Prediction of Dust Pollution and Impact on Solar Photovoltaic System (TPD1807030)

1. Introduction
As a kind of renewable, clean and prom,smg energy, solar
photovoltaic (PV) technology has been rapidly developed and
applied around the world over last several decades (1-3]. More
than 75 GW of solar PV capacity were installed globally and the
growth rate of the global solar PV market was more than 50% in
2016 [4]. Nevertheless, dust pollution, a serious problem caused the
reduction of solar PV power efficiency and life time, has attracted
much attention recently (5-9]. Salim et al. (10] studied the in
fluences of dust deposition on solar PV power effi ciency in Saudi
Arabia. The results showed that PV efficiency was decreased more
than 30% because of dust deposition after eight months. Moreover,
Wakim [ 11 ] found the PV output efficiency is reduced by 17% within
several days because of sand deposition in Kuwait. Recently, Pavan
et al. (12] found that the reduction of PV power of poly crystalline
silicon PV modules by dust can reach 5% within a year. These
findings indicate that the solar PV power efficiency is greatly
degraded due to dust accumulation. This is because the deposited
particulates can greatly degrade the solar transmittance and
further influence the PV efficiency performance [ 13-15 ].
The studies of dust deposition on solar PV modules can be
generally divided into two aspects. The first aspect is the outdoor
studies of dust deposition effects on solar PV output efficiency in
specific regions. El Nashar (16] investigated dust deposition on the
glass transmittance and power efficiency of solar PV system. The
results showed that glass transmittance reduction can reach 10%
and the annul reduction of PV power output is 70% in the United
Arab Emirates. Besides, the influences of different PV tilt angles on
dust accumulation and transmittance of glass were studied by
Elminir et al. (17]. The results showed that higher inclination angles
can reduce the deposition rates of dust and help to against the
transmittance degradation. Recently, Chaichan et al. (18] studied
the effects of traffic air pollution on the PV cells and the results
showed that the reduction of PV efficiency can reach 12%. Adinoyi
et al. (19] examined the effect of dust accumulation on solar PV

panels in the eastern part of Saudi Arabia. They found that power
decreases as much as 50% for six months. Sayyah et al. [20] pre
sented a detailed review on the PV efficiency reduction by dust
accumulation in different regions and built a database for future
research of this topic. Moreover, Sharma et al. and Zaihidee et al.
recently conducted detailed review papers on the influences of dust
accumulation on solar PV power performance [21,22].
The second aspect is indoor experimental studies of relationship
and influencing mechanism between PV module efficiency and
dust deposition density. El Shobokshy [23] investigated the influ
ence of deposited dust particle sizes and density on solar PV per
formance. It was found that small dust had higher influence on PV
efficiency than large dust. Besides, Goossens [24] studied the in
fluences of different air flow velocities on dust deposition on PV
modules in air duct. It was observed that the reduction of PV power
is increased as the air velocity rose because of the higher dust
deposition. Recently, Jiang [25] examined effects of different PV
module types on power reduction because of dust deposition. The
results showed that the differences of PV efficiency reduction for
different types of PV cells are not obvious. Furthermore, Qasem
et al. [26] investigated the mechanisms of dust effect on spectral
transmittance of solar PV glass. The results showed that smaller
dust particles play an important role on the transmittance atten
uation. Moreover, Piliougine et al. [27] proposed a novel coating to
reduce dust accumulation on PV panels. They found the coating can
effectively decrease the soiling deposition.
However, these two aspects both focused on dust effects on PV
panel performance after dust deposition [28e32]. The character
istics of dust deposition process itself have been seldom studied
and examined. Nevertheless, dust deposition behaviours and
characteristics on PV panels are significant for further proposing
effective solution strategies. Therefore, this paper aims to examine
dust deposition process on a ground mounted PV panel, which is
seldom considered in the past studies. Besides, an empirical model
was proposed to fast predict dust deposition effect on solar PV ef
ficiency reduction based on present CFD simulation and related
experimental results. Lu et al. [33] investigated dust deposition on
building mounted solar PV panels. They found that 10 mm particles
have the maximum deposition rate on solar PV panel. However, the
dust deposition process on ground mounted PV panel has been not
studied yet. The deposition behaviors may be quite different be
tween building mounted and ground mounted PV panels, as the
flow structures are completely different for the two cases. There
fore, the objective of this study is to investigate dust deposition
behaviors on ground mounted solar PV panels and their influences
on PV efficiency performance. The present tilt angle of PV panel is
25, which is the optimum anneal power output for latitudes be
tween 30 and 45. Some solar PV penal farms within this latitude
are 97MW Sarnia solar farm in Ontario, Canada; 30MW Cimarron
Solar Facility, New Mexico, USA and 21MW Blythe Solar Project,
California, USA. These solar farms have a mounting tilt angle of 25.
Computational fluid dynamics (CFD) method was adopted and
established in this study, as it has become a powerful tool for
predicting multiphase flow and solar energy applications [34e39].
The turbulent wind flow fields around the solar PV system and the
dust deposition behaviors on the PV panel of different particle sizes
and wind velocities were predicted and analyzed carefully.
Furthermore, the effects of dust deposition on PV output reduction
was predicted for different dust diameters and different PV types,
based on the present CFD results and experimental measurement
from the literature.
The paper is organized as follows: section 2 is “Numerical
Methodology and Solution Strategy”. The Reynolds stress model for
wind flow fields and discrete particle model for dust motions were
described in this section. section 3 is “Case description”, which
introduced computational geometry and conditions. section 4 is
“Results and Discussions”. In section 4.1, grid independence study
and numerical validation were conducted to ensure the correctness
and reliability of the present methods. In section 4.2, wind flow
fields around an isolated solar PV system were discussed and
analyzed, as they are crucial for dust deposition behaviors. In sec
tion 4.3, deposition rates on solar PV panel for different dust sizes
and wind velocities were presented and discussed. Moreover, dust
trajectories for different diameters and wind velocities were dis
cussed in section 4.4. Then, influences of dust deposition on PV
efficiency were predicted by the theoretical model proposed by the
authors in section 4.5. Finally, section 5 is conclusion.
2. Numerical Methodology and Solution Strategy
The dust deposition processes and characteristics on an isolated
ground mounted solar PV system were simulated by CFD method.
In this method, the Navier Stokes equations concerned to the mass,
momentum and kinetic energy budgets are solved for wind flow.
Moreover, particles are considered as a discrete phase and the
trajectory of each particle is tracked through solving the particle
dynamic equation, such as the discrete particle model (DPM) model
in present study. In DPM model, particle motion and diffusion are
predicted by considering different particle forces. Moreover, the
turbulent dispersion of particles is modeled by Discrete Random
Walk Model (DRW). The turbulent wind flow fields around the solar
PV system were resolved by shear stress transport (SST) k u tur
bulence model, while the dust deposition behaviors were predicted
by discrete particle model (DPM). The computational cost of 3 D
air dust two phase flow is very high, especially when released
dust number is huge. Therefore, a 2 D simulation was adopted in
the study. User defined function (UDF) program were developed
and imposed to FLUENT to model the realistic inlet velocity and
turbulent kinetic energy (TKE) profiles in the wind boundary layer,
based on the wind tunnel experiment measurements of Tominaga
et al. [40]. The detailed numerical models and solution methods
were addressed as follows.
2.1. Wind flow fields around the solar PV system
The accurate simulation of air flow fields over the solar PV
system is important for further prediction of dust deposition rates
and behaviors. In general, direct numerical simulation (DNS) or
large eddy simulation (LES) can predict the air flow fields more
accurately, compared with the Reynolds Averaged Navier Stokes
(RANS) models. However, the computational cost of DNS or LES is
greatly higher than that of RANS. Moreover, it was proved that
RANS method can already accurately predict particle dispersion
and deposition in the air flow fields of building environment [40].
Therefore, the RANS model was used to predict the air flow in the
present study. Karava et al. [41] compared the numerical results for
wind flow around a building by different turbulent models. They
found that the SST k u turbulence model has the best prediction
results, as it combines the k u model near the wall and the k ε
model in the free shear layers [42,43]. In this study, the SST k u
turbulence model was used to predict the wind flow fields.
The time averaged mass and momentum equations for the
turbulent air flow fields around the isolated solar PV system can be
described by Ref. [44]:

(a) Inlet wind velocity profile

Gk and Gu in the equations are the generation of TKE k and the
dissipation rate u respectively. Gk and Gu are the effective diffu
sivity of k and u respectively. Yk and Yu are the dissipation of k and
u respectively. Du is the cross diffusion term. Sk and Su are user
defined source terms respectively. In this study, Sk and Su both
are zero.
In the study, the initial wind velocity is 1.3 m/s or 2.6 m/s in the
whole computational domain for two different wind velocity cases.
Moreover, the initial pressure is equal to standard atmospheric
pressure. The input data is the boundary conditions in the
computational domain. Different boundary conditions will results
in different simulation results of wind flow fields. In the present
study, the inlet wind velocity and TKE profiles were fitted by
Tominaga’s the experimental data [40], as shown in Fig. 1. The inlet
wind velocity profile was fitted using a power function while inlet
TKE distribution was fitted via a polynomial function. It can be
found that the fitting curves for the both inlet wind velocity and
TKE profiles agree well with the experimental measurement.
Moreover, the outflow boundary condition was used in the outlet.
The no slip boundary condition was employed for the walls. The
symmetry boundary condition was applied in the upper boundary.

(b) Inlet TKE profile

Fig. 1. Inlet wind velocity and TKE profiles.

2.2. Dust deposition behaviors
The DPM was employed to predict dust deposition behaviors on
PV panel. Because this model is effective to calculate dust deposi
tion rate by tracking the trajectories of each dust particle. As the
dust laden wind air flow is dilute enough, the influences of dust
behaviors on the air flow fields and collisions of inter particles can
be neglected. Moreover, as the places those have serious dust
deposition on solar PV panels are always dry regions, the effect of
moisture on dust depositionwas ignored in the study. In this study,
total five dust particle forces were considered in the simulation: the
drag force, the gravity force, the buoyancy force, the Brownian force
and the Saffman’s lift force. The governing equation of the dust
motions is described as follows [45].

Here, ug and up are the velocity of fluid and particle, respectively. rg
is the density of fluid andrpis the density of particle. z is normal
distributed random number. kinematic viscosity. the
deformation tensor. Eq. (5) shows the particle motion equation. The

left hand of the equation is particle accelerated velocity while the
right hand of the equation is particle forces. The input data is par
ticle density and diameter as well as the wind velocity and wind
density. The output data is particle motion velocity. Eq. (5) was
calculated by coupling with Eqs. (1)e(4) to solve the air dust two
phase flow fields in the computational domain. Compared with the
above five forces, the pressure gradient force, the virtual mass force
and the Basset force were negligible in the study. The drag coeffi
cient CD can be written by Ref. [45]:

The particle relaxation time t is calculated by Ref. [45]:

where Cc is the Cunningham slip correction factor.
Spherical particles were injected at the inlet within the three
times height of the solar PV panel height. The initial dust particle
velocities equal the air velocity at solar PV height. It was assumed
that particles are deposited if they contact the surface of the solar
PV panel. The dust rebound and resuspension were not considered
here (46-48]. For the rest of dust boundary conditions, the airborne
dust particles were assumed to leave the computational domain
once they reach the boundaries.

2.3. Solution strategy
The finite volume method (FVM) was used to resolve the con
servation equations for the turbulent wind flow fields. The con
vection term was discretized by the second order upwind scheme
and the diffusion term was discretized by the second order central
difference scheme. The Semi Implicit Method for Pressure Linked
Equations (SIMPLE) algorithm was applied to decouple the pres
sure and the velocity fields (49]. The Runge Kutta method was
adopted to resolve the dust particle motion equations.
3. Case description
Fig. 2 illustrates the simplified physical model of airborne dust
deposition on an isolated ground mounted solar PV system. The
dust laden wind boundary layer flows over the solar PV system and
some dust particles would be deposited on the surface of the PV
panel, as shown in Fig. 2. A 1 :10 scaled model experiment of wind
flow over ground mounted stand alone PV system was conducted
by Abiola Ogedengbe to investigate the wind load of the PV panel
(50]. The present CFO geometries was designed to be in consistence
with those of Abiola Ogedengbe’s wind tunnel experiments for
further validation. The solar PV panel height Hp and width Wp are
1.65 m and 2.48 m, respectively. The tilt angle of the solar PV panel
with the horizon is 25°. This inclination angle was chosen based on
the optimum annual power output at places with latitudes be
tween 300 and 45° (51 ]. The computational domain was 21.4 Hp in
length and 6 Hp in height. The distance from the inlet to solar PV
panel is 5 Hp, and the distance behind the solar PV system is 15 Hp.
Structured grids were developed to discretize the

computational domain by the ANSYS ICEM 15.0, as shown in Fig. 3
(a). Fig. 3 (b) illustrates the enlarged view of the grids near the solar
PV system. The total number of the computational grids is 113,071.
The first grid spacing from the solar PV panel is 3.0 x 10- 3 m (
0.0018 Hp). The corresponding non dimensional wall distance y+ is
8. The mesh growing factor is 1,2 from solar PV panel surface to the
center region.
The wind velocities at the solar PV height UHp are 1.3 m/s and
2.6m/s. The Reynolds number, based on the Hp and UHp, is 143,000
and 286,000 respectively. The air flow fields over the solar PV
system was predicted by the SST k w turbulence model with UDF
inlet conditions. The dust particle size range was chosen as
1-300 μm according to the experimental study by Appels et al. (52]
and Kaldellis et al. (53]. Eleven dust particle sizes were investigated
in the simulation: 1, 3, 5, 10, 30, 50, 100, 150, 200, 250 and 300 μm.
The dust in this study was calcium carbonate particle and the
density is 2800 kg/m3 in the present study. The main data consid
ered in our model is tilt angle of PV panel, the wind velocities and
the dust diameters, as shown in Table 1. The present tilt angle of PV
panel is 25°, which is the optimum anneal power output for lati
tudes between 300 and 45°, such as 97 MW Samia solar farm in
Ontario, Canada.
Generally, the initial data can be divided for two parts: wind
flow fields and dust motions. For wind flow fields, the initial data
includes wind velocity distribution and pressure distribution in the
computational domain. Moreover, the air density and air viscosity
need to be inputted in the model. For dust motions, the initial data
includes released dust spatial distribution and initial dust veloc
ities. The dust density and diameter should be given in the discrete
particle model.
4. Results and Discussions
4.1. Grid independence study and numerical validation
The grid quality and number have great influences on the CFO
simulation results. Therefore, the structural grids were developed
by ANSYS ICEM to ensure the high quality of grids. To further
validate the resolution of grids, the grid independence test was
conducted in this study. The numbers of the coarse, medium and
fine grids are 49,443, 81,257 and 113,071, respectively. The wind
flow velocity profiles at I.,, 8 m for different grid sizes were
compared, as shown in Fig. 4. It can be found that the difference

between the results of the coarse grids and the medium grids is
significant. However, the wind velocity profiles for fine and me
dium grids are almost overlapped. These indicates that the medium
grids are refined enough to obtain accurate turbulent wind flow
fields. The fine grids was adopted in the present study.
As accurate simulation of the wind flow fields over the solar PV
system is significant for correctly predicting the dust deposition
rates and behaviors, further numerical validation of the wind flow
fields is necessary. Abiola Ogedengbe investigated the wind load of
an isolated solar PV system in the wind tunnel experiments [50].
Therefore, the mean pressure coefficient profile of the upper sur
face of the solar PV panel was obtained by the present CFD simu
lation and compared with that in Abiola Ogedengbe’s experimental
measurements [50], as shown in Fig. 5. In the figure, w is the dis
tance from the leading edge along the width of the solar PV panel,
and Wp is the width of the solar PV panel. It can be observed that
the mean pressure coefficient Cp profile agrees well with the
experimental results. This implies that the present CFD models can
predict the air flow fields correctly and accurately.

Table 1
Geographical locations and database of the study.
Geographical locations Latitude (°) Wind velocities at panel height (m/s) Dust diameters (mm) Tilt angle of panel (°)
Ontario, Canada 30 45 1.3; 2.6 1 300 25

Distance from the PV leading edge w/W p

Fig. 5. Validation of the mean pressure coefficient along the upper surface of PV panel.

4.2. Wind flow fields around an isolated solar PV system

The turbulent wind flow fields have significant effects on dust
motions and deposition behaviors. Fig. 6 (a) and (b) display the
velocity fields near the solar PV system for UHp 1.3 m/s and 2.6 m/
s respectively. From Fig. 6 (a) and (b), it can be observed that the
wind velocity fields become quite complicated due to the obstacle
of solar PV system. Moreover, it can be noted that the wind velocity
is accelerated along the solar PV panel and dramatic velocity vari
ation appears near the PV system. Comparing the Fig. 6 (a) and (b),
it can be found that the wind velocity fields are quite similar for
different wind velocities. These complicated wind velocity fields

may result in complex dust motions and distributions.
Moreover, the turbulent kinetic energy (TKE) distributions
around the solar PV system for different wind velocities are illus
trated in Fig. 7. It can be found that the TKE distributions are quite
similar for different wind velocities. However, the TKE values for
case of UHp 1.3 m/s are lower compared with the case of
UHp 2.6 m/s. Furthermore, TKE values above the PV panel are
relatively high for the both cases. This TKE distribution may
enhance dust deposition on the solar PV panel. Because particle
turbulent fluctuation would be more intense due to the high TKE of
the fluids. This effect will enhance the dust deposition, especially
for small sized dust [46,47].
Furthermore, the streamlines of air flow over the solar PV sys
tem for different wind velocities was obtained and displayed in
Fig. 8. The color gradients in Fig. 8 represent the local wind velocity
magnitude. Two small turbulent eddies and one large scale eddy
can be observed below and behind the solar PV system, respec
tively. However, the large turbulent eddy behind PV panel for
UHp 1.3 m/s is slightly larger than the case of UHp 2.6 m/s. The
streamlines above the solar PV panel are just parallel with PV
surface. Small sized dust particles would just follow the stream
lines of the wind flow due to their low inertia. However, the
gravitation and inertia has much more influences on particle mo
tions for large sized dust.

4.3. Deposition rates for different dust sizes and wind velocities
Fig. 9 showed the deposition rates on PV module for different
dust diameters and wind velocities. The dust deposition rate l was
defined as follows,

l
Nd
Np
100% (8)
In Eq. (8), Nd andNp are the number of deposited dust particles

Fig. 6. Turbulent wind velocity fields over an isolated solar PV system.

Fig. 7. TKE distributions of wind flow over the solar PV system.

Fig. 8. Streamlines of air flow around solar PV system.

on PV panel and inlet released dust particles, respectively. From
Fig. 9, dust deposition rates on solar PV panel are dramatically
different for different particle sizes. The profile of deposition rates
first rises, and then falls when the dust particle sizes increase.
When wind velocity is 2.6 m/s, the peak value of deposition rate is
14.28% for 150 mm dust. However, the minimum depositon rate for
300 mmparticles is only 0.17%. For the smallest dust particles (1 mm)
in this study, the deposition rate is about 5.12%. It can be found that
there are almost two orders of magnitude difference between the
maximum and minimum deposition rates of dust particles. This
indicates that the deposition behaviours and mechanisms are
completely different for different size particles.

Fig. 9. Oust deposition rates on an isolated ground-mounted solar Pl/ panel.

When wind velocity is 1.3 m/s, it can be found that the dust
deposition rates are quite similar with the case of higher wind
velocity (2.6 m/s). The deposition rates are firstly increased and
then decreased with the increase of dust diameter. This means that
the dust deposition behaviors are almost the same for different
wind velocities. However, the maximum deposition rate for
UHp 13 m/s is slightly lower compared with the case of
UHp 2.6 m/s. Moreover, the dust size for the maximum deposition
is 100 μm, which is smaller than the case of higher wind velocity.
The medium size dust has the maximum deposition rate on solar
PV panels. This findings may be valuable and helpful for proposing
according solution to prevent dust deposition on PV panels. For
example, specific self cleaning coating can be developed to prevent
dust accumulation on PV panels for medium size particles. More
over, the predicted data in Fig. 9 can be also used in the theoretical
model to estimate solar PV efficiency reduction caused by dust
deposition.
Moreover, dust deposition rates on building mounted PV panels
were previously studied by the authors. It was found that the
highest deposition rate is 0.28% for 10 μm dust when wind velocity
is 2.6 m/s, while the lowest deposition rate is 0.13% for 50 μm dust.
These results are quite different from the present study. Generally,
dust deposition behaviors on PV modules are influenced by many
factors, such as the initial particle velocities, the inertia of the dust
particles, the gravitational effect of dust particles, the flow struc
tures around the PV panel, the turbulent kinetic energy (TKE) dis
tribution around the PV panel. Compared with building mounted
PV panel, the dust diameter of maximum deposition is increased
from 10 μm to 150 μm dust when solar PV is mounted on ground.
Moreover, the maximum deposit ion rate is also increased from
0.28% to 14.28%. This is mainly due to the combined effects of the
quite different flow structures and particle inertia.
4.4. Trajectories for different dust sizes and wind velocities
To investigate dust deposition mechanisms for different dust
sizes and wind velocities, the trajectories of dust particles with 1,
50, 100, 150 and 250 μm sizes near the solar PV panel were obtained
and displayed in Fig. 10. In order to illustrate dust trajectories more
clearly, two hundred samples of dust trajectories were selected in
Fig. 10. From Fig. 10(a-d) for UHp 13 m/s, it can be observed that

dust particles with different sizes have significantly different
deposition behaviors. When dust size is 1 μm, it can be found that
the trajectories of dust particles are quite similar with the
streamlines of wind flow shown in Fig. 8. It indicates that the
movements of the dust particles almost follow the air flow due to
the low inertia. Thus the deposition rates are quite low, However,
the trajectories of dust are modified on certain degree as dust size
increases to 50 μm. More dust particles move towards the solar PV
panel for deposition by the interception, as shown in Fig. 10 (b ).
When dust particle size is 100 μm with the maximum deposition
rate, it can be seen that the move directions of most particles are
forward or downward to the PV panel due to the effects of gravity
and inertia. Thus a large number of dust particles deposit on solar
PV panel seen from Fig. 10 (c). Nevertheless, a majorty of dust
particles with 150 μm diameter has deposited on the ground before
reaching the solar PV panel. Therefore, the deposition rates for
150 μm dust are extremely low due to the intense effects of gravi
tation. In consequence, it can be concluded that medium size dust
particles (100 μm) have the maximum posibility to deposit on PV
panel for UHp 1.3 m/s, and the deposition rates of dust particles
rise up firstly and then fall down when dust size increases. When
the wind velocity increases to 2.6 m/s, the deposition trajectories
and mechanism are similar with the case oflower wind velocity, as
shown in Fig. 10 (e h). However, the dust size with maximum
deposition rate is increased to 150 μm for UHp 2.6 m/s.
4.5. Influences of dust deposition on PV efficiency
The relationship of the PV performance and dust density was
investigated carefully by Jiang et al. (25]. The empirical equation
established by the experimental study was adopted in this study, as
described by Ref. [ 25 I:

where E,ed and Ec1ean are PV power efficiency with and without dust
accumulation, respectively. Pdep and K are the density of deposited
dust and the fitting factor from the experiments, respectively. From
Jiang’s experimental study, the value ofKis different for different
kinds of PV modules. The reduction of output efficiency with dust
deposition density for different PV cells are shown in Fig. 11 (a). It
can be observed that dust deposition has the highest influences on
the output efficiency of poly crystalline silicon PV module, fol
lowed by amorphous silicon and mono crystalline silicon ones. For
individual module, the value of Kis 0.0115, 0.015 or 0.0139 for
mono crystalline silicon, poly crystalline silicon and amorphous
silicon PV cells, respectively (25].
Based on the present CFO simulation and the experimental
study of Jiang et al. (25 ], a simplified model was proposed by the
authors (33] to predict the PV efficiency degradation with exposure
time caused by dust:

where Np is the released the dust particle number in the time
period td. ). and Kare the dust deposition rate and influencing factor
of dust density on PV efficiency, respectively. Pp is dust density and
dp is the dust particle diameter. Sd is the PV panel area. T (day) is
exposure time.
The reduction ratio of the solar PV efficiency can be estimated by
the proposed empirical model shown Fig. 11 (b) and (c), if dust
number is 1000 per second and UHp 2.6 m/s. It can be observed

Fig. 10. Dust particle movement trajectories near the solar PV panel.

that the reduction ratios of the PV power caused by dust deposition
gradually rise with exposure time for all different dust particle
sizes. However, dust particles of different sizes have dramatically
different influences on solar PV efficiency for several orders of
magnitude. The reduction ratio of solar PV efficiency reaches 36.17%
after 100 days’ exposure for 200 mm dust particles, which is the
maximum influence on PV output. Although the maximum dust
deposition occurs with 150 mm diameter dust. This means that the

150 mm diameter dust has the largest deposition number. However,
the 200 mm diameter dust has the maximum deposition mass
because the larger diameter. Nevertheless, the minimum reduction
ratio of PV output efficiency is 2.42 X 10 -7 for dust particles with
1 mm diameter, which almost can be neglected. These huge differ
ence of influences on PV efficiency is because the deposition den
sity of large dust particles are much higher, compared with the
small ones.
To assess the effects of dust deposition on different kinds of PV
modules, the reduction ratios of PV efficiency caused by 200 mm
dust particles for mono crystalline silicon, poly crystalline silicon
and amorphous silicon PV cells were calculated and shown in
Fig. 11 (c). The logarithmic coordinate was used in the figure to
show the difference of the PV efficiency reduction more clearly. It
can be found that dust deposition has the highest influence on
poly crystalline silicon PV module, followed by amorphous silicon
one and mono crystalline silicon one. The reduction ratios of PV
output efficiency caused by 200 mmdust particles are 36.17%, 47.18%
and 43.72% for mono crystalline silicon, poly crystalline silicon and
amorphous silicon PV cells, respectively. Therefore, it can be
concluded that dust deposition have significant effects on the
output efficiency of different kinds of PV modules.
Effective solution strategies for dust pollution on solar PV panels
need to be developed. The existing method to mitigate dust
pollution on PV panels is mainly manual or mechanical cleaning by
water. However, the cleaning frequency of PV panels is always a
challenging problem as dust pollution situation is quite different in
different regions. The empirical model may be useful to estimate PV
power efficiency reduction by dust pollution in different regions
and then further determine cleaning frequency. Moreover, some
super hydrophilic or super hydrophobic coating are recently
developed and produced to prevent dust deposition on solar PV
panels. The findings of present study may be helpful for coating
development strategies. For example, the present study found that
the medium size dust has maximum deposition rate on PV panels.
Thus the self cleaning coating should be developed to mainly
prevent deposition of medium size dust.

( c) Influences of dust deposition on PV efficiency for different types of PV cells

Fig. 11. Theoretical prediction of effects of dust deposition on solar PV efficiency.

5. Conclusions
Dust deposition characteristics on an isolated ground mounted
solar PV system and its effects on the PV efficiency were investi
gated by CFD simulation. The turbulent wind flow fields over the
solar PV system, the dust deposition rates and processes on PV
panel with different particle sizes and wind velocities, as well as
their influences on the output efficiency reduction of different
kinds of PV modules were predicted and analyzed carefully. The
following conclusions can be drawn based on the present study.
1. For turbulent flow fields around the solar PV system, there are
two small turbulent eddies below and one large scale separa
tion eddy behind the solar PV system. The wind velocity is
intensely varied near the PV system and the flow streamlines
above the solar PV panel are parallel with PV surface. The TKE
values above the PV panel are higher than those of the sur
rounding fields. Therefore, the wind flow fields around the solar
PV system are quite complicated.
2. The dust deposition rates and behaviors on PV panel are sig
nificant different for different particle diameters. However, the
dust deposition profiles are similar for different wind velocities.
The medium size dust particles have high deposition rates
while the deposition rates are quite low for the small and large
size dust particles. The peak deposition rate is 13.71% for 100 mm
particles when UHp 1.3 m/s, However, the maximum deposi
tion rate can reach 14.28% for 150 mm particles.

3. The deposition mechanisms of dust particles with different sizes
on solar PV panel were analyzed. For samll dust, the movements
of the dust particles almost follow the air flow due to the low
inertia. However, the dust deposition rates are increased when
dust size increases because of the effects of interception, grav
itation and mass inertia. When dust size is large enough, most of
particles has deposited on the ground before reaching solar PV
panel due to intense effects of gravitation. Thus the dust depo
sition rates firstly increase and then decrease when dust size
increases.
4. The reduction ratios of PV efficiency caused by dust deposition
was predicted by the empirical model based on the present CFD
simulation and experimental measurement. The reduction of PV
efficiency caused by dust is gradually increased with exposure
time. However, dust particles with different sizes have
dramatically different influences on solar PV efficiency for
several orders of magnitude. This is because the deposition
density of large dust particles are much higher compared with
small particles. Besides, the influences of dust accumulation on
PV output efficiency are different for different types of PV cells.
Therefore, the simplified model used in the study can be applied
in the estimation of PV clean frequency for dust deposition in
practical engineering. Moreover, the effects of tilt angle on dust
deposition characteristics and PV power efficiency are quite sig
nificant, which will be investigated separately by numerical simu
lation and experimental measurement in our future work.
Acknowledgement
The authors appreciate the financial supports provided by the
National Natural Science Foundation of China (Grant Nos.
50876053).