5 (the 1:50 scale is shown in Figure 1).
The arrangements of the modules used are: tilt angle () of 40o and leg height (H)
of 24 inches (0.61 m) at full-scale. Two wind profiles were considered in the
current study (Figure 2): (a) low turbulent flow E0 which has less turbulent
content at low frequency and (b) open terrain scaled 1:400 E3 which represents
the available wind flow that entails to simulating the atmospheric flows of open
terrain exposures in a wind flow that is designed for testing models of length
scales about 1:400. The generated wind spectra for the along-wind velocity
component are plotted in Figure 2 in comparison with von Karaman spectra taken
from the literature (Holmes, 2007). The mean wind speed at the average height of
the panels was about 12 m/s. The test model scaled 1:5 required a reduced tunnel
wind speed of about 9 m/s to avoid aggressive loads and associated vibrations.
The time periods of recording pressure data were 360 s, 240 s, 120 s, 80 s and 48 s
for the scales 1:5, 1:10, 1:20, 1:30 and 1:50, respectively.

Figure 1. Photograph of the 1:50 model in the BLWT

Figure 2. Wind spectra for a 1:20 scaled panel.

RESULTS OF THE GEOMETRIC SCALE STUDY

The net pressure coefficient at any location on an individual module, Cpnet(t), is
the simultaneous difference between the upper pressure coefficient at the upper
surface, Cpupper(t), and the pressure coefficient at the bottom surface, Cpbot(t), at
the same locations
Cpnet (t ) = Cpupper (t ) Cpbot (t ).
(1)
The coefficient of the normal load acting on test model is defined as follows:

Cp
n

CF (t ) =

i =1

(t ) Ai

A
net ,i

(2)

i =1

where Ai is the tributary area of tap number i, i = 1, 2, 3, n (total number of

taps).
Figure 3 shows mean and peak normal force coefficients for solar panel
modules under different wind profiles. It is shown that the mean values of the
normal load coefficients are not significantly affected by the model. However, for
a very large module (scale 1:5), the mean value is slightly higher as this module
has a high blockage ratio (7.8 %). This reveals that the mean wind loads on the
solar panels remain unchanged regardless of the turbulence intensity and the
model size under wind flows that have similar high frequency turbulent. This can
be explained as high frequency turbulence is the key parameter that controls the
flow patterns around bluff bodies in general (Melbourne, 1979; Tieleman et al.,
1996). In contrast, peak wind loads are governed by both turbulence intensity and
low frequency turbulent. As smaller models are located closer to the ground and
hence immersed in a higher turbulent flows, they have higher fluctuating
pressures. In addition, this result is also in agreement with the results presented by
Hunt (1981) who concluded that test models larger than the correct size
(geometric scale matching with integral length scale) will underestimate the
magnitude of the pressures. Furthermore, flows with higher low frequency
turbulence cause higher pressure fluctuation on the solar panels indicated by
higher maximum load coefficients.
It has been stated by researchers that the 3-s peak pressures can be correctly
simulated in wind tunnel laboratories, provided that the high frequency part of the
spectra is matching with that at full-scale (Banks 2011, Melbourne 1980,
Tieleman et al. 1996). This typically means that the wind simulated for testing
relatively large-scale laboratory models might have different turbulence intensity
(lower than that in nature) in order for the end tail of the spectra to be matching.
Figure 2 shows that the laboratory flow, when used to test large models (1:20), is
missing the low frequency turbulence. However, when the objective is to estimate
the 3-s peak values of pressures, the spectra produced in the wind tunnel (wind
flow E0) could be matching with the spectra in nature.

1:5

1:10

1:20

1:30

1:50

1:5

1:10

1:20

1:30

5
1.5
peak

CF

3
2

0.5

1
0

E0

E3

E0

E3

Figure 3. Mean hourly and peak normal force coefficients

under different wind exposures.

peak

(3s)

1.5

CF

CF

mean

0.5

1:50

1:30

1:20

1:10

1:5

Scale

Figure 4. 3-s peak normal force coefficients for solar panel

modules under wind exposure E0.

Figure 5. Four configurations were used: (a) horizontal H1, (b)

vertical V1, (c) horizontal H2 and (d) vertical V2.
5

1:50

The 3-s peak values of loads may be calculated by using the 3-s mean wind
speed. By dividing the time history of the force coefficient to several windows,
each one has a size of 3-s, one can obtain a relatively high number of 3-s peak
values. The mean value of these values can be presented as the robust estimate of
the 3-s peak value. Figure 4 shows the 3-s peak normal force coefficients under
wind exposure E0. Except for the model scaled 1:50, all other model scales
used (1:30, 1:20, 1:10, and 1:5) have similar 3-s peak values of the normal force.
This means that the aerodynamic testing of larger models of the ground-mounted
solar panels can be achieved by using a low turbulence wind flow that has the
same high frequency content as the flow in nature when the target is to estimate
the 3-s peak values. Accordingly, it is possible to test larger models in artificial
wind which has relatively low turbulence. It is worth noting that another research
carried out by the authors in Fu et al. (2012) shows that testing of low-rise
residential homes might by achieved in low turbulence flow, when the objective is
to estimate the peak values of loading. Such conclusion remains in force for
testing rigid models where the flow structure interaction (aeroelasticity) is not an
issue and the test models are relatively small in nature.
ROOF MOUNTED SOLAR PANELS
Three sizes of roof-mounted solar panels are considered, namely small S (3 ft x
5 ft or 0.9144 m x 1.524 m), medium M (5 ft x 8 ft or 1.524 m x 2.4384 m) and
big B (5 ft x 9 ft or 1.524 m x 2.7432 m). All the modules and the building
model were scaled 1:15 and mounted on gable roofs with two different slopes.
This scale is not commonly used for testing residential buildings in traditional
boundary-layer wind tunnels. However, as discussed early in the paper (see also
Aly and Bitsuamlak, 2012), a large model that allows proper pressure
measurements and architectural details representation is necessary. The scale
taken in this study was a result of a compromise choice between making the
modules sufficiently large for pressure instrumentation and keeping the wind
tunnel blockage less than 5%. Kopp et al (2012b) presented wind tunnel testing of
solar panel modules at relatively large scales (Dyrbye and Hansen, 1997; Visscher
and Kopp, 2007).
Table 1 Net pressure coefficients over roof zones (3, 2 and 1) for
solar panels mounted on a gable roof with a slope of 3:12 (14o)
and comparisons with data provided by the ASCE-2005.
Zone 3
Zone 2
Zone 1
Configuration
GCP+
GCPGCP+
GCPGCP+
GCPWind tunnel
H1
1.03
-2.78
1.22
-1.94
0.40
-0.32
V1
1.02
-2.83
1.16
-3.01
0.34
-0.42
H2
_
_
_
_
0.71
-0.34
V2
_
_
_
_
0.43
-0.22
Bare roof (slope = 14o)
0.77
-4.14
0.56
-3.12
0.16
-0.96
ASCE bare roof
(7o < slope < 27o)

0.50

-3.70

0.50
6

-2.00

0.50

-0.90

A gable roof with a slope of 3:12 (14o) was first instrumented with pressure
taps for external pressure assessment on a bare roof. Second, the modules were
mounted on a quarter of the building as shown in Figure 5. Note that only a
quarter of the roof was instrumented due to symmetry. The rest of the roof was
covered by dummy modules (to produce an aerodynamic scenario similar with a
roof covered entirely with solar panels) according to the test configuration type.
Four different arrangements of the modules were tested: (1) horizontal
arrangement where the modules are covering the entire roof, with small gap
between the rows, this arrangement is called configuration H1 (Figure 5-a); (2)
vertical arrangement where the modules are covering the entire roof without any
significant gaps (configuration V1 as shown in Figure 5-b); (3) horizontal
arrangement where the critical roof corner zones close to the eave, ridge and
gable-end edges are avoided, and small gap between the rows exist (configuration
H2 as shown in Figure 5-c); (4) vertical arrangement where the roofs critical
corner zones are avoided, and small gap between the modules exist (configuration
V2 as shown in Figure 5-d). All the modules were placed parallel to the main roof
surface with a full-scale gap of 6 inches (0.1524 m).
An open country wind profile only for the lower 40 m portion of the
atmospheric boundary layer (ABL) was simulated at RWDIs boundary-layer
wind tunnel in Miramar, Florida (Aly and Bitsuamlak 2012). Wind pressure data
over the top and bottom surfaces of the modules were collected for a time period
of 90 s at a sampling rate of 512 Hz. The velocity scale was assumed to be 4. For
a geometric length scale of 15 the time scale is 3.75. This means that the 90-s test
period in the laboratory represents 5.625 minutes at full-scale. The mean wind
speed over 3-s period was used for pressure data normalization.
RESULTS OF THE ROOF-MOUNTED SOLAR PANEL STUDY
Table 1 lists net pressure coefficients for the solar panel modules mounted over
roof zones (3, 2 and 1) and comparisons with data provided by the ASCE-2005
(Figure 6-11C of the standard) for claddings and components (C&C). It is shown
that the external minimum pressure coefficients estimated over the bare roof
(slope 3:12) are relatively higher than the net pressure coefficients estimated over
solar panels mounted on the corresponding zones. For zones 3, the ASCE standard
predicts higher minimum pressure coefficients by a factor of about 32%.
Minimum pressure coefficients over the solar panel modules in zone 2 are
dependent on the configuration (horizontal or vertical). The horizontal
configuration of solar panels in zone 2 (H1) is in agreement with the minimum
pressures predicted by the ASCE standard over a bare roof with an overhang
(Table 1). However, net pressures over panels in zone 2 with configuration V1
are relatively higher than those estimated in configuration H1. This may be due
to the effect of the gap between the panels (20% of the panels width) that exist in
configuration H1 (see Figure 5-a). This means that the data provided in the
ASCE standard for a bare roof, when used to predict pressures on solar panels,
may underestimate the minimum pressure on zone 2 for the solar panels mounted
with no significant spacing between the modules by 34%. In other words, a factor
of 1.34 might be useful to make use of the data provided in the ASCE standard for
gable roofs. It is worth noting that Banks (2007) recommends an uncertainty
7

factor of 1.4, in case data provided by ASCE 7 for bare roofs are used. The
minimum design pressure predicted by the standard on modules in zone 1 is about
two times the actual net minimum pressures.
In Table 1, it is shown that the maximum pressure predicted by the standard
on bare roofs in zones 3 and 2 are generally lower than the net pressure on the
modules. The standard slightly overestimates the maximum net pressures for the
modules in zone 1. It is to be noted that these positive pressures may not govern
the design and are very small in magnitude compared to the suction pressures.
CONCLUSIONS
The paper investigates wind effects on solar panels with an aim to reduce the
overall construction costs and to examine the possibility of using design codes
applicable to bare gable roofs. A geometric scale study was carried out to permit
using the most practical geometric scale for testing the solar panels. The study
shows that mean pressure loads are not significantly affected by the test model
size. However, very small-size solar panels may have different mean loads as they
are affected by the instrumentation (pressure tubes). At the same time, very large
models need to be used with caution not to introduce significant blockage issues.
Unlike the mean load coefficients, peak load coefficients vary according to the
model size. For the scale range considered in this study (i.e., 1:5 to 1:50), the large
models (1:5 and 1:10) violated the scale law of similitude significantly, hence
resulted in very low peak values. The results for the moderate scales (1:20 and
1:30) were consistent throughout and did not show significant variations. In the
absence of matching the entire spectrum, it is therefore necessary to carry out
project specific geometric sensitively analysis to avoid significant deviations in
the peak values. By reducing the turbulence intensity in the test flow, relatively
large-scale laboratory models can be used when the objective is to estimate the 3-s
peak loads. Irrespective of mismatch in the lower frequency, all model scales used
(1:30, 1:20, 1:10, and 1:5), except for the model scaled 1:50, have similar 3-s peak
values of loads.
Following the scale effect study, a wind tunnel study on roof-mounted solar
panels was presented. Solar panels mounted close to roof corners (zone 3) are
generally subjected to lower net pressures than the external pressure provided by
ASCE 7-2005 for a bare roof. The code over estimates the minimum pressures by
a factor of 32%. Panels mounted at roof edges and away from corners (zone 2) are
generally subjected to lower net pressures than the external pressure provided by
ASCE 7-2005 for a bare roof. The standard predicts very similar values on
modules mounted with a horizontal gap between rows (the gap is 20% of the
panels width). Nevertheless, the standard underestimates the minimum pressures
by a factor of 34% when the horizontal gap was eliminated. The minimum design
pressure predicted by the standard on modules in zone 1 is about two times the
minimum net pressure. The maximum pressure predicted by the standard in zones
3 and 2 are generally lower than the net pressures. The standard slightly
overestimates the maximum net pressures in zone 1. Generally speaking, it may
be recommended to avoid mounting the solar panels in zones located close to the
roof edge, ridge, and corners as the wind loads are higher at these zones (zone 3
and 2).
8

ACKNOWLEDGEMENTS
The financial support from Western University and Florida Center of Excellence
for Hurricane Product Development (through the International Hurricane
Research Center at Florida International University) is gratefully acknowledged.
The help received from Mr. Warsido, Mr. Erwin and Mr. Vincent Crepel, is
acknowledged with thanks. We also acknowledge the informative discussion with
Mr. Barkaszi from Florida Solar Energy Center (Solar America Board for Codes
and Standards, www.solarabcs.org).
REFERENCES
ASCE 7-2005, (2006), Minimum design loads for buildings and other structures,
ASCE Standard, ASCE/SEI 7-05, American Society of Civil Engineers,
Virginia.
Aly, A. M., and Bitsuamlak, G. (2012). Aerodynamic pressures on ground
mounted solar panels: test model scale effects. Proceedings of the 2012
Intl Conference on Advances in Wind and Structures (AWAS12), Seoul,
South Korea, August.
Banks D., 2011. Measuring peak wind loads on solar power assemblies. The 13th
International Conference on Wind Engineering, Amsterdam, Netherlands,
July.
Banks, D. (2007). How to calculate wind loads on roof mounted solar panels in
the US.
http://www.cppwind.com/support/PDF/HowToCalculateWindLoads.pdf
Barkaszi S., and OBrien C. (2010). Wind Load Calculations for PV arrays. A
report, Solar America Board for Codes and Standards,
(www.solarabcs.org, June 2012).
Chung, K. M., Chang, K. C., and Choub, C. C. (2011). Wind loads on residential
and large-scale solar collector models. J. Wind Eng. Ind. Aerodyn., 99(1),
5964.
Dyrbye, C., Hansen, S., 1997. Wind Loads on Structures. Wiley & Sons,
Chichester, England.
Fu, T.C., Aly, A.M., Gan Chowdhury, A., Bitsuamlak, G., Yeo, D.H., Simiu, E.,
2012. A proposed technique for determining aerodynamic pressures on
residential homes. Wind and Structures 15(1), 27-41.
Geurts, C. P. W., and Van Bentum, C. A. (2007). Wind loads on solar energy
roofs. Heron, 52(3), 201222.
Ginger, J., Payne, M., Stark, G., Sumant, B., and Leitch, C. (2011). Investigations
on wind loads applied to solar panels mounted on roofs, Buildings Codes
Queensland, Report No. TS821, Reviewed by Holmes, J., Cyclone Testing
Station, School of Engineering & Physical Sciences, James Cook
University, Townsville, 22 December, 2011.
Holdo, A.E., Houghton, E.L., Bhinder, F.S., 1982. Some effects due to variations
in turbulence integral length scales on the pressure distribution on windtunnel models of low-rise buildings. Journal of Wind Engineering and
Industrial Aerodynamics 10, 103-115.
Holmes, J.D. (2007), Wind Loading of Structures, 2nd Edition, Taylor and
Francis, London.
9

Hunt, A., 1981. Scale effects on wind tunnel measurements of surface pressures
on model buildings. Proceedings of the Colloquium Design with the
Wind, CSTB, Nantes, France, June.
Kopp, G.A., Surry, D. and Chen, K. (2002). Wind loads on a solar array. Wind
Struct., 5(5), 393406.
Kopp, G. A., and Banks, D. (2012a), Use of the wind tunnel test method for
obtaining design wind loads on roof-mounted solar arrays. J. Struct. Eng.
(in press). doi: 10.1061/(ASCE)ST.1943-541X.0000654
Kopp, G. A., Farquhar, S. and Morrison, M. J. (2012b), Aerodynamic
mechanisms for wind loads on tilted, roof-mounted, solar arrays. J. Wind
Eng. Ind. Aerodyn., 111 , 4052.
Lubitz, W. D. (2011). Effect of manual tilt adjustments on incident irradiance on
fixed and tracking solar panels. App. Energy, 88, 17101719.
Melbourne, W.H., 1979. Turbulence effects on maximum surface pressures. In
proceedings of the 5th International Conference on Wind Engineering, Fort
Collins, Colorado, 1, 541-551.
Melbourne, W.H., 1980. Turbulence effects on maximum surface pressures a
mechanism and possibility of reduction. Wind Engineering 1, 541-552.
Peterka, J. A., Bienkiewicz, B., Hosoya, N., and Cermak, J. E., (1987). Heliostat
mean wind load reduction. Energy, 12, 261267.
Radu, A., Axinte, E., and Theohari, C. (1986). Steady wind pressures on solar
collectors on flat-roofed buildings. J. Wind Eng. Ind. Aerodyn., 23,
249258.
Richards, P.J., Hoxey, R.P., Connell, B.D., Lander, D.P., 2007. Wind-tunnel
modelling of the Silsoe Cube. Journal of Wind Engineering and Industrial
Aerodynamics 95(9-11), 1384-1399.
Shademan, M. and Hangan, H. (2009). Wind loading on solar panels at different
inclination angles. 11th American Conference on Wind Engineering, San
Juan, Puerto Rico, June.
Stathopoulos, T., Surry, D., 1983. Scale effects in wind tunnel testing of low
buildings. Journal of Wind Engineering and Industrial Aerodynamics
13(1-3), 313-326.
Tieleman, H.W., Surry, D., Mehta, K.C., 1996. Full/model scale comparison of
surface pressured on the Texas Tech experimental building. Journal of
Wind Engineering and Industrial Aerodynamics 61, 1-23.
Visscher, B., and Kopp, G. A., (2007). Trajectories of roof sheathing panels
under high winds. J. Wind Eng. Ind. Aerodyn., 95, 697713.
Wood, G. S., Denoon, R. O., and Kwok, K. C. S. (2001). Wind loads on
industrial solar panel arrays and supporting roof structure. Wind Struct.,
4, 481494.
Zhao, Y.W., Plate, E.J., Rau, M., Keiser, R., 1996. Scale effects in wind tunnel
modeling. Journal of Wind Engineering and Industrial Aerodynamics
61(2-3), 113-130.

10