Renewable Energy 162 (2020) 1747e1762

Contents lists available at ScienceDirect

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

Solar tree design framework for maximized power generation with

minimized structural cost
Sumon Dey a, c, *, Bala Pesala b, c
a
Council of Scientific and Industrial Research - Structural Engineering Research Centre, Chennai, 600113, India
b
Council of Scientific and Industrial Research - Central Electronics Engineering Research Institute, Chennai, 600113, India
c
Academy of Scientific and Innovative Research (AcSIR), Ghaziabad, 201002, India

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

Article history: Conventional solar trees, inspite of their high-power density, are relatively unviable due to the huge
Received 4 March 2020 shading losses (>30%) and structural cost (>50% of the total cost) associated with them. The paper
Received in revised form proposes a location specific design framework for maximized electrical output from solar tree using
14 June 2020
minimized structural material. Actual solar insolation data is utilized to orient solar panels in a solar tree
Accepted 6 July 2020
Available online 4 August 2020
to maximize annual energy extraction. Preliminary structural optimization coupled with multi-objective
optimization leveraging genetic algorithms with objectives as minimization of shading losses and pro-
jected ground footprint area are employed to position the solar panels. Further, the dimensions of the
Keywords:
Solar tree
branches and trunks in the solar tree structure are individually tuned to ascertain structural stability at
Structural optimization high wind speed using finite element modelling based von Mises stress analysis. The designed 3 kW solar
Shading losses tree, having a normalized ground footprint of 1.67 and shading loss of only 0.17% demonstrates the
Solar panel orientation design framework. Energy generation estimates are validated using ray-optic simulations. Simultaneous
structural optimization carried out to withstand a wind speed of 150 kmph, has resulted in 20% reduction
in structural mass requirement. The study increases the feasibility of deployment of solar tree and can be
extended to other geographical locations.
© 2020 Elsevier Ltd. All rights reserved.

1. Introduction extraction. As the latitude of location increases, the angle of

installation usually increases and thereby, the power density in-
Solar Photovoltaic (PV) technology due to its low life cycle creases. On an average, large photovoltaic installations have very
emissions of 14 g CO2-eq/kWh [1] is one of the most promising low capacity factor of 25% resulting in a power density of 35e37 W/
renewable energy technologies that can help mitigate global m2 (including inter-row spacing and other ancillary facilities) [4,5].
warming. Based on the PV technology roadmap by International If only solar panels area is considered, 1 kWp solar PV system oc-
Energy Agency (IEA), it is envisaged to contribute to more than 16% cupies 10 m2 of land approximately [6]. This implies that to be able
of the global electricity generation by 2050 [2]. The net installed to meet the expected solar PV capacity addition, huge areas of land
solar PV capacity till 2018 is only 485 GW [3]. In order to keep up are essential. Such large plots of land are not easily available in the
with the expected growth rate, one of the key requirements is the urban dwellings [7]. In addition, due the unavailability of land and
availability of shadow free land for large scale solar photovoltaic huge transmission losses in the utility-scale PV installations,
installations. distributed generation is becoming more popular [8].
The land area required for solar photovoltaic installations Solar tree can be a promising solution to distributed power
largely depends on the geographic location, insolation conditions generation especially in urban areas where there is a space
and topology of the land. For locations near the equator, tilt angle of constraint. In the urban areas, typical power density for electricity
solar PV installations are low in order to maximize power consumption ranges from 200 to 400 W/m2 [4]. Reduced ground
footprint and hence, high power density of electricity generation
using a solar tree at the point of consumption makes it an attractive
* Corresponding author. Council of Scientific and Industrial Research - Structural
solution. A few solar tree designs are reported in the recent past
Engineering Research Centre, Chennai, 600113, India. and some of the designs are commercially available [9e14]. Ross
E-mail address: [email protected] (S. Dey). Lovegrove has developed an innovative 5.5 m tall solar tree for LED

https://doi.org/10.1016/j.renene.2020.07.035
0960-1481/© 2020 Elsevier Ltd. All rights reserved.
S. Dey and B. Pesala Renewable Energy 162 (2020) 1747e1762

lighting consisting of 10 “stems of grass” of 40 mm diameter and 10

poles of 76 mm diameter [9,15]. Spotlight solar has designed
various solar trees for different applications for rating ranging from
1.7 kW to 5 kW [10]. These solar trees are designed using recycled
steel to withstand different wind speeds ranging from 120 to 175
MPH and have varying heights (130 to 22’) [10]. 3 kW solar power
tree which tracks the Sun by rotating the trunk is developed in
Ref. [11]. These solar trees consume lower ground footprint and are
aesthetically pleasing. However, it has been pointed out by Hyder
et al. [16]. that capture efficiency of these solar trees is low due to
large shading losses (as high as 30%). Patents [17e19] also propose
placement of solar panels in an inverted cone structure. However,
there is no mention on optimized positioning of panels to maxi-
mize energy extraction by minimizing the shading losses. Despite
the low ground footprint of solar trees compared to the conven-
tional solar PV installations, they are yet to become competent
scalable solutions for renewable power generation due to the
higher cost of the solar trees. This higher cost of the solar trees is
primarily due to the increased structural cost.
Earlier we have demonstrated the design of solar tree in order to
obtain tuned solar power generation curve [20]. The solar panels in Fig. 1. Representation of solar panels positions and orientations in solar tree.
the solar tree are arranged in different orientations and are posi-
tioned such that the shadow effect of solar panels on each other are
minimized. However, improper structural design of the branches panels. The first solar panel is considered as the reference solar
and trunk may result in unstable structure, higher material cost and panel with its centre positioned at origin ((x1, y1, z1) ¼ (0, 0, 0)). All
hence, higher levelised cost of energy of the solar tree system. other panel coordinates are denoted relative to the reference solar
Number of studies were reported on the structural design optimi- panel. East-West axis is represented by the x-axis, y-axis denotes
zation of wind turbine towers [21,22]. However, to the best of our the North-South axis while z-axis denotes the relative height of the
knowledge, comprehensive structural design and optimization of a solar panels w.r.t. the reference solar panel.
solar tree has not be carried out so far considering the energy The solar tree structure consists of a trunk (projecting from 1st
generation aspects. panel placed at origin to the ground) and n-1 number of branches
The main objective of the present paper is to achieve a trade-off (projecting from the trunk to the positions of each panel) as shown
between energy generation and structural cost to increase the in Fig. 2. In this study, the trunks and branches are considered to be
feasibility of deployment of solar tree. The optimization procedure hollow cylindrical sections. In order to maintain symmetry, while
employed initially involves optimizing the orientations of multiple maintaining the total height of the solar tree between 4 and 5 m,
solar panels in a solar tree so as to maximize the solar power the branches make an angle (d) of 45 with the trunk. Therefore,
generation. Subsequently, the solar panels are positioned in a solar given the position (xn, yn, zn coordinates) of the solar panels and
qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
 ffi
tree such that both shading loss and structural material require- d as 45 , the length (l) of each branch is given by 2 x2n þ y2n . The
ment is minimized. Analytical calculations for annual shading loss point of attachment of each branch to the trunk relative to the
estimation is done considering actual solar insolation data and
movement of Sun throughout the year. Positioning of solar panels
with wider space constraints results in lower shading loss. How-
ever, this leads to higher ground footprint area resulting in higher
lengths of branches and trunk and hence, higher structural mass
requirement. Leveraging genetic algorithms based multi-objective
optimization approach, a set of optimal positions of solar panels
in a solar tree are obtained. Assuming that all the branches of these
solar tree designs are of equal dimensions (outer diameter and
thickness), a design is selected where an optimal balance between
the shading loss and mass of structural material is achieved. The
optimal performance of the solar tree design thus obtained, is
validated through optical simulations using ZEMAX. In order to
further reduce the structural mass while ensuring structural
integrity at high wind speed, finite element modelling of the
structure is done to analyse the von Mises stress acting on the
structure. Based on von Mises criteria, the dimensions of the trunk
and branches in the solar tree structure are optimized individually.

2. Description of solar tree

The solar tree consists of multiple (n) solar panels each of length
(L)  breadth (B) oriented in different directions. The orientations
of the solar panels are denoted by tilt angle (bn) and surface azi-
muth angle (gn) as shown in Fig. 1. Solar panel positions are
denoted by the xn, yn, zn coordinates of the centre of the solar Fig. 2. Solar tree structure representation.

1748
S. Dey and B. Pesala Renewable Energy 162 (2020) 1747e1762

reference solar panel is determined using the relation

qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
- x2n þ y2n þ zn (given d ¼ 45 ). The height of base (point of
attachment to the trunk) of the lowest branch from the ground (h)
is set to 5 ft in order to provide easy accessibility while avoiding
major hindrance during its installation in parking lots, parks and
highways.
The solar tree structure is designed to be made up of mild steel
which is typically used for solar panel mounting structure. The
properties of the mild steel material used for the simulation pur-
pose is shown in Table 1. Using the values of outer diameter,
thickness and length of the branches and trunk, the volume of the
structure is calculated using basic volume calculation equations for
hollow cylindrical section. Density of structural material being
known, mass of the structure for a given set of solar panel positions
is given by the product of density and volume.
In the present study, the orientations of the ‘n’ solar panels of
the solar tree determine the solar power generation curve. The
positions of the solar panels determine the shading losses and
hence, affect net annual energy generation. The length of the
branches and trunk of the solar tree is also determined by the
positions of the solar panels. Therefore, solar panels positioning in
the solar tree also affects the structural material requirement. The
outer diameter and thickness of the branches further determine the
net structural material requirement.

3. Design methodology of solar tree

The energy generation from a solar tree primarily depends on

the orientations of the solar panels. The optimization of solar tree
involves alignment of multiple solar panels in different orientations
so as to be aesthetically pleasing without compromising on the
energy generation aspect. This requires that the solar panels are
optimally oriented taking into account the actual solar insolation
pattern of a particular geographic location. Typical solar insolation
curve for a location consisting of long term average of solar radia- Fig. 3. Solar tree design flow.
tion data over 20 years (1991e2010) is obtained from Meteonorm
database [23]. A time-series decomposition study using monthly
structural material is the main objective of the solar tree design.
data for 20 years obtained from National Renewable Energy Labo-
Fig. 3 shows various steps involved in the design of an electrically
ratory (NREL) [24] is conducted to validate the use of these typical
and structurally optimized solar tree.
solar insolation curves from Meteonorm as a true representation of
Depending on the solar insolation pattern of a particular
annual solar insolation curve of a location. Trend component
geographical location, the orientations of the solar panels are
calculated using 13-point moving average method indicated
optimized in order to maximize the energy generation. Thereafter,
gradual decrease in irradiance values over time. De-trended time
positioning of the solar panels is done by using genetic algorithms
series demonstrated seasonality implying that pattern followed by
based multi-objective optimization so as to minimize the shading
the annual solar insolation curve remains same every year. There-
losses and projected ground footprint area. The process is repeated
fore, long term average data is further used in the study as typical
for different volume constraints representing different ground
annual insolation curve of a location. The hourly global horizontal
footprint areas in each iteration. Each solar tree with reduced
irradiance (Ig) for the location is analysed. The hourly global hori-
ground footprint area results in reduced structural material
zontal irradiance (Ig), direct normal irradiance (Ib) and diffuse
requirement and hence, reduced cost. However, this occurs at the
horizontal irradiance (Id) values for the time duration during which
expense of increased shading loss and hence, reduced energy
Ig is greater than 100 W/m2 is considered for estimation of energy
generation. A solar tree design is selected where the ratio of
output and optimization of orientations and positions of solar
structural mass reduction to shading loss is maximized. Shading
panels.
loss is calculated w.r.t. solar panels arranged in the same plane
Maximization of solar power generation with the use of minimal
(where there is no shading). Structural mass reduction is calculated
w.r.t. a base case solar tree design where the solar panels are
Table 1 positioned with a large volume constraint such that shading losses
Material properties for solar tree structure. are nearly zero. The selected solar tree design is then validated for
Property Value its optical performance using ray optic simulations performed us-
ing ZEMAX software. The diameter and thickness of the branches
Density (g/cm3) 7.85
Yield Strength (S) (MPa) 240
and trunk in the structure of the selected solar tree is further
Ultimate Tensile Strength (MPa) 348 optimized using Finite Element Analysis (FEA) in COMSOL software
Young’s Modulus (E) (GPa) 200 so as to ensure structural stability with the use of minimal struc-
Poisson’s Ratio (n) 0.30 tural material and hence, minimized cost. The following sections

1749
S. Dey and B. Pesala Renewable Energy 162 (2020) 1747e1762

explain the solar tree design steps in detail. maximized.

3.1. Orientation of solar panels 3.2. Positioning of solar panels

Orientation of solar panels is given by tilt angle (b) and surface The orientation of the solar panels are then used to position
azimuth angle (g). The irradiance (I) on solar panel as a function of them such that the shading losses are minimized. This multi-
orientation is given by Eqn (1) [25e28]: objective optimization process leverages genetic algorithms with
two objective functions as shown in Fig. 4. Based on the principles
cosqi 1 þ cosb of transmission of light, the annual energy loss due to shading is
Iðb; gÞ ¼ Ib þ Id (1) defined as a function of solar panel positions (Pn (xn, yn, zn)), length
cosqz 2
(L), breadth (B), orientation (bn, gn) and DNI of the location. It is
where, qi is the angle of incidence and is a function of solar panel known that light travels in a straight line and objects placed in the
orientation and position of Sun in the sky and qz is the solar zenith path of light, cast shadow. Using this basic principle, area of shadow
angle. Summation of irradiance at each hourly time instant gives cast by one solar panel “Panel 1” on another solar panel “Panel 2” is
the value of incident energy per day (E). Annual average energy calculated using the algorithm represented in Fig. 4a.
(Eavg) is defined as the mean of incident energy for 365 days. Given the centre point coordinates of the solar panel ABDC
Conventional solar panels are installed at a tilt angle equal to the denoted as O [x0, y0, z0], length (L) and height (H) of solar panel, the
latitude tilt angle facing south/north for locations in the northern/ end point coordinates A, B, C and D of the solar panel are calculated
southern hemisphere [25,26,29]. It has been shown in Ref. [20] that using equations (2)e(6) as shown in Fig. 4b.
the orientations of solar panels can be optimized based on the

   

H H
O’ ¼ x0  cosbsinðg þ 180 Þ ; y0  cosbcosðg þ 180 Þ ; ðz0  ðH = 2ÞsinbÞ (2)
2 2

period of the year during which maximum energy extraction is

required. In the present study, orientation of solar panels in a solar    
tree configuration consisting of ‘n’ number of solar panels is opti- A¼ ðO’x þðL=2Þcosðg þ180 ÞÞ; O’y ðL=2Þsinðg þ180Þ ; ðO’z Þ
mized so as to provide levelised solar power generation curve (3)

   
B ¼ ðAx þ H cos bsinðg þ 180 ÞÞ; Ay þ Hcosbcosðg þ 180 Þ ; ðAz þ HsinbÞ (4)

throughout the year. For this purpose, the annual insolation curve
of the location under study is analysed and low insolation months
are targeted. To understand the variation of average energy in the    
targeted months (Etarget) with tilt angle (b) and surface azimuth C ¼ ðAx  Lcosðg þ 180 ÞÞ; Ay þ Lsinðg þ 180 Þ ; ðAz Þ (5)
angle (g) of solar panel, Etarget is calculated by varying b between

   
D ¼ ðCx þ H cos bsinðg þ 180 ÞÞ; Cy þ Hcosbcos ðg þ 180 Þ ; ðCz þ HsinbÞ (6)

0 to 90 and g between 0 to 360 in steps of 0.01. It is observed !

For a particular instant of time, the directional vector ( s ) for the
that maximization of energy extraction in a certain part of the year Sun’s ray are computed as mentioned in Ref. [20] and shown in
occurs at the cost of decreased energy extraction in the other part of equation (7) below:
the year. In order to obtain a levelized solar insolation curve, the
orientation of the solar panels in the solar tree are selected such
!
that incident energy in the targeted months is improved while s ¼ ½cosgs cosa; sings cosa; sina (7)
simultaneously improving the annual average incident energy.
Therefore, each solar panel of the solar tree is oriented such that where, gs and a denote azimuth angle and altitude angle of the Sun
average energy in a particular targeted month or for a particular respectively. Thereafter, a line segment is created passing through
!
period of the year (summer months, winter months etc.) is maxi- each point of “Panel 1” in the direction of Sun vector ( s ). One end
mized. In order to balance the power generation curve, the orien- point of the line segment is considered to be the “Panel 1” coor-
tations of ‘n’ solar panels in solar tree are chosen such that the sum dinate (For example: Point A). The relation used for finding a sec-
of annual average energy (Eavg) provided by the ‘n’ solar panels is ond end point for the line segment is shown in equation (8) for
1750
S. Dey and B. Pesala Renewable Energy 162 (2020) 1747e1762

Fig. 4. (a) Flowchart for the calculation of shaded area, Pictorial representation of (b) Solar panel end point coordinates, (c) Sun rays, (d) Projection/shadow of “Panel 1” on “Panel 2”
plane, (e) Shaded area of “Panel 2”

point A of “Panel 1”. The value of k in equation (8) is any constant lying above it. In other words, for “Panel 1” to cast shade on “Panel
integer value. Therefore, the line segment representing Sun rays 2”, the relative height of “Panel 2” must be less than that of “Panel
passing through point A is given by SAA. Similar relation is used for 1”. In order to avoid calculation of false shading effect, it is checked
all the 4 points of the panel 1 to obtain line segment representing that the “Panel 1” (ABDC) lies between the shadow (IAIBIDIC) and the
Sun rays as shown in Fig. 4c. Sun.
Position of Sun may be represented by SASBSDSC which is
!
SA ¼ A þ k  s (8) calculated using the relation shown in equation (8). For this check,
the centre points of “Panel 1”, shadow and Sun are calculated
The point of intersection of solar radiation (whose direction is
! represented as p, i and s respectively. Dot product is used as a check
given by the directional vector s ) passing through point A, B, C and if p lies between i and s using the relation shown in equation (10).
D, with the plane of the “Panel 2” (PQSR), gives the end points of the
shadow (IAIBIDIC) cast by the “Panel 1” (ABDC) (marked in light
green in Fig. 4d). To find the shadow of “Panel 1” (ABDC) in the ! !
! check ¼ pi , sp (10)
plane of “Panel 2” (PQSR), the normal vector ( n ) to “Panel 2” plane
is calculated using equation (9). Value of check >0 implies p lies between i and s. If the check is
satisfied (check >0), the intersection of the shadow and the “Panel
! ! ! 2” gives the shaded area (marked in black) as shown in Fig. 4e.
n ¼ PQ  PR (9)
Similar routine is followed for shadow calculation of each panel on
The algorithm used for calculation of intersection point of these all the other panels in the solar tree for each time instant. It may be
line segments with the plane of “Panel 2” is shown in Fig. 5. The noted that the shading due to branches of the solar tree has not
!
algorithm takes inputs as the normal vector to Panel 2 ( n ), any been considered in this study. This is primarily due to the fact that
point in the plane of panel 2 (For example Point “P”), end points of branches (having low cross-sectional area) lie below the solar
the line segment as mentioned above (For example: P0 ¼ A, P1 ¼ panels (having large area). Hence, shading effect due to the solar
SA). The process is repeated for the 4 points of “Panel 1” to obtain panels will be relatively much higher than that of the branches.
the shadow (IAIBIDIC) on “Panel 2” plane. The point O in this algo- It is known that shading does not affect diffuse radiation inci-
rithm represents the origin [0 0 0]. dent on solar panel. Hence, the energy loss due to shading is
It is known that an object cannot cast shadow on another object calculated by multiplying the beam radiation on each solar panel to
1751
S. Dey and B. Pesala Renewable Energy 162 (2020) 1747e1762

Fig. 5. Algorithm for finding intersection point of line and plane.

Fig. 6. Flowchart for optimal positioning of solar panels.

the shaded area (Ah) and efficiency of solar panel (he). The sum of
energy loss at each time instant gives the net energy loss (El) [20] as
and minimization of net projected ground footprint area (Ap) are
shown in equation (11).
the two fitness functions for the multi-objective optimization
XX problem carried out using genetic algorithms as shown in Fig. 6.
El ¼ Ib ðbn ; gn Þ  Ah  he (11) The initial population size is set to 200. The next population is
h n selected using a tournament selection function of size 2 where 2
Net estimated annual energy generated (Eg) from the solar tree individuals from the present generation are selected randomly and
consisting of ‘n’ solar panels is calculated by subtracting the the best out of the set is chosen as the parent. With crossover
shading loss (El) from the estimated annual energy without fraction of 0.8, uniform mutation function with 0.01 mutation rate
considering shading loss. This essentially means that Eg consists of and distance crowding function as a measure of concentration of
energy generation due to diffuse component of solar radiation the population, the optimization continues till the difference be-
received by the entire solar panel and energy generation due to tween the best individual in two consecutive generation is less than
beam radiation from the non-shaded parts of solar panel. It is worth 0.001. Lower (xl, yl, zl) and upper bounds (xu, yu, zu) of the x, y and z
mentioning here that the present study has not considered the coordinates are given in order to constrain the centres of the solar
interconnection losses in the solar tree. This represents a scenario panels within a predefined volume. The best solution in terms of
where each solar panel of the solar tree is individually connected to the annual energy generated (Eg) per unit ground footprint area is
a micro-inverter/Maximum Power Point Tracker (MPPT). Hence, selected as the solution for the given set of constraints.
the net energy for the solar tree (Eg) is given by the summation of
energy generated by individual solar panels of the solar tree. 3.2.1. Selection of solar tree design
Considering the 1st solar panel positioned at the origin (0, 0, 0) If the positioning of solar panels is done with larger volume
as the reference panel, the number of variables, i.e. the positions of constraint having wider allowable space in the x-y plane, the
solar panels to be optimized is 3  (n-1). The net projected ground shading effect reduces. However, the lengths of branches projecting
footprint area (Ap) is also defined as a function of solar panel po- from the trunk of the solar tree increases, hence the amount of
sitions and orientation. Similar to calculation of shaded area, for the structural material required increases. Similarly, stricter constraint
calculation of projected area, the solar panels are projected onto the on volume implies higher shading loss and lower structural ma-
ground (i.e. xy plane). The direction vector used for projecting the terial requirement. Hence, there needs to be a balance between the
solar panels on the xy plane is given by the negative z-axis [0, two. The process for the selection of optimized solar tree design is
0, 1]. The area of the largest rectangle enclosing all the solar shown in Fig. 7.
panels is considered as the net projected area (Ap). Initially, the positions of the solar panels for a base case solar
Therefore, maximization of net annual estimated energy (Eg) tree design is selected by applying large volume constraint such
1752
S. Dey and B. Pesala Renewable Energy 162 (2020) 1747e1762

mean higher savings in structural cost without compromising on

the energy loss due to shading. The solar tree positioning that re-
sults in maximized R value is referred to as the selected solar tree
design.

3.3. Validation of solar tree design

The shading losses in the solar tree design are then validated
using detailed optical simulations using ZEMAX software. ZEMAX
OpticStudio is an effective tool for optical analysis of systems as it
provides a 3D view of the light ray paths which enables clear un-
derstanding of the system performance. The 3D CAD model of the
designed solar tree is constructed using Autodesk Inventor and
movement of the Sun in sky is simulated using non-sequential
mode of ZEMAX OpticStudio. Large rectangular source, whose di-
rection and power are set as per the Sun’s position and DNI at a
particular time instant, are used to simulate the solar insolation
values. DHI values are not considered in this optical simulation as it
does not affect the shading losses significantly. Absorbing detectors
are placed on the solar panel surface which are divided into pixels
of 1 cm2. The light intensity incident on the solar panel surface at
different time instants of the day throughout the year are observed.
These values are matched with that of the results obtained using
the analytical calculations.

3.4. Structural optimization

The optimization of solar tree structure is done using Finite

Element Analysis (FEA) using COMSOL software. The selected solar
tree design obtained above consists of all the branches of equal
dimension. Here, outer diameter and thickness of each branch and
the trunk are optimized individually depending on the static loads
acting on it. Static loads include the self-weight of the structure,
weight due to solar panel assembly on the branches and wind load.
The von Mises stress on the structure due to these loads is calcu-
lated. The dimensions (outer diameter and thickness) of each
branch is then tuned in order to design a solar tree structure to
withstand the location specific wind speed with a factor of safety 2.
For a ductile material, von Mises stress is the most commonly
used parameter to determine if a given material will yield or frac-
ture. Mild steel used in this study being a ductile material, its failure
Fig. 7. Flowchart for optimized solar tree design selection.
due to plasticity is studied using von Mises criteria [30] which
states that if the von Mises stress of a material under load is equal
that the shading loss is nearly zero. A preliminary structural anal- or greater than the yield limit of the same material under simple
ysis of the base case design is done so as to set the diameter and tension, then the material will yield. This gives the constraint on the
thickness of the trunk and various branches of the solar tree such maximum stress that the solar tree branches can be subjected to as
that the von Mises stress in all the branches and trunk of the base described in equation (12).
case design is less than the maximum permissible stress as
tv  S (12)
mentioned in Section 3.4. As the mass of structural material usually
determines its price, the optimization aims to reduce the structural
where S is the yield strength of material, tv is the calculated von
mass. All the branches in this solar tree design has equal di-
Mises stress. The objective of the structural optimization is to make
mensions (outer diameter and thickness).
the solar tree structurally stable at a design wind speed defined for
Thereafter, the solar panels are positioned with gradually
a particular location as per IS: 875 (Part 3) [31] with the use of
stricter volume constraints in each iteration. The upper and lower
minimal structural material. This will in turn reduce the cost of the
bounds are reduced by a factor (D) in each iteration. For each iter-
solar tree structure which forms a significant part in the cost of the
ation, the reduction in energy due to shading (Eshad) is evaluated.
solar tree system. The tree is designed to have a factor of safety 2 as
Also, keeping the branch dimensions equal to that of the base case
shown is equation (13).
solar tree designs, the mass of the structural material required for
the solar panel positions obtained in each iteration is calculated.
2tv  S (13)
The reduction in structural mass (Smass) w.r.t. the base case solar
tree design is evaluated. Since it is desirable to maximize Smass and Therefore, for the mild steel considered in simulation, the
minimize Eshad, a ratio of Smass to Eshad is considered as the opti- maximum permissible stress at any of the branches or trunk in the
mization parameter (R). Maximization of this parameter R would solar tree structure has to be less than 120 MPa.
1753
S. Dey and B. Pesala Renewable Energy 162 (2020) 1747e1762

Fig. 8. Solar tree finite element model for structural design and optimization.

Fig. 10. Point of stress analysis.

Fig. 9. Loading conditions on the solar tree structure.

3.4.1. Stress analysis

For the purpose of analysis, a finite element model of the solar
Fig. 11. Improved solar power generation curve by tree configuration.
tree structure is generated using COMSOL. The 3D CAD model
designed using Autodesk Inventor is imported using COMSOL CAD
import module and the finite element model is created using
tetrahedral mesh with finer setting as shown in Fig. 8. The
maximum element size is set to 0.3 m while the minimum element Table 2
size is set to 0.01 m with a maximum growth rate of 1.45 and Improvement in the energy obtained by tree configuration.
curvature factor 0.5. The resolution of narrow region is set to 0.6. Parameter Improvement (%) w.r.t.

Horizontal Latitude Tilt

3.4.2. Loading conditions Annual Average 0.46 0.69
The base of the solar tree is then fixed as shown in Fig. 9 Average for months (OcteFeb) 10.06 7.03
implying that displacement of the base is zero. Also, the joint of Average for rest of year 2.28 1.52

each branch with the trunk is considered as welded joints. There

are 3 different loading conditions applied on the solar tree struc-
ture. The first being the body weight of the solar tree structure force due to body weight acts in the downwards direction along the
which is given by equation (14). z-axis.
The second force acting on the solar tree branches is the load
W¼ r  V  g (14) due to the weight of the solar panel and mounting assembly (Fg) on
the top of branches. This force is applied in the downward direction
where r is the mass density of the material, V is the volume of the along z-axis as shown in Fig. 9.
material and g is the acceleration due to gravity (9.81 m/s2). This
1754
S. Dey and B. Pesala Renewable Energy 162 (2020) 1747e1762

Table 3
Orientations of solar panels for 3 kW solar tree.

Solar Panel Tilt Angle (b) Surface Azimuth Angle (g) Solar Panel Tilt Angle (b) Surface Azimuth Angle (g)

1 24 185 6 2 168

2 9 178 7 20 178
3 13 180 8 17 190
4 9 179 9 13 169
5 3 19 10 7 186

Fig. 12. Stress on the branch and trunk cross-sections of base case solar tree design with branch cross-sectional dimensions of 400 diameter and 6 mm thickness.

1755
S. Dey and B. Pesala Renewable Energy 162 (2020) 1747e1762

Table 4
Energy loss due to shading and structural mass reduction using different constraint conditions.

Iteration Net projected ground Net annual energy % Energy lost due to Structural mass (kg) % Structural mass R
footprint area (Ap) (m2) generated (Eg) (MWh) shading (Eshad) reduction (Smass)

Base Case 41.18 5.81 0.14 2187 0 0.00

1 36.56 5.79 0.49 2117 3.20 6.53
2 33.57 5.81 0.17 2041 6.66 39.20
3 33.33 5.78 0.54 2061 5.76 10.66
4 29.18 5.69 2.21 1906 12.90 5.83
5 25.96 5.50 5.37 1658 24.20 4.51
6 25.68 5.66 2.7 1801 17.70 6.55
7 23.71 5.50 5.37 1657 24.20 4.51
8 21.82 5.54 4.72 1563 28.50 6.05
9 16.56 5.22 10.24 1374 37.20 3.63

Fig. 13. Shading loss and structural mass reduction for different projected ground
footprint area.

Fig. 15. CAD model of selected solar tree design.

Fig. 14. Variation of parameter R for different ground footprint area.

The third important load acting on the solar tree structure is the
wind load which is given by IS: 875 (Part 3) as shown in equation
(15).

1 Fig. 16. Irradiance profile due to DNI of selected solar tree design at 12 p.m. on 15th
Fw ¼ ACp v2 (15) March using ZEMAX optical simulation.
2

where r is the density of air and is considered to be 1.2 kg/m3, A is

speed for the solar tree designed for a particular location is defined
the projected solar panel area in the direction of wind, v is the
as per IS: 875 (Part 3) as mentioned in equation (16).
design wind speed and Cp is the pressure coefficient. Design wind
1756
S. Dey and B. Pesala Renewable Energy 162 (2020) 1747e1762

4. Results and discussion

In the present study, to demonstrate the multi-objective design

framework, a 3 kWp solar tree is designed for the location Chennai,
India geographically positioned at 13.08 N, 80.27 E. An analysis of
the average hourly global horizontal irradiance (Ig) for Chennai for
each month shows that it receives a maximum of 8.8 W/m2 (June)
and 63.4 W/m2 (July) at 6 a.m. and 6 p.m. respectively. Hence, 11 h
data from 7 a.m. to 5 p.m. for 365 days is used in the study during
which value of Ig is greater than 100 W/m2.
The solar tree consists of 10 solar panels each of rating 300 Wp.
The dimension (L  B) of the solar panel is considered to be
2 m  1 m which is usually the dimension of commercially available
300 Wp solar panels. A solar panel efficiency of 15% is assumed under
real world operating conditions. Though a particular panel (300 Wp)
and solar tree capacity (3 kWp) is chosen for the present study, the
proposed multi-objective optimization framework can be easily
extended for any other geographical location and solar tree capacity.

Fig. 17. Irradiance profile due to DNI of selected solar tree design at 12 p.m. on 15th
March using analytical calculations. 4.1. Orientation of solar panels

Chennai being located at a low latitude angle (13.08 ), has

relatively less deviation in the solar power incident curve. However,
it can be seen in Fig. 11 that the average incident energy on hori-
v ¼ vb k1 k2 k3 k4 (16)
zontal surface is relatively low in the winter months (October to
where vb is the basic wind speed for the location provided in IS:875 February) than in the summer months (March to September). The
(Part 3). The factors k1, k2, k3 and k4 correspond to risk coefficient, methodology presented in Section 3.1 is used for calculation of
terrain roughness and height factor, topography factor and impor- annual optimal orientation for maximum annual energy extraction
tance factor for cyclonic region respectively. The values of these for this geographic location. The result shows that the optimal tilt
constant factors are selected as per guidelines provided in IS: 875 angle for Chennai is 9.42 and surface azimuth angle is 178.46 .
(Part 3). In order to consider the worst-case scenario where the Also, in case of Chennai, the latitude tilt orientation (13.08 , 180 )
entire wind force acts on the solar tree, pressure coefficient Cp is gives equal annual average energy close to that of the optimal
considered to be 1. For the static wind load on solar tree which is orientation. The optimization of orientations of the 10 solar panels
placed facing the South, it is considered that the wind is flowing in in the 3 kWp solar tree is done to maximize the annual average
the direction from South to North. This is considered as the worst- energy extraction. The optimization of orientations (tilt angle and
case scenario as the projected area will be highest in this direction surface azimuth angle) has resulted in solar power generation
and hence, the force due to wind will be highest in this direction. A curve which closely matches with that of the latitude tilt configu-
solar tree designed with a factor of safety 2 for wind in this direc- ration with increased generation in the winter months. However,
tion will potentially be able to withstand wind force from other higher improvements can be obtained for locations with higher
directions also. Since, the connection of the branch to the trunk is latitude and large deviation in solar insolation curve as presented in
fixed, ideally maximum stress occurs at the base of the branch and Ref. [20].
the minimum stress occurs at the free end of the branch where load Table 2 shows the improvement obtained using the tree
is applied. However, during the mesh creation, infinitely small el- configuration. Though the annual average energy improvement is
ements are formed at the branch bases. Solutions of FEA for these relatively less, it can be seen that about 10% improvement in energy
elements tend towards infinity and are referred to as singularities generation is obtained in the winter months. Also, the standard
in FEA [32]. Therefore, in order to obtain true value of von Mises deviation of the solar power generation curve is reduced by 15%
stress at the branch bases, a cross-section exactly 2 cm away from w.r.t. horizontal. This implies that the tree configuration provides a
the trunk is considered as shown in Fig. 10. Since, 2 cm is very small more levelized solar power generation curve throughout the year.
as compared to the branch lengths which are of the order of The optimized orientations of the solar panels are shown in Table 3.
1.5e5 m, stress values 2 cm away from trunk can be considered as
the value of maximum stress occurring in each branch. In order to 4.2. Positioning of solar panels for optimal solar tree design
individually fine-tune the dimensions, outer diameter and thick-
ness of each branch and trunk is iteratively increased/decreased so The objective of this study is to reduce the ground footprint area
as to ensure that maximum stress at any point in the branch/trunk of the solar tree without compromising on the energy loss. For the
is just below 120 MPa. optimization of positions of panels for the base case solar tree

Table 5
Comparison of estimated power generation due to DNI using analytical calculation and ZEMAX simulation.

Time DNI (W/m2) Estimated Power Generation due to DNI

Analytical Calculation (W) ZEMAX optical simulation (W)

March 15th 9 a.m. 554.74 1060.59 1060.58

March 15th 12 p.m. 557.81 1646.42 1646.40
March 15th 3 p.m. 570.10 1280.74 1279.64

1757
S. Dey and B. Pesala Renewable Energy 162 (2020) 1747e1762

Fig. 18. Stress on the branch and trunk cross-sections of selected solar tree design before structural optimization.

design, the constraints in the x-y plane are set to ±2.5 m. A pre- wind speed for Chennai is 40.95 m/s (~150 kmph).
liminary optimization of the base case design is done to set the Mass of commercially available 300 Wp solar panel, used in this
outer diameter and thickness. Each branch is considered to have simulation, is usually 22e25 kg. Mounting assembly for these solar
equal dimensions. The wind load acting on the solar tree is defined panels consists of an aluminium frame along with necessary
as per IS: 875 (Part 3) as mentioned in equation (8). According to arrangement for adjustment of orientation of solar panel. The net
the standard, vb is 50 m/s for Chennai region. The value of risk mass of the solar panel along with mounting assembly will be
coefficient (k1) is 0.90 considering the design life of the structure as about 50 kg. Hence, Fg is considered to be 500 N which has been
25 years. Considering terrain category 3 having closely spaced applied in the downward direction along z-axis.
building structure up to a height of 10 m, the terrain roughness The optimization procedure is started with outer diameter as 500
factor (k2) is taken as 0.91. The topography factor (k3) and impor- and thickness 8 mm. It was noticed that stress in all branches were
tance factor (k4) are taken as 1. Considering these factors, the design much lower than that of the maximum permissible stress

1758
S. Dey and B. Pesala Renewable Energy 162 (2020) 1747e1762

Fig. 19. Stress on the branch and trunk cross-sections of the optimized solar tree.

(120 MPa). Hence, outer diameter and thickness of the branches Hence, it is assumed that solar tree designs in further iterations
were reduced in steps of 0.500 and 1 mm respectively until the stress with branch and trunk dimensions as per the base case solar tree
in all the branches is close to 120 MPa. Fig. 12 shows the stress in design will be structurally stable. Each branch of the solar tree is
each branch and trunk of the base case solar tree design. For branch considered to have a diameter of 400 and thickness 6 mm. The trunk
of outer diameter 400 and thickness 6 mm, the maximum stress in of the solar tree has a diameter of 7” and thickness 7 mm. Multiple
branch 9 is 127.05 MPa which is slightly higher than 120 MPa. iterations of positioning of solar panels is done by reducing the
Similarly, trunk of outer diameter of 7” and thickness 7 mm expe- upper and lower bounds by 0.125 m (D ¼ 0.125 m) in each iteration
riences a maximum stress of 116.03 MPa. Since, in further itera- until the energy loss due to shading (Eshad) is less than 10%. The
tions, solar panels are positioned with higher constraint in the x-y result of this simulations and the corresponding structural mass
plane, the branch lengths in these iterations will be less than that of reduction (Smass) calculated w.r.t. the base case solar tree design is
the base case solar tree design. shown in Table 4.
1759
S. Dey and B. Pesala Renewable Energy 162 (2020) 1747e1762

increase and percentage reduction in structural mass also in-

creases. From Fig. 13, it can be seen that the rate of reduction in
structural mass is higher than that of the rate of increase in shading
loss. It can be seen from Fig. 14 that the ratio Smass/Eshad peaks for a
certain arrangement of solar panels (iteration 2). For this selected
solar tree design, the normalized projected ground footprint area is
1.67, while the shading loss is only 0.17%.

4.3. Validation of energy loss due to shading

The energy losses due to shading are confirmed using the optical
simulations. Fig. 15 shows the CAD model of the selected solar tree
design. The optical simulation is done for 3 time instants of the day
i.e. 9 a.m., 12 p.m. and 3 p.m. for 15th day for every month. With the
solar panels (blue in color) as absorbing detectors, a rectangular
light source is positioned according to the Sun’s position at the time
instant being simulated. The simulation is done with 10 million rays
Fig. 20. Stress levels on the optimized solar tree under full wind load conditions. such that intensity of the light source is equal to the DNI. Diffuse
radiation is not considered in this simulation. Fig. 16 and Fig. 17
show the irradiance on the solar panels obtained using optical
simulation and analytical calculations respectively for the selected
solar tree design at 12 p.m. on March 15th during which DNI is
557.81 W/m2.
The estimated power generation due to the direct radiation
obtained using analytical calculation and ZEMAX simulation are
shown in Table 5. It can be seen from the table that the optical
simulation results closely match to that of the analytical calcula-
tions for all time instances.

4.4. Optimization of structural design

Stress analysis of the solar tree design is done using solid me-
chanics module of COMSOL Multiphysics 5.2 in stationary mode.
The loading conditions and constraints are applied as mentioned in
Section 3.4. Initially, the solar tree design with all branches of equal
dimension (400 , 6 mm) and trunk of dimension (7” 7 mm) is ana-
lysed. Fig. 18 shows the stress at the circular cross-section at the
Fig. 21. Displacement of branches in the optimized solar tree under full wind load base of each branch and trunk and the maximum stress in each
conditions of 150 kmph
branch. The maximum stress on branch 8 and trunk are 104.37 MPa
and 106.06 MPa respectively while the branches 3, 6 and 9 have
very low maximum stress values in the range of 30e60 MPa. It is
The structural mass is calculated by considering that the branch
clear that each branch dimensions (outer diameter and thickness)
dimensions (outer diameter and thickness) are same for all cases.
can be tuned individually such that maximum stress in each branch
The energy loss and structural mass reduction percentages are
is close to 120 MPa. The stress acting on each branch of the solar
calculated w.r.t. base case solar tree design where the normalized
tree after optimization is shown in Fig. 19. As can be seen from the
projected area is 2.06 (w.r.t. actual solar panel area) and the annual
figure, the maximum stress acting on all branches and trunk is
shading loss (Eshad) is only 0.14%.
within the limits defined. Depending on the tilt angle of the solar
It is seen that as the projected area reduces, shading losses
panel on each branch and the length of the branches, the optimized

Table 6
Positions and dimensions of branches in optimized solar tree for Chennai.

Solar panel X position Y position Z position Branch Length (m) Diameter (inch) Thickness (mm) Mass (kg)

1 0.00 0.00 0.00 4.25 7 7 502

2 2.10 0.02 0.20 2.97 3 6 123
3 0.87 0.99 0.30 1.87 2.5 4 44
4 1.69 2.09 0.56 3.79 4 4 146
5 1.23 2.20 0.54 3.57 3.5 5 148
6 1.86 0.90 0.31 2.93 3.5 4 98
7 2.06 0.30 0.04 2.94 4 5 140
8 2.20 1.62 0.22 3.86 4 6 218
9 0.20 1.21 0.15 1.73 3 4 49
10 1.90 1.63 0.15 3.54 4 5 169
Mass of optimized solar tree 1638
Mass of non-optimized solar tree 2041
% Reduction in structural mass 20

1760
S. Dey and B. Pesala Renewable Energy 162 (2020) 1747e1762

branch dimensions have reduced by varying degrees. For solar Declaration of competing interest
panels installed at higher tilt angle, wind load is higher and hence,
higher branch dimensions are required as in case of branch 7 and 8 The authors declare that they have no known competing
which have solar panels installed at 20 and 17 respectively. Also, financial interests or personal relationships that could have
higher diameter and thickness of branches are essential in case of appeared to influence the work reported in this paper.
longer branches as in case of branch 4 and 10.
The von Mises stress acting on the entire tree structure is shown Acknowledgement
in Fig. 20. The stress in the entire tree structure is within the limits.
Also, Fig. 21 shows the displacement of the various branches and The authors acknowledge the funding support of CSIR through
trunk of the solar tree under wind load of 150 kmph considering a MLP-0113 grant. Author Sumon Dey thanks CSIR-SRF fellowship for
factor of safety of 2. It can be seen that the maximum displacement the financial support. Author Sumon Dey would like to thank Dr.
of about 9 cm under full load conditions has occurred in the Madan Kumar Lakshmanan for his valuable inputs while carrying
branches of length greater than 3.5 m. Since, the optimized selected out the work.
solar tree design satisfies von Mises criteria with a factor of safety 2,
the above displacement is acceptable. The positions, dimensions References
and lengths of each branch of the optimized solar tree design for
Chennai is given in Table 6. It is observed that the structural opti- [1] R.R. Hernandez, M.K. Hoffacker, M.L. Murphy-Mariscal, G.C. Wu, M.F. Allen,
mization has resulted in 20% reduction in structural mass w.r.t. the Solar energy development impacts on land cover change and protected areas,
Proc. Natl. Acad. Sci. Unit. States Am. 112 (2015) 13579e13584, https://
non-optimized selected solar tree design structure. doi.org/10.1073/pnas.1517656112.
[2] S. Weckend, A. Wade, G. Heath, End-of-life Management: Solar Photovoltaic
Panels, 2016.
[3] International Renewable Energy Agency (IRENA), Renewable Energy Statistics,
5. Conclusion 2019.
[4] S. Lo Piano, K. Mayumi, Toward an integrated assessment of the performance
of photovoltaic power stations for electricity generation, Appl. Energy 186
The concept of solar tree is rapidly gaining attention for (2017) 167e174, https://doi.org/10.1016/j.apenergy.2016.05.102.
achieving higher energy generation per unit ground footprint area. [5] S. Ravi, J. Macknick, D. Lobell, C. Field, K. Ganesan, R. Jain, et al., Colocation
opportunities for large solar infrastructures and agriculture in drylands, Appl.
However, high shading losses and increased structural material
Energy 165 (2016) 383e392, https://doi.org/10.1016/j.apenergy.2015.12.078.
requirements are major bottlenecks resulting in lower public [6] F. Hyder, P. Baredar, K. Sudhakar, R. Mamat, Performance and land footprint
acceptance. High cost of the existing solar tree designs is mainly analysis of a solar photovoltaic tree, J. Clean. Prod. 187 (2018) 432e448,
https://doi.org/10.1016/j.jclepro.2018.03.249.
due to high structural material requirement. The present study
[7] R. Rachchh, M. Kumar, B. Tripathi, Solar photovoltaic system design optimi-
focuses on maximizing power generation from the solar tree while zation by shading analysis to maximize energy generation from limited urban
minimizing the structural material requirement. A 3 kWp solar tree area, Energy Convers. Manag. 115 (2016) 244e252, https://doi.org/10.1016/
is designed for Chennai located at 13.08 N, 80.27 E. The solar j.enconman.2016.02.059.
[8] Solar PV market outlook 2018: Cheaper and bigger d continuing to break
panels in the solar tree are optimally oriented taking into account records, n.d. https://www.renewableenergyworld.com/articles/print/volume-
actual solar insolation data of the location so that annual energy 21/issue-1/features/solar/solar-in-2018-better-technology-record-breaking-
generation is maximized. Thereafter, genetic algorithms based installations.html. (Accessed 17 June 2018).
[9] Solar tree e ross lovegrove, n.d. http://www.rosslovegrove.com/index.php/
multi-objective optimization technique is used to position the solar custom_type/solar-tree/. (Accessed 23 July 2018).
panels of solar tree such that shading losses and projected ground [10] Solar Power Trees | Spotlight Solar Products d Solar Trees | Spotlight Solar
footprint area are minimized. Lower projected ground footprint n.d. https://spotlightsolar.com/products/(accessed July 23, 2018).
[11] Solar Power Tree | Central Mechanical Engineering Research Institute n.d.
area implies lower branch lengths in solar tree and hence, lower http://www.cmeri.res.in/technology/solar-power-tree (accessed July 23,
structural material requirement. However, shading losses also in- 2018).
creases with decrease in projected ground footprint area. An [12] Solar Energy | etree n.d. http://sol-logic.com/etree/(accessed July 23, 2018).
[13] “Solar Tree” to Replace Real Tree in Bristol City Centre - BBC News (n.d).
optimal balance between the two is obtained such that annual
[14] JIP- Innovative solar tree manufacturer and supplier, n.d. http://www.
shading losses are only 0.17% while the normalized ground foot solartree.co.in/. (Accessed 23 July 2018).
print area is 1.67. The dimensions (outer diameter and thickness) of [15] Artemide - solar tree, n.d. https://www.artemide.com/en/subfamily/665440/
solar-tree. (Accessed 8 December 2018).
each branch of the solar tree is further optimized individually for
[16] F. Hyder, K. Sudhakar, R. Mamat, Solar PV tree design: a review, Renew.
static loading conditions by applying von Mises criteria using FEA. Sustain. Energy Rev. 82 (2018) 1079e1096, https://doi.org/10.1016/
The structural optimization of the solar tree designed for a wind j.rser.2017.09.025.
load of 150 kmph with a factor of safety 2 has resulted in 20% [17] P.T. Kumar, Solar Tree, 2015. WO2015198348A1.
[18] F.L. Christy, S.C. Keiser, D.M. Archer, R. Arter, Solar Tree with Optional Wind
reduction in structural material requirement. The shading losses Turbine Generator, 2015. US 9,151,273 B2.
and power generation from solar tree are validated using ray optic [19] F.L. Christy, S.C. Keiser, D.M. Archer, Solar Wind Tree, 2013. US 8.487,469 B2.
simulations. The design framework established in the paper ad- [20] S. Dey, M.K. Lakshmanan, B. Pesala, Optimal solar tree design for increased
flexibility in seasonal energy extraction, Renew. Energy 125 (2018)
dresses the issues with solar tree and can be employed to design of 1038e1048, https://doi.org/10.1016/j.renene.2018.02.017.
solar tree of 5 kW and 10 kW for various geographical locations. [21] I. Lavassas, G. Nikolaidis, P. Zervas, E. Efthimiou, I.N. Doudoumis,
Further, light weight materials such as Fiber-Reinforced Plastic C.C. Baniotopoulos, Analysis and design of the prototype of a steel 1-MW wind
turbine tower, Eng. Struct. 25 (2003) 1097e1106, https://doi.org/10.1016/
(FRP) can be used to minimize net weight of the solar tree and S0141-0296(03)00059-2.
hence, making transportation and installation
̊ easier.̊ [22] D. Stratton, D. Martino, F.M. Pasquali, K. Lewis, J.F. Hall, A design framework
for optimizing the mechanical performance, cost, and environmental impact
of a wind turbine tower, J. Sol. Energy Eng. 138 (2016), 041008, https://
doi.org/10.1115/1.4033500.
[23] Meteonorm: typical years, n.d. http://www.meteonorm.com/en/product/
CRediT authorship contribution statement
productpage/mn-produkt-dataset. (Accessed 6 February 2017).
[24] NSRDB: 1991- 2010 update, n.d. http://rredc.nrel.gov/solar/old_data/nsrdb/
Sumon Dey: Methodology, Software, Formal analysis, Investi- 1991-2010/#data. (Accessed 6 November 2018).
gation, Data curation, Writing - original draft, Visualization. Bala [25] B. Jamil, A.T. Siddiqui, N. Akhtar, Estimation of solar radiation and optimum
tilt angles for south-facing surfaces in Humid Subtropical Climatic Region of
Pesala: Conceptualization, Verification, Resources, Writing - review India, Eng. Sci. Technol. Int. J. 19 (2016) 1826e1835, https://doi.org/10.1016/
& editing, Supervision, Project administration, Funding acquisition. j.jestch.2016.10.004.

1761
S. Dey and B. Pesala Renewable Energy 162 (2020) 1747e1762

[26] C. Stanciu, D. Stanciu, Optimum tilt angle for flat plate collectors all over the building applications in mid-latitude zone, Energy Convers. Manag. 124
World e a declination dependence formula and comparisons of three solar (2016) 20e28, https://doi.org/10.1016/j.enconman.2016.06.066.
radiation models, Energy Convers. Manag. 81 (2014) 133e143, https:// [30] C.K. Lin, C.Y. Dai, J.C. Wu, Analysis of structural deformation and deformation-
doi.org/10.1016/j.enconman.2014.02.016. induced solar radiation misalignment in a tracking photovoltaic system,
[27] M. Gulin, M. Vasak, M. Baotic, Estimation of the global solar irradiance on Renew. Energy 59 (2013) 65e74, https://doi.org/10.1016/
tilted surfaces, in: 17th Int Conf Electr Drives Power Electron (EDPE 2013), j.renene.2013.03.031.
2013, p. 6. [31] P. Krishna, K. Kumar, N.M. Bhandari, IS : 875 ( Part3 ): Wind Loads on
[28] W.G. Le Roux, Optimum tilt and azimuth angles for fixed solar collectors in Buildings and Structures, vol. 875, 2012.
South Africa using measured data, Renew. Energy 96 (2016) 603e612, https:// [32] Singularities in Finite Element Models: Dealing with Red Spots | COMSOL Blog
doi.org/10.1016/j.renene.2016.05.003. n.d. https://www.comsol.com/blogs/singularities-in-finite-element-models-
[29] S. Soulayman, M. Hammoud, Optimum tilt angle of solar collectors for dealing-with-red-spots/(accessed April 9, 2019).

1762