You are on page 1of 8

Renewable Energy 90 (2016) 283e290

Contents lists available at ScienceDirect

Renewable Energy
journal homepage: www.elsevier.com/locate/renene

Investigation of wind speed cooling effect on PV panels in windy


locations
€ kmen, Weihao Hu*, Peng Hou, Zhe Chen, Dezso Sera, Sergiu Spataru
Nuri Go
Department of Energy Technology, Aalborg University, Pontoppidanstraede 101, Aalborg, Denmark

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

Article history: Environmental concerns have considerably increased the penetration of renewable energy sources in the
Received 3 September 2015 electricity grid. Especially, the quick rise of photovoltaic (PV) installations aroused more research in-
Received in revised form terests in efficiency improvement during the recent years. Even one percent more gain is of crucial
30 November 2015
importance for sustainable energy development, potential impacts of some parameters as wind speed
Accepted 2 January 2016
has not been taken into account broadly in PV systems yet. This paper is intended to help project
Available online xxx
planners to accurately estimate true potential of the PV plants especially in windy locations by taking
into account generally underestimated wind speed cooling effect. Firstly, optimum tilt angle variations
Keywords:
Optimization
have been investigated and secondly yearly energy comparisons are made for cases with and without
Photovoltaic (PV) systems considering wind speed. A more accurate mathematical model is given to estimate yearly energy gain
Solar radiation especially in the planning stage. Particle swarm optimization (PSO) algorithm is used to find optimum tilt
Temperature angles for several time intervals. Results are given in detail through a case study in a windy northern
Tilt angle European city, Aalborg, Denmark. The presented methodology can also easily be applied to other systems
Wind speed located around the world.
© 2016 Elsevier Ltd. All rights reserved.

1. Introduction especially in the planning stage, it is of great importance to accu-


rately estimate true potential of the PV systems and accordingly
Mankind has witnessed considerable social and economic sizing the required system devices [5].
development in the last two centuries thanks to electricity. Despite As it is well known, irradiance and temperature are two main
these benefits, increasing energy dependence came at a price of parameters which directly affect PV power generation. Best per-
negative environmental effects. Intensive usage of limited fossil formance can be achieved by minimizing temperature and maxi-
fuels has boosted CO2 emissions in the atmosphere [1]. Therefore, mizing solar radiation on a PV panel surface. Especially, irradiance
many efforts are being made to integrate more non-polluting en- is directly dependent on orientation (azimuth) and tilt angle of the
ergy sources to the electricity grid. As one of the promising options PV panel [6]. Solar tracking systems can be used to change these
and with many advantages, e.g. being clean, secure, noise free, and angles throughout the day to maximize falling solar radiation on PV
easy to install, solar PV energy is expected to play an important role module surfaces. They are thermal or electromechanical devices
in future sustainable energy development [2]. In Europe, the ca- which do actively change angular positions of solar panels ac-
pacity of photovoltaic (PV) plants has already passed 80 GW in 2013 cording to the sun’s position either on single-axis or dual-axis [7,8].
[3]. This upward trend is predicted to continue expeditiously by However, these devices increase installation and maintenance costs
force of increasing support, research and manufacturing experience of the system and may consume energy during tracking. Besides,
in time [4]. they are not always applicable due to installation area limitations.
However, although a lot of work has been performed to improve Therefore, it may be very beneficial to change the orientation and
PV system efficiency, initial investment cost is still relatively high tilt angle of solar panels manually in a less frequent manner such as
compared to the traditional power generation sources. Therefore, monthly, seasonally or semi-yearly.
In the northern hemisphere for best azimuth angle, generally
solar panels are oriented south (i.e. facing equator), but optimum
tilt angle changes significantly according to geographical location,
* Corresponding author. meteorological conditions, and utilization period of time.
E-mail address: whu@et.aau.dk (W. Hu).

http://dx.doi.org/10.1016/j.renene.2016.01.017
0960-1481/© 2016 Elsevier Ltd. All rights reserved.
284 €kmen et al. / Renewable Energy 90 (2016) 283e290
N. Go

Therefore, optimum tilt angles should be calculated for each spe- surface: beam (direct), diffuse, and ground-reflected irradiance.
cific site separately [9]. Beam radiation is the component which directly comes from sun
Thus far, a lot of work has been performed to find optimum tilt without being scattered through the atmosphere. Contrary, diffuse
angles for specific sites in several countries such as Australia [10], radiation is highly scattered by different molecules, clouds, dust or
Canada [11], Egypt [12], Germany [13], Greece [14], Iran [9], Ireland haze in the atmosphere. On the other hand, ground-reflected ra-
[15], Malaysia [16], Reunion Island [1], Saudi Arabia [17], Singapore diation is formed by both beam and diffuse radiation which are
[18], Syria [19], Taiwan [20], Turkey [21], and United States [22]. reflected from the ground and reach to the surface again.
Commonly two techniques are being used to determine optimal tilt Up to now, different models have been proposed to find the total
angles: 1) using latitude information (rule of thumb); and 2) radiation on a tilted surface [31]. Beam radiation calculation is
maximizing falling solar radiation on a tilted panel surface. Due to purely geometric and quite straightforward so that there is no
considering of only direct radiation, first method is more appro- difference between the models. Reflected radiation in reality is
priate for mostly sunny locations and generally, second method quite complex and it depends on many factors such as surface to-
gives better results for cloudy locations because of additionally pology, zenith angle, direct and diffuse radiation. However, as most
taking into account both diffuse and reflected radiation. of the models do in literature this component can be treated as
On the other hand, temperature is the second most dominant isotropic if the albedo is not too high and the tilt angle is moderate
parameter in PV systems which do affect performance of the solar [32]. The only difference is emerging from the treatment of sky-
cells seriously. Manufacturers generally give PV module specifica- diffuse component. Basically, diffuse radiation models can be
tions under standard test conditions (STC), i.e. 1000 W/m2 irradi- classified into two groups as isotropic and anisotropic models.
ance, 1.5 AM and 25  C module temperature. However, under real Experimental studies show that accuracy of isotropic models is not
operating conditions currentevoltage characteristics of PV mod- good enough to estimate the diffuse radiation component on tilted
ules may change significantly according to the solar irradiance, surfaces by virtue of assuming the density of diffuse radiation is
ambient temperature, wind speed, and different mounting types. uniform all over the sky dome [33]. Therefore, anisotropic models
Although in literature, commonly just solar irradiance and ambient are improved which also take into account circumsolar diffuse and
temperature are used to estimate the operating temperature of PV horizon-brightening components. In this study, widely known and
modules, in recent studies, it is reported that wind speed may also accepted diffuse radiation model of Perez is used [34].
contribute to improve working efficiency by reducing solar cell Beam radiation on a tilted surface can be calculated with the
temperature due to its cooling effect [23-27]. Yet, it has not been following formula:
broadly applied to PV systems’ power calculations. A few of studies
a
take wind speed into account but in mostly sunny and less windy Gb;b ¼ Gb , (1)
regions [28-30]. Especially, for very windy locations, the contribu- b
tion from wind speed may be envisaged more compared to the less
windy locations. a maxð0; cos qÞ
¼ (2)
In this paper, it is intended to help project planners to accurately b maxð0:087; cos qz Þ
estimate true potential of the PV plants especially in windy loca-
tions by taking into account generally underestimated wind speed where Gb is horizontal beam radiation, q is incidence angle, and qz is
cooling effect. Firstly, optimum tilt angle variations have been zenith angle.
investigated and secondly yearly energy comparisons are made Secondly, the diffuse radiation is given as:
when tilt angles are changed in some specific time periods with and    
1 þ cos b a
without considering wind speed. Particle swarm optimization Gb;d ¼ Gd ð1  F1 Þ þ F1 þ F2 sin b (3)
2 b
(PSO) algorithm is used to find the optimum tilt angles in Aalborg
by using yearly high resolution data. Moreover, the proposed
where Gd is horizontal diffuse radiation, b is tilt angle, F1 and F2 are
technique is also experimentally validated for two different days
statistically derived coefficients which express the degree of cir-
one in winter and one in summer. Following this introductory part,
cumsolar and horizon anisotropy respectively. These coefficients
mathematical models for irradiance, temperature and power
are functions of zenith angle, sky clearness and brightness indexes
calculation are given in Section 2. Optimization procedure is
which can easily be calculated from given table in Perez’s study
explained in detail in Section 3. Study location, data, and experi-
[34].
mental work are introduced in Section 4 and finally results are
As third component, ground-reflected radiation can be calcu-
given together with discussions in Section 5.
lated with this simple formula:
2. Mathematical models  
1  cos b
Gb;r ¼ ðGb þ Gd Þ,r, (4)
2
In this section mathematical models are given to estimate total
solar radiation on tilted surfaces and the operating temperature of where r is reflectance (albedo) of the ground and generally a con-
PV modules. Then, PV array power model is formulated by stant value of 0.2 is taken in literature.
considering all important system parameters. Finally, total radiation on a tilted surface can be computed by
summation of beam, diffuse and ground-reflected irradiance as:
2.1. Estimation of total solar radiation on tilted surfaces
Gb ¼ Gb;b þ Gb;d þ Gb;r (5)
Generally, meteorological stations record global (total) and
diffuse irradiance components on horizontal surfaces. However, PV
modules or solar collectors need to be tilted with respect to the
horizontal plane for better performance. The total irradiance may 2.2. Estimation of PV module temperature
be measured on a tilted surface or it may be accurately estimated
from these available horizontal irradiance components. Three Temperature is another important parameter which directly
components are needed to determine total irradiance on a tilted affects power output of a PV system. It depends on many factors but
€kmen et al. / Renewable Energy 90 (2016) 283e290
N. Go 285

mostly to ambient air temperature (Ta), plane of array irradiance 3. Optimization procedure
(Gb), mounting type of PV arrays (u), and also wind speed (Vf).
Generally, in literature the following explicit equation is widely In this section, firstly optimization algorithm is introduced and
being used for estimation of the operating temperature of a PV then yearly energy calculation is given. In the last part objective
module [35]: function and constraint are expressed.
 
Tnoct  20
Tc ¼ Ta þ ,Gb (6) 3.1. Particle swarm optimization (PSO)
800
Although this equation may give accurate results for low windy In recent years, modern optimization techniques have been
regions, it is not precise enough for high wind locations as in case of developed to solve complex engineering problems. Particle swarm
northern European countries. Therefore, in this study a more pre- optimization (PSO) is one of the metaheuristic algorithms originally
cise temperature estimation formula is used that considers wind proposed by Kennedy and Eberhart in 1995 and was first intended
speed parameter and also different mounting types [36]: to simulate social behaviour of bird swarms [38]. Here, the word
particle denotes a bird in a flock. Each particle has a position and
!
0:32 velocity vector according to their location in a search space. They
Tc ¼ Ta þ u, ,Gb (7) try to find best location for food individually but also want to move
8:91 þ 2Vf
together in swarm intelligence. At every turn, they communicate
with each other about best locations and accordingly swarm up-
where u is mounting coefficient and given as 1, 1.2, 1.8 and 2.4 for dates its location by deciding the finest (global best) among these.
free standing, flat roof, sloped roof, and façade integrated installa- This behaviour has been imitated successfully to find the global
tion types, respectively [35] and Vf is free-stream wind speed. maximum point of an objective function in engineering problems.
Wind speed value can be converted to the required height ac- PSO algorithm used in this study can be expressed as [39]:
cording to measured height by using following power law [37]:
   
!n vkþ1
i ¼ w,vki þ c1 ,r1 pbestik  xki þ c2 ,r2 gbest k  xki (12)
Vf z
¼ (8)
Vref zref
xkþ1
i ¼ xki þ vkþ1
i for i ¼ 1; 2; …; m: (13)
 
0:37  0:0881 ln Vref w ¼ 0:5 þ ðr3 =2Þ (14)
n¼  .  (9)
1  0:0881 ln zref 10 where w is inertia weight which balances local search and global
search abilities, c1 and c2 are acceleration coefficients usually set
where zref is reference height measured from the ground, z is the equal to 2, r1, r2 and r3 are random numbers in the range of [0, 1], m
required height in meters and Vref is the known wind speed at the is the number of particles in the population, k is current iteration
reference height. number, xki and vki are the position and velocity of the ith particle at
iteration k, respectively. pbestki represents locally best of ith particle
and finally gbestk denotes globally best of swarm at iteration k.
2.3. PV array power model
3.2. Yearly energy calculation
Power production from grid-connected or stand-alone PV sys-
tems depends on many parameters such as PV array peak power In this study, 15-min resolution irradiance, ambient tempera-
(Ppeak), solar radiation on tilted module surface (Gb), PV module ture, and wind speed data is used for a whole year which makes a
operating temperature (Tc), inverter efficiency (hinv), and losses total number of 35040 data values. 15-minute, half-hourly, hourly,
(hloss). PV module efficiency model is given as: daily, monthly, seasonally, semi-yearly and yearly time periods are
   considered to periodically change the tilt angle. Power is assumed
hpv ¼ hinv ,hloss ,href , 1  kT Tc  Tref (10) constant in 15-min intervals and yearly energy is calculated from
summation of these power values for the entire year. Output power
of a PV module given in (11) is basically functions of ambient
where href is the reference PV module efficiency at a cell tempera-
temperature, wind speed, irradiance components, and tilt angle.
ture of 25  C (Tref) and kT is temperature coefficient. These param-
Therefore, 15-min power values can be calculated according to the
eters depend on solar cell material and can be found on the
following equation:
datasheets of manufacturers. Here, another important issue is the
inverter efficiency which is not constant. It fluctuates according to  
15min
the input power or voltage fed into it. Especially during sunrise and Ppv ¼ hinv ,hloss ,href ,Apv ,f Ta15min ; Vf15min ; b
  (15)
sunset hours inverter efficiency may decrease significantly.
,g G15min
b ; G15min
d ; G15min
r ;b
Generally, manufacturers provide efficiencyepower curve so it
should be added to the PV array power model.
Based on these parameters, output power of a PV array which where f function denotes temperature losses 1-kT(Tc-Tref) in (10) and
has an area of Apv (m2) is given by: g function returns total radiation on a tilted surface (1e5).
According to each tilt angle changing period yearly energy value
Ppv ¼ hpv ,Apv ,Gb (11) is computed by using (15) as given in Table 1. Here it is important to
remember that day numbers are not equal for each month. So this
The values of the PV system parameters used in this study are fact is also taken into account and day indices (d1 and d2) are
0.86, 0.12, 0.004, 6.667 and 1, for hloss, href, kT, Apv, and u, created for monthly, seasonally, and semi-yearly time periods as in
respectively. Table 2.
286 €kmen et al. / Renewable Energy 90 (2016) 283e290
N. Go

Table 1
Yearly energy calculation for each time period

Time period Total tilt angle number Yearly energy formulation

15 minute 35040 yearly P


35040 15min
Epv15min ¼ i¼1 Ppvi
30 minute 17520 yearly P17520 P2i
Epv30min ¼ i¼1 j¼1þ2ði1Þ Ppvi;j
15min

1 hour 8760 yearly P8760 P4i 15min


Epvhour ¼ i¼1 j¼1þ4ði1Þ Ppvi;j
1 day 365 yearly P P96i
Epvday ¼ 365 i¼1 j¼1þ96ði1Þ Ppvi;j
15min

1 month 12 yearly P12 P96d2


Epvmonth ¼ i¼1 j¼1þ96d Ppv15min i;j
1

1 season 4 P4 P96d2
Epvyearly
season ¼ i¼1 j¼1þ96d1
Ppv15min
i;j
1 semi-year 2 P P 2
Epvyearly
semiyear
¼ 2i¼1 96d j¼1þ96d1
Ppv15min
i;j
1 year 1 P1 P35040
Epvyearly
year ¼ i¼1 j¼1 Ppvi;j
15min

Table 2
Day indices for yearly energy calculation

Time interval d1 d2 Time interval d1 d2

Monthly Seasonally
Jan 0 31 Winter 0 59
Feb 31 59 334 365
Mar 59 90 Spring 59 151
Apr 90 120 Summer 151 243
May 120 151 Autumn 243 334
Jun 151 181
Jul 181 212
Aug 212 243
Sep 243 273 Semi-yearly
Oct 273 304 Winter 0 90
Nov 304 334 273 365
Dec 334 365 Summer 90 273

3.3. Objective function and constraints

The objective is finding optimum tilt angles which maximize


Fig. 1. Particle swarm optimization algorithm.
yearly energy for each tilt angle changing periods as given in Table 1
and tilt angle is allowed to change between 0 and 90 as an only
constraint. horizontal irradiation, air temperature at 2 m, and wind speed at
Objective: 10 m height. A yearly average of 8.6  C air temperature and 4.4 m/s
n o wind speed is recorded at the location. As it is expected from a
yearly
max Epvfor each time period
(16) northern European city, 49.5% of the total solar radiation is coming
as diffuse radiation due to the cloud coverage. Yearly sums for
Constraint: diffuse and global horizontal radiation are recorded 505 and
1020 kW h/m2, respectively.
0  b  90 (17)
Moreover, experimental data from a PV array which consists of
Flowchart of the PSO algorithm used in this study is given in 8 x 2 BP MSX 120 PV modules (1.9 kWp) is also used to test the
Fig. 1. For this particular case, swarm size is chosen 55, the lower proposed method. It is mounted on rooftop of PV systems labora-
and upper boundaries of positions which corresponds to tilt angles tory which is 10 m above of the ground at Department of Energy
are chosen 0 and 90, respectively and maximum velocity is deter- Technology in Aalborg University. PV modules are poly-crystalline
mined according to the range of particle positions which is (90-0)/ with nominal efficiency of 12% and 120 Wp power and di-
5. mensions of each are 1.097 x 0.981 x 0.048 m. Temperature coeffi-
In the beginning of the main algorithm, optimization parame- cient of power is given (0.5 ± 0.05)%/ C by the manufacturer.
ters are initialized and according to the chosen tilt angle changing Irradiance is measured on plane of array (tilted 45 ) using the Si-
period, tilt angles are randomly being sent to the fitness function. It RS485-TC-2T-v, digital silicon irradiance sensor which is able to
calculates solar radiation and solar cell temperature and then gives measure in the range of 0e1400 W/m2 with ±5% error. Module
15-min power values. After that, the fitness function returns yearly temperature is also measured with a built-in temperature sensor in
energy value to the main algorithm. The process continues until this device with an accuracy of ±1  C. For ambient temperature a
there is no any significant improvement in global best value and Pt1000 sensor is used with accuracy of ±1%. Wind speed is
finally main algorithm returns optimum tilt angles. measured at 2 m above of the PV array with Thies Clima small wind
sensor in 0e40 m/s range and ±0.5 m/s error. The system is also
4. Case study and experimental work grid connected through a Sunny Boy SB 1700 inverter and AC power
is measured by using Yokogawa WT3000 power analyser with a
In order to test the proposed method, Aalborg city (57 020 N / power accuracy of 0.02%.
9 550 E),
Denmark is chosen as a study case. For optimization pro- Two different characteristic days, one in winter 11 Dec 2010 and
cess, 15-min time series data (35040 records) of year 2013 is pur- one in summer 7 Aug 2011 are chosen for the experimental study
chased from SolarGIS [40]. Data contain global and diffuse and 15 min averaged data are recorded for these days. For the
€kmen et al. / Renewable Energy 90 (2016) 283e290
N. Go 287

optimization process Matlab simulation environment is used and temperature difference, it would be better to show the cooling ef-
detailed analyses of the results are given in following section. fect of wind speed to the PV module operating temperature as in
Fig. 3(a). Here, simulation results show that wind speed helps solar
cells to operate much cooler. In traditional way (6), due to not
5. Results and discussion
considering cooling effect of the wind speed, at some days module
temperature is overestimated up to 11  C with reference to (7) [see
5.1. Wind speed cooling effect on optimum tilt angles
Fig. 3(b)]. According to PV module power temperature coefficient,
this 11  C difference in other words means that daily energy gen-
Optimum tilt angles are calculated with and without consid-
eration is being underestimated around 4.5% during some days.
ering wind speed for monthly, seasonally, semi-yearly and yearly
Especially, during hot periods as in spring and summer, this dif-
time periods as given in Table 3. Here, even though it is not rec-
ference is much more compared to the cold periods.
ommended for making a direct comparison due to using different
In Table 4, yearly energy values are compared for the cases with
solar radiation and temperature database, output from PVGIS
and without considering wind speed for different tilt angle
model is [41] also given to get an overall idea about the results. It
changing periods. From the table, it may be deducted that no
should also be noted that PVGIS model use a simple temperature
matter which period is chosen for changing the tilt angle, taking
estimation formula which does not take into account cooling effects
into account wind speed always increases energy estimates by 3.5%
of the wind speed.
for this location. In other words, when wind speed is not taken into
From monthly results, it can be seen that the difference between
account as an important parameter, yearly energy is under-
PVGIS model and the proposed model is more than 4 for February,
estimated by about 3.5%.
August and October. However, the reason of the difference (could be
By using the data in Table 4, also the effect of changing tilt angles
due to different solar radiation and/or ambient temperature pro-
in some time intervals can be investigated. If fixed installation type
files) cannot be detected without using the same PVGIS datasets.
is chosen as reference, energy gain ratios can be calculated 2.3%,
Therefore, the contribution from wind speed may be investigated by
2.2%, 3.1%, 4.2%, 6.2%, 6.4%, and 6.5% for semi-yearly, seasonally,
checking the other two columns in the table. According to the
monthly, daily, hourly, 30-min and 15-min tilt angle changing pe-
observation, the angles calculated by considering wind speed are
riods, respectively. Here, an interesting phenomenon can be
always bigger than the angles where the wind speed is not taken
noticed between semi-yearly and seasonally tilt angle changing
into account. Minimum difference becomes 0.11 in December
periods. Even though, yearly energy gain generally follows an
while the maximum reaches 0.80 in May. The same behaviour can
increasing trend when the changing period is decreased, it might
also be seen in seasonal period with minimum 0.18 in winter and
be observed that, exceptionally, semi-yearly period is 0.1% more
0.83 in summer. Six-month changing periods are following the
efficient than the seasonal period. In literature, it has been
same trend with 0.11 and 0.53 in OctobereMarch and AprileSep-
encountered with another study which reports that semi-yearly
tember period, respectively. For those who install PV systems at
adjustment gives a slightly better performance compared to the
fixed tilt angle a difference of 0.34 can be seen from the table.
seasonally adjustment [42]. However, no additional explanation
The variation of the optimum tilt angle for different time periods
has been made for this behaviour. Although it is not a significant
is also illustrated in Fig. 2. Fluctuations seen in daily period are
difference, the reason for this phenomenon may be consequence of
mostly due to cloud coverage and they may follow a different
not separating seasons optimally.
pattern for another year. Thus, to change the tilt angles for smaller
time periods, special tracking systems with real time controllers are
required and this may not be a cost effective solution. From the
5.3. Experiment
beginning until to the midpoint of the year gradually decreasing
pattern of the optimum tilt angles can be observed. This pattern
In order to validate the models also experimental data have
changes its direction in opposite way after mid of the year. Addi-
been used as introduced in Section 4. First of the experiments is
tionally, for those who want to install PV modules at fixed tilt angle,
from a very windy and cold winter day, 11 Dec 2010. Day length is
yearly optimum angle is calculated as 41.08 .
very short (6 h and 50 min) for this day, the time for sunrise and
sunset is 08:48 and 15:39, respectively. During the daylight an
5.2. Wind speed cooling effect on energy production average of 146.5 W/m2 irradiance, 7.6  C ambient temperature, and
15.2 m/s wind speed value is recorded. As it may be seen in Fig. 4
Firstly, for the reason that results are directly dependent on when the wind speed is not considered, the temperature of solar

Table 3
Optimum tilt angle comparison for each time period

Tilt angle changing Time bopt PVGIS ( ) bopt NCWS ( ) bopt CWS ( ) Tilt angle changing Time bopt PVGIS ( ) bopt NCWS ( ) bopt CWS ( )
period interval period interval

Monthly Jan 73 71.37 71.57 Seasonally Winter - 68.22 68.40


Feb 65 60.52 60.83 Spring - 41.54 42.12
Mar 57 56.22 56.72 Summer - 24.44 25.27
Apr 42 41.11 41.67 Autumn - 56.48 56.67

May 26 24.92 25.72 Semi-yearly Oct-Mar - 62.10 62.21


Jun 19 19.03 19.78
Jul 22 23.92 24.49 Apr-Sep - 30.77 31.30
Aug 36 31.41 31.99
Sep 49 47.91 48.17 Yearly Jan-Dec 42 40.74 41.08
Oct 63 58.18 58.48
Nov 72 71.77 71.92
Dec 77 76.28 76.39

NCWS: Not considering wind speed, CWS: Considering wind speed


288 €kmen et al. / Renewable Energy 90 (2016) 283e290
N. Go

Fig. 2. (a) Daily variation of the optimum tilt angle. (b) Variation of the optimum tilt angle for longer time periods.

Fig. 3. (a) Comparison of temperature estimation functions. (b) Daily overestimated temperature values when traditional approach is applied.

Table 4
Comparison of yearly energy values: with and without wind

Optimum tilt changing period Epv yearly (kWh) (without wind) Epv yearly (kWh) (with wind) Relative difference (%)

15 minute 1056 1093 3.55


30 minute 1055 1092 3.55
1 hour 1053 1091 3.55
1 day 1033 1069 3.54
1 month 1023 1059 3.53
1 season 1013 1049 3.52
1 semi-year 1014 1050 3.53
1 year 991 1026 3.51

cells are overestimated up to 10  C more according to the measured ambient temperature, and 5.9 m/s wind speed value is recorded. As
values at some specific hours of the day. Root mean square error it may be seen in Fig. 5, the difference between measured tem-
(RMSE) for temperature estimation functions are calculated 4.47  C perature and traditional estimation method reaches up to 25  C
for (6) and 0.37  C for (7). For this day, measured and simulated around the midday. Root mean square error (RMSE) for the tem-
power values are also given in the figure. At the end of the day, perature estimation functions are calculated 11.15  C and 2.69  C for
287.7 W h energy is measured in the system. According to the (6) and (7), respectively. For the power performance comparison
temperature error in traditional method, daily energy is under- right side of Fig. 5 can be analysed. Especially, in hot periods during
estimated 2% less (281.9 W h) whereas in proposed method it is the midday, simulation error is much more in comparison with the
overestimated 0.5% (289.1 W h). measured values. At the end of the day, total 841.2 W h energy is
Second experimental data is from a hot and less windy summer measured in the system. Daily energy is underestimated 4.4% less
day, 7 Aug 2011. Day length is very long (15 h and 50 min) for this (804.1 W h) in traditional method whereas in proposed method it is
day, the time for sunrise and sunset is 05:30 and 21:21, respectively. overestimated 0.7% (847 W h).
During the daylight an average of 373.4 W/m2 irradiance, 20  C Experiment results show that proposed method is much more
€kmen et al. / Renewable Energy 90 (2016) 283e290
N. Go 289

Fig. 4. Comparison of experiment and simulation for temperature and power profile in a winter day (11 Dec 2010)

Fig. 5. Comparison of experiment and simulation for temperature and power profile in a summer day (7 Aug 2011)

effective and accurate to estimate the power production than the the system devices and may reduce the overall cost of the PV sys-
traditional method. Especially, it can be seen that power estimation tems. Especially, at high altitudes such as hills and mountains, or
error of the traditional method increases significantly during the very windy locations, the contribution from wind speed should not
hot periods of the year compared to cold periods. Therefore, be ignored. Lastly, bearing in mind that presented methodology
contribution of the wind speed cooling effect particularly in windy should be applied for each installation site respectively, because
locations should not be ignored in PV system calculations. different local weather conditions can considerably change the
given results in this study.

6. Conclusions
Acknowledgment

This study reveals that how significant wind speed can affect a
This work was supported in part by the Danish Strategic
design parameter and operating performance of a PV system. As a
Research Center under Grant DSF 09-067255, “Development of a
design parameter, the tilt angle and as a performance parameter,
Secure, Economic, and Environmentally-friendly Modern Power
the power output of a PV array has been investigated with the
System” (SEEMPS) and in part by the Program for Professor of
situation in a windy northern European city Aalborg, Denmark.
Special Appointment (Eastern Scholar) at Shanghai Institutions of
From the tilt angle analysis, it can be concluded that especially
Higher Learning.
summer period angles are affected much more compared to the
winter period angles. The difference between calculated optimum
8. References
tilt angles with traditional and proposed methods varies between
0.11 and 0.80 for monthly, 0.18 and 0.83 for seasonally, 0.11 and [1] M. Bojic, D. Bigot, F. Miranville, A. Parvedy-Patou, J. Radulovic, Optimizing
0.53 for semi-yearly tilt angle changing periods. In fixed in- performances of photovoltaics in Reunion Islands-tilt angle, Prog. Photovol-
stallations a difference of 0.34 can be noticed for the study loca- taics Res. Appl. 20 (2012) 923e935.
[2] D. Yang, H. Yin, Energy conversion efficiency of a novel hybrid solar system for
tion. On the other hand, it is observed that when wind speed is not photovoltaic, thermoelectric, and heat utilization, IEEE Trans. Energy Convers.
taken into account, yearly energy is being underestimated by 3.5%. 26 (2011) 662e670.
This difference is emerging from overestimating PV module tem- [3] EPIA, Global market outlook for photovoltaics 2014-2018. [Online]. Available:
http://www.cleanenergybusinesscouncil.com/global-market-outlook-for-
perature by not considering cooling effect of wind speed as in
photovoltaics-2014-2018-epia-2014 (accessed on 22.11.15).
traditional method. Here, it is important to emphasize that wind [4] X. Gong, M. Kulkarni, Design optimization of a large scale rooftop photovoltaic
would exist whether it is to be considered or not. The useful system, Sol. Energy 78 (2005) 362e374.
conclusion would be that wind speed should be considered when [5] F. Bizzarri, M. Bongiorno, A. Brambilla, G. Gruosso, G.S. Gajani, Model of
photovoltaic power plants for performance analysis and production forecast,
the energy production is assessed especially in the planning stage. IEEE Trans. Sustain. Energy 4 (2013) 278e285.
This result could help project planners to prevent oversizing of [6] T. Khatib, A. Mohamed, K. Sopian, A review of solar energy modeling
290 €kmen et al. / Renewable Energy 90 (2016) 283e290
N. Go

techniques, Renew. Sustain. Energy Rev. 16 (2012) 2864e2869. M. Zebisch, A. Tetzlaff, Wind effect on PV module temperature: analysis of
[7] A. Dolara, F. Grimaccia, S. Leva, M. Mussetta, R. Faranda, M. Gualdoni, Per- different techniques for an accurate estimation, Energy Procedia 40 (2013)
formance analysis of a single-axis tracking PV system, IEEE J. Photovoltaics 2 77e86.
(2012) 524e531. [26] G.M. Tina, S. Scrofani, Electrical and thermal model for PV module tempera-

[8] S. Seme, G. Stumberger, J. Vorsi
c, Maximum efficiency trajectories of a two- ture evaluation, in: Proceedings of the Mediterranean Electrotechnical Con-
axis sun tracking system determined considering tracking system consump- ference - MELECON, 2008, pp. 585e590.
tion, IEEE Trans. Power Electron. 26 (2011) 1280e1290. [27] E. Skoplaki, J.A. Palyvos, “On the temperature dependence of photovoltaic
[9] H. Khorasanizadeh, K. Mohammadi, A. Mostafaeipour, Establishing a diffuse module electrical performance: a review of efficiency/power correlations, Sol.
solar radiation model for determining the optimum tilt angle of solar surfaces Energy 83 (2009) 614e624.
in Tabass, Iran, Energy Convers. Manag. 78 (2014) 805e814. [28] T. Bhattacharya, A.K. Chakraborty, K. Pal, Effects of ambient temperature and
[10] R. Yan, T.K. Saha, P. Meredith, S. Goodwin, Analysis of yearlong performance of wind speed on performance of monocrystalline solar photovoltaic module in
differently tilted photovoltaic systems in Brisbane, Australia, Energy Convers. Tripura, India, J. Sol. Energy 2014 (2014).
Manag. 74 (2013) 102e108. [29] A.J. Veldhuis, A. Nobre, T. Reindl, R. Ruther, A.H.M.E. Reinders, The influence of
[11] I.H. Rowlands, B.P. Kemery, I. Beausoleil-Morrison, Optimal solar-PV tilt angle wind on the temperature of PV modules in tropical environments, evaluated
and azimuth: an Ontario (Canada) case-study, Energy Policy 39 (2011) on an hourly basis, in: Proceedings of IEEE Photovoltaic Specialists Conference
1397e1409. - PVSC, 2013, pp. 824e829.
[12] H.K. Elminir, A.E. Ghitas, F. El-Hussainy, R. Hamid, M.M. Beheary, K.M. Abdel- [30] G. Xydis, The wind chill temperature effect on a large-scale PV plant-an
Moneim, Optimum solar flat-plate collector slope: case study for Helwan, exergy approach, Prog. Photovoltaics Res. Appl. 21 (2013) 1611e1624.
Egypt, Energy Convers. Manag. 47 (2006) 624e637. [31] S.A. Khalil, A.M. Shaffie, A comparative study of total, direct and diffuse solar
[13] S. Beringer, H. Schilke, I. Lohse, G. Seckmeyer, Case study showing that the tilt irradiance by using different models on horizontal and inclined surfaces for
angle of photovoltaic plants is nearly irrelevant, Sol. Energy 85 (2011) Cairo, Egypt, Renew. Sustain. Energy Rev. 27 (2013) 853e863.
470e476. [32] C.A. Gueymard, Direct and indirect uncertainties in the prediction of tilted
[14] J. Kaldellis, K. Kavadias, D. Zafirakis, Experimental validation of the optimum irradiance for solar engineering applications, Sol. Energy 83 (2009) 432e444.
photovoltaic panels’ tilt angle for remote consumers, Renew. Energy 46 [33] A.K. Yadav, S.S. Chandel, Tilt angle optimization to maximize incident solar
(2012) 179e191. radiation: a review, Renew. Sustain. Energy Rev. 23 (2013) 503e513.
[15] S. Armstrong, W.G. Hurley, A new methodology to optimise solar energy [34] R. Perez, P. Ineichen, R. Seals, J. Michalsky, R. Stewart, Modeling daylight
extraction under cloudy conditions, Renew. Energy 35 (2010) 780e787. availability and irradiance components from direct and global irradiance, Sol.
[16] M.A. bin, H.M. Yakup, A. Malik, Optimum tilt angle and orientation for solar Energy 44 (1990) 271e289.
collector in Brunei Darussalam, Renew. Energy 24 (2) (Oct. 2001) 223e234. [35] M. Alsayed, M. Cacciato, G. Scarcella, G. Scelba, Multicriteria optimal sizing of
[17] M. Benghanem, Optimization of tilt angle for solar panel: case study for photovoltaic-wind turbine grid connected systems, IEEE Trans. Energy
Madinah, Saudi Arabia, Appl. Energy 88 (2011) 1427e1433. Convers. 28 (2013) 370e379.
[18] Y.S. Khoo, A. Nobre, R. Malhotra, D. Yang, R. Ruther, T. Reindl, A.G. Aberle, [36] E. Skoplaki, A.G. Boudouvis, J.A. Palyvos, A simple correlation for the operating
Optimal orientation and tilt angle for maximizing in-plane solar irradiation temperature of photovoltaic modules of arbitrary mounting, Sol. Energy
for PV applications in Singapore, IEEE J. Photovoltaics 4 (2014) 647e653. Mater. Sol. Cells 92 (2008) 1393e1402.
[19] K. Skeiker, Optimum tilt angle and orientation for solar collectors in Syria, [37] C.G. Justus, A. Mikhail, Height variation of wind speed and wind distributions
Energy Convers. Manag. 50 (2009) 2439e2448. statistics, Geophys. Res. Lett. 3 (1976) 261e264.
[20] Y.M. Chen, C.H. Lee, H.C. Wu, “Calculation of the optimum installation angle [38] J. Kennedy, R. Eberhart, Particle swarm optimization, in: Proceedings of IEEE
for fixed solar-cell panels based on the genetic algorithm and the simulated- International Conference on Neural Networks 4, 1995, pp. 1942e1948.
annealing method, IEEE Trans. Energy Convers. 20 (2005) 467e473. [39] Y.H. Liu, S.C. Huang, J.W. Huang, W.C. Liang, A particle swarm optimization-
[21] K. Bakirci, General models for optimum tilt angles of solar panels: Turkey case based maximum power point tracking algorithm for PV systems operating
study, Renew. Sustain. Energy Rev. 16 (2012) 6149e6159. under partially shaded conditions, IEEE Trans. Energy Convers. 27 (2012)
[22] M. Lave, J. Kleissl, Optimum fixed orientations and benefits of tracking for 1027e1035.
capturing solar radiation in the continental United States, Renew. Energy 36 [40] Solargis. Online data and tools for solar energy projects. [Online]. Available:
(2011) 1145e1152. http://solargis.info/, Accessed on: Feb. 22, 2015.
[23] E. Kaplani, S. Kaplanis, Thermal modelling and experimental assessment of [41] PVGIS. Photovoltaic Geographical Information System. [Online]. Available:
the dependence of PV module temperature on wind velocity and direction, http://re.jrc.ec.europa.eu/pvgis/, Accessed on: Nov. 22, 2015.
module orientation and inclination, Sol. Energy 107 (2014) 443e460. [42] S.F. Khahro, et al., Evaluation of solar energy resources by establishing
[24] J.K. Kaldellis, M. Kapsali, K.A. Kavadias, Temperature and wind speed impact empirical models for diffuse solar radiation on tilted surface and analysis for
on the efficiency of PV installations. Experience obtained from outdoor optimum tilt angle for a prospective location in southern region of Sindh,
measurements in Greece, Renew. Energy 66 (2014) 612e624. Pakistan, Int. J. Electr. Power & Energy Syst. 64 (2015) 1073e1080.
[25] C. Schwingshackl, M. Petitta, J.E. Wagner, G. Belluardo, D. Moser, M. Castelli,

You might also like