Applied Energy 145 (2015) 211222

Contents lists available at ScienceDirect

Applied Energy
journal homepage: www.elsevier.com/locate/apenergy

Enhancing the efciency of solar concentrators by controlled optical

aberrations: Method and photovoltaic application
Alessandra Giannuzzi a,, Emiliano Diolaiti a, Matteo Lombini a, Adriano De Rosa b, Bruno Marano c,
Giovanni Bregoli a, Giuseppe Cosentino c, Italo Foppiani a, Laura Schreiber a
a
b
c

INAF-Osservatorio Astronomico di Bologna, Via Ranzani 1, I-40127 Bologna, Italy

INAF-Istituto di Astrosica Spaziale e Fisica Cosmica, Bologna, Via P. Gobetti 101, I-40129 Bologna, Italy
DIFA-Dipartimento di Fisica e Astronomia, Alma Mater Studiorum Universit di Bologna, Via Ranzani 1, I-40127 Bologna, Italy

h i g h l i g h t s

g r a p h i c a l a b s t r a c t

 We developed a new optical design

method for high performance solar

concentrators.
 The method is based on optimizing
the optical shapes to match the
receiver features.
 A dense array PV concentrator made
by few monolithic mirrors was
modeled.
 The optimization led to free-form
optics focusing high uniform
irradiance spots.
 The optimal optics/receiver coupling
increases the system conversion
efciency.

a r t i c l e

i n f o

Article history:
Received 21 June 2014
Received in revised form 12 January 2015
Accepted 23 January 2015
Available online 2 March 2015
Keywords:
Photovoltaic concentrator (CPV)
Dense-array receiver
Numerical optimization
Optical design
Zernike polynomials

a b s t r a c t
We present a general method, based on controlled static aberrations induced in the reectors, to boost
receiver performances in solar concentrators. Imaging mirrors coupled with dense arrays suffer from severe performance degradation since the solar irradiance distribution is bell-shaped: mismatch losses occur
in particular when the cells are series connected. The method consists in computing static deformations
of the reecting surfaces that can produce, for an adopted concentration ratio, a light spot matching the
receiver features better than conventional reectors. The surfaces and the deformations have been
analytically described employing the Zernike polynomials formalism. The concept here described can
be applied to a variety of optical congurations and collecting areas. As an example, we extensively investigated a dense array photovoltaic concentrator, dimensioned for a nominal power of about 10 kWe. The
at distribution of light we obtain can exploit the PV device cells close to their efciency limit. A signicant gain is thus obtained, with no need of secondary optics or complex dish segmentation and of special features in the receiver electrical scheme. In the design, based on seven 2.6 m mirrors, we addressed
also non-optical aspects as the receiver and the supporting mechanics. Optical and mechanical tolerances
are demonstrated not to exceed accurate, but conventional, industrial standards.
2015 Elsevier Ltd. All rights reserved.

Corresponding author.
E-mail address: alessandra.giannuzzi@unibo.it (A. Giannuzzi).
http://dx.doi.org/10.1016/j.apenergy.2015.01.085
0306-2619/ 2015 Elsevier Ltd. All rights reserved.

212

A. Giannuzzi et al. / Applied Energy 145 (2015) 211222

1. Introduction
Concentrating Photovoltaics technology (CPV) is experiencing a
growing interest thanks to the development of solar cells with
continuously improved efciency. At present, the best reported cell
is a 0.165 cm2 multi-junction (MJ) cell having a new record of
44.4% conrmed efciency at direct irradiance concentration of
302 suns (1 sun = 1000 W/m2) [1]. For both high concentration
(HCPV) and low concentration (LCPV) systems the yearly installed
capacity increased signicantly during the last ve years [2].
A simple advantage induced by this technology is that, given the
collected energy, the concentration performed by optical devices
such as lenses or mirrors allows us to replace the area of photovoltaic material with cheaper optical surfaces. Moreover, high efciency cells are too expensive to be used in non-concentrating
applications. Despite most of the installed systems are point focus
lens based as Fresnel [36] or micro-dish [79] systems, dense
array systems have been recently investigated as protable solutions for lowering the cost per watt-peak supplied [10,11]. In this
technology the light is focused using one large reective element
called dish, onto an array of photovoltaic MJ cells densely packed
to form a single detector. If compared with lenses, mirrors have
the main advantage to not suffer from chromatic aberrations.
These systems track the sun in two-axis during its daily motion
and usually operate in high concentration mode, i.e. with solar ux
up to hundreds times the ambient value. Reective dish concentrators with diameters ranging from few meters to few tens of meters
have been already proposed and are at the beginning of their commercial development working at typical concentrations of 500
[1214].
Traditional dish concentrators have paraboloidal shapes.
Theoretically, their diameters could reach several tens of meters
as the heliostats in central tower plants, the construction of monolithic mirrors being difcult at these scales. The size generally
imposes to approximate the proles with cheap at reecting
facets mounted on a common frame and reproducing globally
the paraboloidal surface. As for the receivers, standard cells have
rectangular shapes and the arrays are groups of cells densely
packed together mostly in series and parallels connections. The
arrays do consequently resemble rectangular shapes too. When a
standard imaging mirror that produces a sun image intrinsically
circular is coupled with a rectangular detector problems arise. In
this condition some cells could be obscured if the spot is smaller
than the receiver, or part of the light could be lost if the detector
is smaller than the spot, these two effects contributing to a substantial loss in efciency. Moreover, the given irradiance distribution is bell-shaped in contrast with the requirement of having all
the cells under the same illumination. In fact, interconnected cells
having identical electrical characteristics and experiencing the
same irradiance/temperature conditions produce the same amount
of output current and voltage. Mismatch losses occur instead when
interconnected cells experience different conditions, in particular
for series connections. Still few investigations have been specically performed on current mismatches in dense array receivers
exposed to high concentrations [1517]. The issue of spatial light
uniformity is instead widely known for single cell devices [18
21] and the problem is commonly approached by the introduction
of secondary optics (SO) [2224] working as homogenizers. The
presence of an extra secondary optics is rather useful to increase
the acceptance angle leading to a relaxation of tracking and alignment tolerances. However, this solution has the disadvantage to
increase the system complexity and to add reection losses, chromatic aberration (if refractive) and mechanical problems as alignment, stability or mounting. A useful review on the state of the art
of the nonuniformity problem for single cell receivers has been
recently published [25]. Few commercial systems and technical

data are available on secondary optics embedded in dense arrays.

Some researches faced the uniformity problem from the receiver
point of view, developing new electrical connections [26],
embedding different cells in the same array [27] or designing
new receivers with radial symmetry [28].
Alternative ways of redesigning the primary collector have been
poorly investigated but some good results has been obtained by
Chong et al. [29]. The proposed planar faceted concentrator
coupled to a dense array has been optimized to give a large uniform illumination over the target area with a peak intensity of
391 suns. However, such a concentrator is made by several mirrors
to be mounted and aligned before being orientated with the use of
line-tilting driving mechanism. Moreover, since the nal spot is the
overlap of the multiple facets reections, the size and the uniformity of the nal spot is inuenced by projection and blocking effects
which increase with the distance of the facets from the center of
the whole assembly. For this reason, such a mosaic system is not
able to both have big collecting area and high concentration ratio
without embedding a high number of facets and high focal
distances, as reported in similar works [3032]. In [32] the economical viability is however claimed for a specic conguration
of faceted dense array system since a cost for the output power
below 2 euro/W has been calculated.
The strategy we suggest in this paper is to boost the spot uniformity by only acting on the primary reector but using monolithic
big surfaces and avoiding the dish faceting into numerous smaller
elements. In the proposed method, the shape of the mirrors is
analytically described by the Zernike polynomials and its optimization is numerically obtained to give a non-imaging optics able to
produce a quasi-square spot, spatially uniform and with prescribed
concentration. The free-form primary optics, optimized in this way
and validated by a ray tracing software, showed a substantial gain
in efciency without the employ of secondary optics. At the same
time, simple electrical schemes for the receiver are required. The
concept has been investigated theoretically modeling a CPV application including a conceptual development of non-optical aspects
as the design of the receiver and of the supporting mechanics.
For the proposed method and the specic CPV system developed,
a patent application has been led in Italy. A preliminary analytical
study, considering a residential utility, has been also performed in
order to understand the energetic and economic performance of
the system [33]. The analysis indicates that the maximum sustainable capital cost of the system ranges between 30,000 euros and
45,000 euros depending on the years which are considered for
the return of the investment (10 or 20 years respectively). Further
more detailed economical evaluations will be performed during
the future constructive phases of the project.

2. Optical concept
From an optical point of view there is no need for an accurate
image at the receiver of a solar concentrator. The optical design criteria rather concern with the optimal transfer of light between the
source and the target chosen. To solve matching issues in concentrators we thought to reinterpret optical concepts largely used in
astronomy, where an accurate image formation is an essential premise for efcient observations. In telescopes, controlled mirrors
deformations are introduced by actuators to balance the optical
aberrations that degrade the wavefront coming from an observed
source [3436]. What we developed instead is a sort of reverse
approach of the astrophysical method: the guideline is to apply
deformations (active or static) to the mirrors of the solar collectors
to introduce aberrations in the wavefront, thus degrading the solar
image and, in the case of a CPV dense array system, focusing a
squared spot with a prescribed irradiance. The result would be a

213

A. Giannuzzi et al. / Applied Energy 145 (2015) 211222

better match between the irradiance features and the receiver

performance.
The technical feasibility of our concept is supported by independent studies and projects involving technology transfer processes
from the astronomical instrumentation knowledge. Single monolithic reectors suitable for concentrators (3.1 m wide) have been
already realized in a customized furnace at the Steward Observatory Mirror Lab, at the University of Arizona [37]. A novel mirror
concept based on an active laminate consisting of an ultra-thin
(less than 1 mm) and ultra-light carbon-ber shell bonded to a
piezo-ceramic active layer have been recently investigated and
manufactured with the aim of reducing the cost of active mirrors
both in telescopes and concentrators [3840].
To describe the mirrors shape and to perform the optimization
for a CPV dish, we used the Zernike polynomials, an analytical tool
largely employed, especially in optics, to characterize functions
and data on a circular domain. They form an orthogonal basis on
the unit circle and real surfaces can be represented by linear
combinations of them. Every Zernike polynomial consists of three
components: a normalization factor, a radial component and an
azimuthal component. The radial components are polynomials
derived from the Jacobi polynomials, whereas the azimuthal
component is sinusoidal. As in the Noll formalism [41], the Zernike
polynomials can be dened in polar coordinates q; h:

p
p
n 1Rm
n q 2 cos mh
p
p

Z jodd n 1Rm
n q 2 sin mh
p
Z j n 1R0n q

Z jev en

Zernike mode
4th

11th

14th

2
3

3. Case of single on-axis mirror

An analysis we performed with the ray tracing software Zemax
showed that, starting from a spherical mirror, very few deformations described by specic Zernike polynomials (modes) can
strongly help in solving the uniformity and shape problem in dense
array receivers. Considering an imaging mirror with deformations,
its surface z (the so-called sag) can be approximated by the following formula:
N
X
cr2
p
Ai Z i q; h
1 1  1 kc2 r 2 i1

Table 1
Principal Zernike modes involved in this study.

where q is the normalized radial coordinate ranging from 0 to 1 and

h is the azimuthal angle ranging from 0 to 2p. In the formulas, m
represents the azimuthal frequency and n the radial degree, both
are integer and the condition m 6 n; n  jmj even must be satised. The index j is a mode ordering number and is a function of n
and m. Eqs. (1) and (2) exist for m 0 while Eq. (3) for m 0.
The double indexing scheme is useful for unambiguously describing
the functions. In the formulas, Rm
n q indicates polynomials with
radial dependence.

Fig. 1. Effects introduced on the Sun image by Zernike polynomials 4th, 11th and
14th.

where N is the number of polynomials, Ai is the coefcient associated to the ith polynomial, r is again the radial coordinate in the chosen units, q and h are the polar coordinates dened before. Eq. (4)
depends on the curvature c (which equals the reciprocal of the curvature radius) and the conic constant k. The rst term in the equation represents an ideal conic surface (spherical if k 0) while the
second term represents the deformations described by Zernike
polynomials. The number of terms needed for a good surface modeling grows together with the number of deformations occurring at
different scales.
For a single spherical mirror focusing on axis, we identied
three main polynomials: the 4th, the 11th and the 14th. Fig. 1
shows how the solar spot produced at a xed distance by a
spherical mirror can be modied by introducing controlled

deformations related to the three modes here mentioned. This

model can be also extended to mirrors with an off-axis focus: in
that case the number of Zernike modes involved in the spot
shaping is higher.
The identied modes are shown in 2D and 3D in Table 1. The
deformation associated with the 4th mode (defocus) basically
enlarges the image and contributes to spread the light quite
similarly to the effect of shifting the receiver plane. The 11th mode
(third order spherical) contributes to redistributing the rays maintaining an image radial symmetry and changing the image irradiance prole. These two polynomials do not have any impact on
the spot shape since they have no azimuthal dependence. A deformation corresponding to the 14th polynomial (vertical quadrafoil)
contributes to make a circular spot square along two preferential
directions rotated 45, depending on the coefcient sign. The effect
of this specic deformation is less evident if the mirror is in focus
mode: that is the reason for a combined use of the modes 14th and
4th. Alternatively, the same effect of this combination can be
obtained by positioning the receiver slightly behind or above the
correct focal plane and avoiding (partially or completely) the
deformations related to the 4th mode. Since it is easier for a single
mirror to produce a square uniform image when the defocus is bigger, this means that the lower the concentration factor the better
the method works. The size of the spot to obtain depends on the
desired concentration factor.
A prescribed irradiance could be also obtained by employing
this concept to design concentrators with several optimized mirrors focusing at the same receiver. In this case, the nal illumination pattern impinging on the receiver would result in the sum
of the incoherent illumination patterns produced by each single
mirror, as we are going to show in the next sections.

4. Case of a CPV dense array system: design choices

A multi-mirror conguration can be useful to solve the issue of
building a single huge mirror. In order to avoid a mosaic of hundreds reective elements [15], we choose to design a CPV dish
made by few monolithic mirrors mounted close together on the

214

A. Giannuzzi et al. / Applied Energy 145 (2015) 211222

Fig. 2. Optical layouts: (a) 3D, (b) xy plane, and (c) yz plane.

same structure. The selected conguration is the hexapolar grid

and it has been already used in Stirling applications as well as in
some ground based optical telescopes. In the hexapolar conguration the elements are placed on rings so that the (n + 1)th ring contains six elements more than the nth ring, the central ring having
only one element. We decided to consider only the central mirror
and a ring of six mirrors arranged around it. Fig. 2 presents the
optical layouts of the proposed system. The mirrors of the second
ring have been labeled from 2 to 7 counter-clockwise. The z-axis
has been set as the direction of the incoming rays and it is perpendicular to the central mirror vertex. This optical condition of alignment with the solar direction should be the system nominal
working state.
Considerations about the concentration ratio to be investigated
and the mechanical compactness have been made also in comparison with similar existing prototypes and plants. Since this
research activity has been carried out with the specic goal of nding new solutions in the eld of clean micro-generated distributed
electricity, our dish has been conceived as a power system suitable
for the market of medium residential contexts or small farms. We
decided the mirror diameter to be around 23 m, to avoid construction difculties. The diameter of the single mirror has been
set to D 2600 mm, for a total system size of about 7800 mm
and a resulting total optical area slightly bigger than 35 m2. Supposing an irradiance at the collecting aperture of 1000 W/mm2,

the entry power would be around 35 kW: with a receiver working

almost at the efciency of the best presently available cells (between 30% and 40%), such a system would be able to deliver more
than 10 kWe. Utility scale applications could be anyway considered, together with the scaling of the single elements for higher
energy outputs.
The detector distance has been set to h 4800 mm in order to
have a low ratio of detector distance to total diameter. Considering
this ratio similar to the focal ratio in imaging systems, a value f =0:5
should be approached to maximize the concentration but also to
allow a more compact structure.
We investigated two concentration levels, 500 and 800. To
obtain these concentrations, we applied a defocus to the mirrors
which is the common method to modulate the concentration delivered at the receiver. A paraboloid in focus mode would have a collected ux too high for the cells working range (up to few
thousands of suns at present). In our case, another reason to avoid
extreme concentrations is that the deformations introduced by the
Zernike modes are more efcient in reproducing the image features required when a defocus occurs.
The concentrator has been initially designed putting mirrors
with the same diameter D on the same plane. The reference system
has been chosen so that incoming rays are parallel to the z-axis,
while the mirrors vertexes lay in the xy plane. Each mirror has
been placed at d 2680 mm (in the xy plane) from the central

215

A. Giannuzzi et al. / Applied Energy 145 (2015) 211222

Table 2
Positions, tilt angles and curvatures of the seven mirrors.

X pos (mm)
Y pos (mm)
ax ()
ay ()
Radius of curv. (mm)

Mirr1

Mirr2

Mirr3

Mirr4

Mirr5

Mirr6

Mirr7

0.00
0.00
0.00
0.00
10101.00

0.00
2680.00
14.59
0.00
11480.10

2320.88
1340.00
7.41
12.60
11480.10

2320.88
1340.00
7.41
12.60
11480.10

0.00
2680.00
14.59
0.00
11480.10

2320.88
1340.00
7.41
12.60
11480.10

2320.88
1340.00
7.41
12.60
11480.10

mirror vertex to prevent shading effects. The mirrors of the external ring have been tilted respect to the central one in order to focus
all the chief rays from the Sun center at the center of the receiver
plane having coordinates 0; 0; h. This optical restriction is optional, but we aimed at simplifying the mechanical structure. The
geometrical laws fullling this optical condition are easily derivable and once xed the distance d in the hexapolar grid the positional/tilting parameters of the mirrors can be immediately
calculated. The tilt of the external mirrors reduce by 5% the collecting projected area of the whole system from 37.17 m2 to about
35.25 m2. Positions, tilts and curvatures of the seven mirrors are
listed in Table 2. The generic mirror surface sag has been described
by Eq. (4).

5. Design method
To optically model our system, an end-to-end IDL code has
been written on purpose. Each step of the procedure and the
results have been veried with the optical design software Zemax
as reference. The code includes four main subgroups of routines:
the rst for individually modeling the optical part; the second for
the receiver implementation; the third for optimizing the optics;
the last one for calculating tolerances of optical/mechanical
parameters.
5.1. Optical modeling
The initial optical parameters, which are the initial conditions of
the simulations, have been set by a ray tracing analysis performed
by Zemax. The Sun has been modeled as a nite source with an
angular diameter of 0:53 , neglecting its shape variations caused
by the altitude changing during the day. The curvatures have been
set so that the mirrors could produce a spot with a size compatible
with the mean geometrical concentration chosen. The concentration ratio has been dened as the total mirrors area perpendicular
to the axis of the central mirror divided by the total area of the
receiver, supposing a receiver and a spot ideally with the same size.
We ignored the obscuration introduced by the receiver itself.
The Zernike modes corresponding to deformations useful to fulll our requirements of shape and uniformity have been selected
after xing the curvature. The deformations needed for the central
mirror are the three described in paragraph Section 2, but other
modes (from 5th to 8th) are necessary for the six off-axis mirrors.

The selection criteria is that the superimposition of all the generated spots could produce an irradiance distribution with the desired
features. Symmetry properties have been imposed for the six mirrors in the external ring to reduce the degrees of freedom of our
problem. For example, these mirrors have been chosen with the
same curvature radius and the same values of the 4th, 11th and
14th Zernike coefcients. As consequence, the non-zero coefcients are linked between mirrors by the geometrical relations
shown in Table 3. In this way, opposite mirrors are equal but rotated by p and the nal optical model results to be made of only four
different types of surfaces. It could be certainly possible to identify
more coefcients to improve the performance however increasing
the complexity of the system. This condition would be more suitable both on construction and calibration stages. The independent
modes identied for our system are eight, three for the central mirror (Z4(1), Z11(1) and Z14(1)) and ve for the external ones, all
derived from the modes of the mirror number 2 (Z4(2), Z6(2),
Z7(2), Z11(2), Z14(2)) according to the relations shown in Table 3.
The mirrors of the ring can not have all the same shapes even if this
would be the best constructive condition. The 14th Zernike mode
in fact corresponds to a deformation able to modify the circular
symmetry of the ray bundle into a square and it has an azimuthal
dependence. The simple rotation of a given surface would lead to a
different analytical description in terms of its Zernike coefcients,
except for the coefcients with pure radial dependence. This
means that a ring generated by replicating mirror number 2 and
simply rotating the replicas according to the position in the ring,
would give a series of identical spot rotated as in Fig. 3a. The superimposition of these spots would certainly not lead to a nal square
shape. On the contrary, xing the 14th coefcient to the same value for all the surfaces, the features in Fig. 3b are obtained. The physical size of the gure is 4  105 lm.
The optical scheme described is simulated by the ray-tracing
code written on purpose. The code output is the nal spot produced by the concentrator. In the algorithm, the continuous optical
surfaces of the mirrors have been discretized by a xed number of
sub-apertures. The rays striking every sub-aperture are reected
toward the receiver according to the classic reection law. The
Sun has been modeled as an homogeneous circular source with a
diameter of 0.53, thus applying a realistic divergence model. The
number of rays traced from the Sun has been set in order to minimize sampling errors. To calculate the nominal mirrors shape, we
supposed an ideal tracking condition in which the central solar
ray strikes the central mirror vertex parallel to the optical axis.

Table 3
Correlation between the Zernike coefcients of the seven mirrors.

Z4
Z5
Z6
Z7
Z8
Z11
Z14

Mirr1

Mirr2

Mirr3

Mirr4

Mirr5

Mirr6

Mirr7

Z4(1)
0.00
0.00
0.00
0.00
Z11(1)
Z14(1)

Z4(2)
0.00
Z6(2)
Z7(2)
0.00
Z11(2)
Z14(2)

Z4(2)
Z6(2) cos 30
Z6(2) sin 30
Z7(2) sin 30
Z7(2) cos 30
Z11(2)
Z14(2)

Z4(2)
Z6(2) cos 30
Z6(2) sin 30
Z7(2) sin 30
Z7(2) cos 30
Z11(2)
Z14(2)

Z4(2)
0.00
Z6(2)
Z7(2)
Z7(2)
Z11(2)
Z14(2)

Z4(2)
Z6(2) cos 30
Z6(2) sin 30
Z6(2) sin 30
Z6(2) cos 30
Z11(2)
Z14(2)

Z4(2)
Z6(2) cos 30
Z6(2) sin 30
Z7(2) sin 30
Z7(2) cos 30
Z11(2)
Z14(2)

216

A. Giannuzzi et al. / Applied Energy 145 (2015) 211222

Fig. 3. Effect introduced in the spot generated by each mirror by the introduction of (a) a Z14 value rotated according to the mirror location and (b) a common Z14 value.

5.2. Receiver implementation

Table 5
Electrical parameters of the AZUR SPACE 3C40 cell at 500 and 1000.

To simulate the performance of a dense array receiver, we considered an electrical model for the PV cells. Neglecting any temperature or spectral variation, the physical behavior of a cell can
be in rst approximation summarized by the following set of equations uniquely depending on the concentration factor :

Isc    Isc 1

KT ln
q
Pmax  Imax   V max 

V oc  V oc 1 nd

6
7

Pmax 
FF
Isc   V oc 
P 
FF
gmax  max
Isc   V oc  
Pin 
P in 

8
9

where Pin is the total power received by the cell and Isc ; V oc 
are short circuit current and open circuit voltage at a given concentration, gmax is the nominal conversion efciency, nd is the diode
ideality factor, T is the absolute temperature of the cell, K is the
Boltzmann constant and q is the electron charge. A more exhaustive
model involving dependencies on T and spectral variations can be
found in [42]. Eq. (8) denes the Fill Factor FF as the ratio between
the power at the maximum power point Pmax and the product of the
open circuit voltage and short circuit current. It is typically better
than 75% for good quality MJ solar cells. It is also an index of the
performance of a solar cell in terms of generated power and it
should be as close as possible to 100%: graphically, the FF is a measure of the squareness of the solar cell IV curve and is also the area
of the largest rectangle which would t in the curve.
Our receiver has been analytically designed and numerically
simulated using a datasheet of commercially available high concentration cells 3C40 produced by AZUR SPACE [44] with a nominal efciency of 39% at 500 (around 38% at 1000) at ambient
temperature. The reference cell has main features described in
Table 4.
In addition to efciency, the cell datasheet gives other output
parameters (Table 5) necessary in the simulations to predict the
cells power output at different illuminations. Moreover, since we
deal exclusively with reective elements, no chromatic aberration
are introduced. The temperature can also be considered reasonably

Table 4
Main features of the AZUR SPACE 3C40 cell implemented in the simulations.
Base material
AR coating
Chip size
Active cell area

GaInP/GaAs/Ge on Ge substrate
TiOx/Al2Ox
5.59  6.39 mm2 = 35.25 mm2
5.5  5.5 mm2 = 30.25 mm2

500
1000

Isc (A)

V oc (V)

Imax (A)

V max (V)

P max (W)

FF (%)

g (%)

2.151
4.239

3.144
3.170

2.102
4.135

2.842
2.762

5.98
11.42

88.0
85.0

39.0
37.8

constant as efcient cooling systems have been shown in

literature.
The receiver electrical design has been chosen in order to minimize the power matching problem even maintaining high degree of
linearity and easiness of construction: attention has been paid to
series connected cells since the output current in this case corresponds to the current produced by the worst illuminated cells of
the series.
The choice of the number of cells to connect has been made
starting from the concept that a receiver should have a certain area
to perform at a certain mean concentration. The array design has to
resemble, with the right connections, an irradiance distribution
which is mostly square and uniform and probably degrading
toward the borders. To simplify the scheme, we decided to simulate different receivers starting from the same base unit, which is
a string of series connected cells. A scheme with many parallels
would lead to a lower dependence from irradiance gradients, but
it has the inconvenience to give high current and small voltages
in output. High voltages are instead more suitable for the standard
range of inverters while small currents limit the resistive losses.
We thus chose to conceptually design different receivers type to
perform at different output voltages. Fig. 4a and b shows the third
of the array implemented for which we will show also the tolerance results. It is a detector made by 56 strings of 36 cells. The
strings spatial positioning is shown in Fig. 4a where each string
is represented by a narrow rectangle. There are 32 strings in the
central square zone, which corresponds roughly to the maximum
uniform area obtainable by the optimization, and 4 lateral zones
made by 6 additional modules each. The total number of cells is
2016. This scheme allows cells in series to be irradiated with similar uxes and at the same time, the strings and the groups contain
the same number of elements thus ensuring small parallel mismatches. This scheme does not have cells at the corners, since
the spillage losses in case of 500 have been evaluated in the order
of 5%. The electrical connections are arranged as follows (Fig. 4b):
cells in each strings are series connected as well as strings with the
same color. The central zone is then made by 8 blocks of cells each
containing 4 adjacent substrings (the subdivision of each colored
areas have been omitted), while the lateral strings are series connected in concentric frames. The 14 resulting blocks are nally parallel connected.
The same electrical scheme has been also used for simulating
the concentration 800. In this case the cells of the base string

217

A. Giannuzzi et al. / Applied Energy 145 (2015) 211222

200

200

(a)

(b)

Y (mm)

-100

-100

-200

-100

100

-200
200

X (mm)

Y (mm)

100

100

-200

-100

100

-200
200

X (mm)

Fig. 4. Type-3 receiver design at 500. The (a) panel shows the subdivision in strings. The (b) panel shows which strings are series connected (zones with the same color). The
14 resulting blocks are parallel connected. (For interpretation of the references to colour in this gure legend, the reader is referred to the web version of this article.)

are only 27 and the central zone is made by 24 strings since the
higher concentration results in a smaller irradiated area. The parallel connected blocks are 12. Spillage losses at the corners are
around 810% but again we preferred to preserve the array symmetry avoiding cells in these areas.
To analytically calculate the electrical performance, we developed a routine implementing the Eqs. (5)(9) modeling the cell
output current and voltage as functions of concentration, neglecting resistive effects. As for the electrical scheme, the routine implements the classical equations for calculating voltages and currents
in series and parallel connections. Only these connections are
involved while no model has been implemented for the bypass
diodes. A temperature of T 298 K has been considered and a reasonable value for the ideality factor nd 3:3 has been assumed to
treat the junctions as real. The other initial parameters used are in
Table 5. Being FF only dependent on V oc , it has been calculated
using a classical empirical formula [43] approximated for zero
resistivity:

FF

v oc   lnv oc  0:72
1 v oc 

10

where v oc  is the open circuit voltage normalized by the factor

nd KT=q.
5.3. Optimization procedure
The optimization procedure employs a downhill simplex
method. We decided to minimize a merit function related to conversion efciency. In particular it has been dened as the negative
efciency of the receiver g as dened in Eq. (9): each evaluation
of this function requires the calculation of the efciency by the ray
tracing procedure and the receiver model previously explained. We
summarize the optimization steps as follows.
The initial values chosen for the parameters to be optimized are
inserted in the optimization routine. The routine operates performing a multidimensional minimization of a function funcx where x
is an n-dimensional vector of parameters, using a downhill simplex
method requiring only function evaluations and not derivatives.
Additional input for the routine are the fractional tolerance to be
achieved in the function value as well as the range of the parameters variation.
The optimization procedure transfers the parameters value to
the ray-tracing procedures which gives the image as output, then
the block simulating the receiver performance gets in input the
image focused by the optics. The image is represented by a matrix
containing the local concentration impinging on each receiver cell.

The receiver model distinguishes between cells series and parallel

connected, imposing the current of a series cells as the current produced by the worst illuminated cell. Subsequently, the current and
voltage output for each series/parallel are summed to give the total
output and the efciency. After calculating the efciency of the
optics coupled with that receiver, the procedure changes the parameters value iteratively in the range specied, modifying the
optics and calculating a new image, a corresponding new efciency
and comparing the values of the simplex obtained. When the minimum is found within the threshold, the routine returns an n-element vector corresponding to the function minimum value. This
kind of method could be applied to other type of receivers and it
could be improved by extending the variables (for example the
curvatures that here we considered xed).
5.4. Tolerance calculation
After obtaining the nominal image produced by the optimized
optics, a tolerance calculation has been implemented to assess
the feasibility of the results. Tolerances have been obtained for
both optical and geometrical parameters. We considered 25 parameters for each of the 4 different mirrors. Additional parameters
are the two tracking angles and the receiver position along the zaxis, for overall 178 parameters. The parameters include tilts and
positions of the mirrors, their curvatures and the Zernike coefcients up to the 6th radial order (from 4th to 21th). The reason
for considering up to this order lays in the connection between
the radial degree of the polynomials and the spatial scale of the
deformations: the degree of a polynomial on a certain surface
(which has a diameter of 2.6 m in the proposed design) roughly
denes the spatial scale (period) of the associated deformation so
that, for example, a 6th degree deformation on 2.6 m diameter
would be roughly half meter (2:6=6 m 0:43 m). It has been
evaluated that higher degree deformations, i.e. occurring on spatial
scales smaller than the considered scale, can be reasonably controlled by surface polishing of candidate materials (aluminum,
molded plastics, etc.). The tolerances have been also calculated
for polynomials with nominal null coefcients since all the polynomials up to a certain degree are necessary to model the irregularities down to a given scale.
The nominal image produced by the optics with the optimized
parameters and the corresponding receiver efciency have been
calculated and stored as terms of comparison. We chose a range
of variation for each parameter and a minimum tolerable efciency. The tolerated efciency degradation was equally split among all
the parameters, assuming their effects as uncorrelated. Degraded
efciency has been calculated for the minimum and maximum

A. Giannuzzi et al. / Applied Energy 145 (2015) 211222

Table 6
Values in mm of the Zernike coefcients optimized at the two considered concentrations considering type-3 receivers.

500
800

Z4(1)

Z11(1)

Z14(1)

Z4(2)

Z6(2)

Z7(2)

Z11(2)

Z14(2)

1.124
1.103

0.137
0.070

0.098
0.108

1.486
1.053

0.616
0.714

0.223
0.280

0.003
0.019

0.217
0.144

values of a given parameter, keeping nominal values for all the

other parameters: if the degraded efciency is acceptable, the
minimum and maximum values of the given parameter are adopted as tolerances for that parameter, otherwise the variation range
of the parameter is reduced and the process is repeated until convergence. After computing the tolerances for each parameter
separately, the global effect has been evaluated by perturbing all
the parameters simultaneously in a random fashion according to
the computed tolerances and evaluating the corresponding
efciency.

6. Results: the SOLARIS concentrator

The results shown in Table 6 have been obtained by optimizing
our optics at two concentrations (500 and 800) with type-3
receivers. The values of the Zernike coefcients not shown can be
derived from the relations in Table 3.
The bi-dimensional and the x-cross section irradiance produced
by the optimized optics have been simulated by Zemax for the
two concentration ratios and they are shown in Figs. 5 and 6.
The x-cross section irradiance is evaluated on the central row parallel to the x-axis of the bi-dimensional irradiance pattern. All the
simulations have been performed supposing 1 sun irradiance at the
concentrator aperture, which is the common value in Standard Test
Conditions (STC).
The performance obtained for other receivers types described in
Section 5.2 are listed in Table 7. The efciency g is the output power of the receiver divided by the total power collected by the optics.
The optimized systems show a conversion efciency of about 30%
in all the cases with 500 and of 28% in the only analyzed case
with 800. The case with higher concentration is interesting for
the development of new generation cells because it shows that
the proposed method gives good results also at higher concentrations. Moreover, the higher the concentration the smaller the number of cells employed in the receiver. The case with concentration
800 in fact includes only 1152 cells, almost half of the cells needed for the concentration 500 (2016 elements).
The relative efciency grel in Table 7 has been dened considering the only effective power impinging on the array, i.e. accounting

for spillage losses at the corners/edges. This parameter is useful to

evaluate the average cells performance in the array. In three of the
four cases, its value is above 31% and it must be compared with the
maximum theoretical efciency reported in Section 5.2 for the
active part of the cell considered, i.e. 33% for concentration 500
and 32% for 1000. This means that the cells in the arrays work
really close to their nominal performance under the irradiance produced by the optimized optics.
Looking at the results in Table 7 with concentration 500, the
main difference between the three receivers analyzed lays in the
output parameters values. Even if the total power produced is quite
similar in all the cases (slightly higher than 10 kWe), the output
current and voltage are very different. The third receiver has been
designed specically with a high number of series connections to
obtain a high voltage value (409.2 V) suitable for the available
inverters and with small current (25.3 A) to limit the resistive losses. This condition is convenient from an electrical point of view,
but it leads to tighter tolerances, as shown below.
The tolerance results are here shown only for the concentration
500 with the type-3 receiver, giving some qualitative indications
for the other cases studied. The parameters which differ from mirror to mirror are summarized in Tables 8 and 9 while the common
parameters related to the receiver position are shown in Table 10.
Five out of seven mirrors have been omitted from the list since
their tolerances are similar to those of the second mirror except
for discretization effects. The last row in Table 8 is the root square
sum (RSS) of the Zernike coefcients and it is one of the most
important tolerance indicators in our analysis since it represents
the tolerated surface sag deviation. For all the mirrors, this parameter is in the order of tenths of a millimeter. The shape deviation
tolerated is also compatible with the manufacturing irregularities
of candidate materials (molded plastics or aluminum) for the
deformed/deformable mirrors. The tracking errors shown in
Table 10 are quite small if compared to other CPV concentrators
(normally in the order of 1 milliradian or more). In any case, the
tracking accuracy can be achievable with standard tracking solutions commonly employed in telescopes since these systems can
also reach subarcseconds tolerances. Good pointing and active
tracking systems are already developed also for solar concentrators
[45], but their performances should be further improved to allow
our tolerances.
The calculations have been performed setting a threshold of 3%
on the efciency, i.e. tolerating a degradation of the performance
from 29.4% down to 26.4%. This value has been chosen as reasonable for this type of systems, but it can be varied depending on
the required performance. In general, for small perturbations, the
tolerance on a parameter scales linearly with the threshold value.

(a)

(b)

0.6
0.5
0.4
0.3
0.2
0.1
0.0

-150

-100

-50

50

100

150

Irradiance (Watts/millimeters squared)

218

X position at Y=center (mm)

Fig. 5. (a) 2D and (b) x-cross section irradiance produced by the optics coupled to the type-3 designed for 500. The physical size of the gures is 350 mm. Units in the color
bar are W/mm2. (For interpretation of the references to colour in this gure legend, the reader is referred to the web version of this article.)

219

0.8

(a)

(b)
0.6

0.4

0.2

0.0
-150

-100

-50

50

100

150

Irradiance (Watts/millimeters squared)

A. Giannuzzi et al. / Applied Energy 145 (2015) 211222

X position at Y=center (mm)

Fig. 6. (a) 2D and (b) x-cross section irradiance produced by the optics coupled to the type-3 designed for 800. The physical size of the gures is 350 mm. Units in the color
bar are W/mm2. (For interpretation of the references to colour in this gure legend, the reader is referred to the web version of this article.)

Table 7
Electrical performance obtained after the optimization run with the three receivers
implemented.

Receiver
Receiver
Receiver
Receiver

1
2
3
3

(500)
(500)
(500)
(800)

Iout (A)

V out (V)

P out (W)

g (%)

grel (%)

98.7
50.5
25.3
32.6

105.2
204.6
409.2
302.6

10288.0
10324.8
10354.5
9868.1

29.2
29.7
29.4
28.0

30.5
31.6
31.2
31.4

Table 8
Zernike coefcients tolerances calculated for the system with 500 coupled with a
type-3 receiver.
Units

mm

mm

Parameter

Z4
Z5
Z6
Z7
Z8
Z9
Z10
Z11
Z12
Z13
Z14
Z15
Z16
Z17
Z18
Z19
Z20
Z21
q
P 2
Z

Mirr1

Table 9
Tolerances on other parameters calculated for the system with 500 coupled with a
type-3 receiver.
Units

mm

Tolerance

Nominal
value

Tolerance

1.124
0.000
0.000
0.000
0.000
0.000
0.000
0.137
0.000
0.000
0.098
0.000
0.000
0.000
0.000
0.000
0.000
0.000

0.063
0.063
0.250
0.031
0.031
0.031
0.031
0.008
0.016
0.016
0.016
0.016
0.016
0.004
0.016
0.008
0.016
0.016

1.486
0.000
0.616
0.223
0.000
0.000
0.000
0.003
0.000
0.000
0.217
0.000
0.000
0.000
0.000
0.000
0.000
0.000

0.063
0.031
0.063
0.016
0.016
0.031
0.016
0.016
0.008
0.008
0.031
0.016
0.004
0.008
0.008
0.008
0.016
0.008

0.2762

0.1122

The tolerances are strictly related to the electrical scheme implemented in the receiver. For example we calculated that with the
receiver involving more parallels and with the same threshold,
the tolerances would be three times more relaxed. In that case
higher output current would be produced, the output power being
approximately the same.
The mechanical model is shown in Fig. 7. From the analysis of
the Zernike polynomials, the desired deformations on the mirrors
can be applied by a restricted number of actuators positioned on
a certain number of control points. For the system with the chosen
dimensions, these points are located radially on three circumferences every 10. A possible way to obtain the nal surfaces is to

Radius of
curv.

Mirr1

Mirr2

Nominal
value

Tolerance

Nominal
value

Tolerance

10101.0

25.0

11480.1

25.0

mrad

tilt x
tilt y
tilt z

0.0
0.0
0.0

0.4
0.9
1.7

254.6
0.0
0.0

0.2
0.1
1.7

mm

offset x
offset y
offset z

0.0
0.0
0.0

5.0
2.5
25.0

0.0
2680.0
0.0

2.5
2.5
3.1

Mirr2

Nominal
value

Parameter

Table 10
Tolerances calculated for to the common parameters.
Units

Parameter

All mirrors
Nominal value

mrad

Tracking error x
Tracking error y

0.0
0.0

mm

Receiver offset z

4800.0

Tolerance
0.11
0.01
2.5

use spherical mirrors and to set the deformations by the actuators.

Another approach is to freeform mirrors already shaped with the
nal form desired, the actuators being employed only to compensate the shape errors once the mirrors have been placed on their
own support. All these mirrors could be made by aluminum sheets,
since this material is particular suitable for its lightness and its
ductility. Molded plastic could be also a candidate substrate material (if compatible with the requested tolerances) after the deposition of a high reective layer. During the realization, the system
should be aligned within tolerances. For this reason, we conceived
a 2-step procedure. The rst phase consists in the mirrors positioning on their own supports and the calibration of their nominal
shape. This test can be performed in laboratory and it requires a
point light source, a beam splitter, a ShackHartmann (SH) wavefront sensor [46] with a camera. The camera acquires the image
of a point source reected back by the mirror which can be used
to recognize the wavefront shape and the mirror surface map.
The actuators are tuned iteratively until the measured surface
map matches its nominal value (within tolerances, see Tables 9
and 10). To accelerate the calibration procedure, an interaction
matrix records the SH sensor reaction to the specic movement

220

A. Giannuzzi et al. / Applied Energy 145 (2015) 211222

Fig. 7. Shaded models of the SOLARIS concentrator: (a) front side and (b) rear side.

of each single actuator. This matrix has to be inverted and used to

transform the SH sensor signal into incremental corrections to
apply to the actuators. The second stage is an alignment on Sun
of each mirror on the whole frame. A mask dimensioned as the
receiver and realized in a material resistant to temperatures of a
few hundreds degrees is needed. Concentric frames of pinholes
on the mask transmit part of the light impinging on the receiver
plane to diodes or other electronic light-sensitive devices. Such a
tool allows to sample the irradiance distribution produced by the
optics and to adjusted iteratively the position of each mirror on
the common frame until the desired irradiance is obtained.
Another interaction matrix is used to record the diodes reaction
to the parameters to align. This matrix is then inverted and used
to translate the measured signal into corrections for the mirror
positioning.
The new concentrator resulting from the investigation carried
out has been called SOLARIS (SOLAR Image Squaring) Concentrator and it has been patented in Italy. The patent is owned by both
the University of Bologna and the National Institute of Astrophysics (INAF), the two research institutes involved in the project.
Main subjects of the patent are both the innovative concentrating
CPV application and the method for the numerical optimization of
reective surfaces. The procedures to test/calibrate the reective
shapes and to align the mirrors on Sun, as well as the receiver
and the mechanical design are all parts of the patent. The model
and the obtained results will be validated with the described procedures during the forthcoming prototyping stage.
7. Summary and discussion
We developed a new optical designing method for solar concentrators. In particular, dense array photovoltaic applications need an
accurate control on both shape and irradiance of the collected light
spot to perform at high efciency. These systems are experiencing
in the last years growing interest (from market and research) as
feasible solutions in the production of cost competitive electricity
on demand, especially in very sunny environments and off-grid
communities. The development of solar cells that can work at very
high irradiance imposes a technological jump also from an optical
point of view, to let these systems work at the same performance
of the employed cells. The proposed method is based on controlling
the optical shapes so that the spot produced by the mirrors can
resemble the optimal features for the chosen receiver without

including secondary optics. The deformations to apply have been

analytically modeled by the Zernike polynomials and the deformed
mirrors have been simulated by ray tracing routines developed on
purpose. At the same time, different schemes of dense array receivers have been designed using reference cells with known features
and simulated by implementing simple electrical models for photovoltaic devices. The deformed optics have been numerically optimized to maximize the performance of the concentrator as a
function of the coupled receiver. The method has been fruitfully
employed to solve the prescribed irradiance problem at high concentration in CPV dishes. It has led to the design of a novel CPV
optics, the SOLARIS concentrator. Both the method implemented
and the specic application developed have been patented in Italy.
The main advantage of using big monolithic mirrors is to have
few optics to manage respect to the complex segmented optics
proposed in other researches involving dense arrays. Despite this
technology is quite recent and commercial plants are not as diffused as the refractive fresnel lens based systems, our method to
design dense array concentrators opens a new scenario for developing PV systems that could perform at very high efciency working at high concentrations. This efciency boosting up to nominal
levels and, at the same time, the relaxation of the constraints on
the receiver design and the recent development of new materials
for optical application suggest interesting perspectives of cost
reduction.
The concentrator developed is a single stage multi-mirror system made by 7 monolithic optics placed in an hexapolar arrangement and all focusing on the same receiver. The principal
investigated design has a mean concentration ratio 500. The
deformations applied to the optics allow them to produce a solar
spot resembling a square shape with smoothed corners. The irradiance pattern inside the spot obtained is highly uniform. At this
concentration, the optimized optics can boost the conversion efciency of the whole receiver up to 30%, almost the same theoretical
performance of the single cell used in the calculations which is
around 33% (considering only the active areas). The receiver has
been designed as simple as possible, using exclusively strings of
identical cells in series. The strings are then organized in parallels
or series connections, with a Cartesian conguration and not
involving bypass diodes in the design.
From an optical point of view, different considerations can been
made to extend the purposes and the applications of the method
conceived. Similar systems with different concentrations can be

A. Giannuzzi et al. / Applied Energy 145 (2015) 211222

surely designed ever keeping in mind the optimization method has

been tested for the two concentration 500 and 800, and that the
results are better in the rst case considered thanks to the higher
defocus involved. Despite this, we demonstrated that our method
can work efciently also at many hundreds of concentration ratio.
Method improvements could be done by a further investigation
of the convenient deformations to introduce, exploring for example
the effects related to Zernike polynomials of higher degrees. The
selected deformations and the optical conguration used in this
work are indeed only an example of the method proposed: other
concentrators could be designed by adding deformations or changing the geometrical/optical parameters as a function of the desired
spot features. Systems with single or multiple mirrors (different or
not) could be implemented and different geometrical congurations explored. Also the mirrors aperture could be varied in shape
and size depending on the amount of output power needed or on
the economical/constructive constraints. The nal spot could result
from a superimposition of images not necessarily centered in the
same point, as in the studied cases. Another interesting application
could result from exploring the performance of deformable optics
including very simple reective secondary optics to recover possible light losses at the receiver borders or to relax the tolerances
(thus enhancing the acceptance angle).
A great advantage of employing actively deformable optics
could be given by the tuning of the concentration ratio. Using convenient deformable materials, exible systems could be obtained
embedding different type of receivers but exploiting the same
optics. Also from the receiver point of view, great improvements
could be obtained in terms of electric efciency, involving optimized electrical schemes or thinking to future monolithic receivers. Finally, an extension of this method could be also helpful in
solving thermal problems. Thermal concentrators do also need a
certain uniformity in the light collected to optimally transfer the
energy to the exchanging uid. The proposed technique could be
implemented to correct possible optical aberrations thus boosting
the concentration up to its limit.
Acknowledgements
This research activity has been nancially supported by a MIUR
PhD Grant and a Fondazione CARISBO grant within the research
framework on renewable sources, energy spare and distributed
micro-generation.
References
[1] Green MA, Emery K, Hishikawa Y, Warta W, Dunlop ED. Solar cell efciency
tables (version 42). Prog Photovolt: Res Appl 2013;21:82737. http://
dx.doi.org/10.1002/pip.2404.
[2] Wiesenfarth M, Helmers H, Philipps SP, Steiner M, Bett AW. Advanced concepts
in concentrating photovoltaics (CPV). In: Proc. of 27th European photovoltaic
solar energy conference and exhibition, Frankfurt (Germany); 2012. p. 115.
http://dx.doi.org/10.4229/27thEUPVSection2012-1AP.1.4.
[3] Bentez P, Miano J, Zamora P, Mohedano R, et al. High performance Fresnelbased photovoltaic concentrator. Opt Express 2010;18:A2540. http://
dx.doi.org/10.1364/OE.18.000A25.
[4] Singh PL, Sarviya RM, Bhagoria JL. Thermal performance of linear Fresnel
reecting solar concentrator with trapezoidal cavity absorbers. Appl Energy
2010;87(2):54150. http://dx.doi.org/10.1016/j.apenergy.2009.08.019.
[5] Ryu K, Rhee JG, Park KM, Kim J. Concept and design of modular Fresnel lenses
for concentration solar PV system. Solar Energy 2006;80(2):15807. http://
dx.doi.org/10.1016/j.solener.2005.12.006.
[6] Leutz R, Suzuki A, Akisawa A, Kashiwagi T. Design of a nonimaging Fresnel lens
for solar concentrators. Solar Energy 1999;65(6):37987. http://dx.doi.org/
10.1016/S0038-092X(98)00143-1.
[7] Feuermann D, Gordon J. High-concentration photovoltaic designs based on
miniature parabolic dishes. Solar Energy 2001;70(5):42330. http://dx.doi.org/
10.1016/S0038-092X(00)00155-9.
[8] Kribus A, Kaftori D, Mittelman G, Hirshfeld A, et al. A miniature concentrating
photovoltaic and thermal system. Energy Convers Manage 2006;47(20):358290.
http://dx.doi.org/10.1016/j.enconman.2006.01.013.

221

[9] Gordon JM, Katz EA, Eugene A, Feuermann D, Mahmoud H. Toward ultrahighux photovoltaic concentration. Appl Phys Lett 2004;84:36424. http://
dx.doi.org/10.1063/1.1723690.
[10] http://www.apollon-eu.org.
[11] Kinsey GS, Sherif RA, Cotal HL, Pien P, et al. Multijunction solar cells for densearray concentrators. In: Proc. of the IEEE 4th world conf. on photovoltaic
energy conversion, 2006. p. 6257. http://dx.doi.org/10.1109/WCPEC.2006.
279532.
[12] Verlinden P, Lewandowski A, Bingham C, Kinsey G, Sherif R, Lasich J.
Performance and reliability of multijunction IIIV modules for concentrator
dish and central receiver applications. In: Proc. of the IEEE 4th world conf. on
photovoltaic energy conversion, 2006. p. 5927. http://dx.doi.org/10.1109/
WCPEC.2006.279526.
[13] Chayet H, Kost O, Moran R, Lozovsky I. Efcient, low cost dish concentrator for
a CPV based cogeneration system. In: AIP conference proceedings, vol. 1407;
2011. p. 24952. http://dx.doi.org/10.1063/1.3658337.
[14] Lasich JB, Verlinden PJ, Lewandowski A, Edwards D, et al. Worlds rst
demonstration of a 140 kWp heliostat concentrator PV (HCPV) system. In:
34th IEEE photovoltaic specialists conference (PVSC), 2009. p. 00227580.
http://dx.doi.org/10.1109/PVSC.2009.5411354.
[15] Chong KK, Siaw FL. Electrical characterization of dense-array concentrator
photovoltaic system. In: 27th European photovoltaic solar energy conference,
Frankfurt (Germany), 2012. http://dx.doi.org/10.4229/27thEUPVSection20121AV.3.18.
[16] Minuto A, Timo G, Groppelli P, Sturm M. Concentrating photovoltaic
multijunction (CPVM) module electrical layout optimisation by a new
theoretical and experimental mismatch analysis including series resistance
effect. In: 35th IEEE photovoltaic specialists conference (PVSC), 2010. p.
0030816. http://dx.doi.org/10.1109/PVSC.2010.5614540.
[17] Cooper T, Pravettoni M, Cadruvi M, Ambrosetti G, Steinfeld A. The effect of
irradiance mismatch on a semi-dense array of triple-junction concentrator
cells. Solar Energy Mater Solar Cells 2013;116:23851. http://dx.doi.org/
10.1016/j.solmat.2013.04.027.
[18] Luque A, Sala G, Arboiro JC. Electric and thermal model for non-uniformly
illuminated concentration cells. Solar Energy Mater Solar Cells 1998;51(3
4):26990. http://dx.doi.org/10.1016/S0927-0248(97)00228-6.
[19] Franklin E, Coventry J. Effects of highly non-uniform illumination distribution
on electrical performance of solar cells. In: Proc. solar Australian and New
Zeeland solar energy society, 2003.
[20] Katz EA, Gordon JM, Feuermann D. Effects of ultra-high ux and intensity
distribution in multi-junction solar cells. Prog Photovolt: Res Appl
2006;14(4):297303. http://dx.doi.org/10.1002/pip.670.
[21] Herrero R, Victoria M, Domnguez C, Askins S, Antn I, Sala G. Concentration
photovoltaic optical system irradiance distribution measurements and its
effect on multi-junction solar cells. Prog Photovolt: Res Appl
2012;20(4):42330. http://dx.doi.org/10.1002/pip.1145.
[22] Hernndez M, Cvetkovic A, Bentez P, Miano JC. High-performance Kohler
concentrators with uniform irradiance on solar cell. In: Proc. SPIE 7059,
nonimaging optics and efcient illumination systems V, 2008, no. 705908,
http://dx.doi.org/10.1117/12.794927.
[23] Fu L, Leutz R, Annenn HP. Evaluation and comparison of different designs and
materials for Fresnel lens-based solar concentrators. Proc. SPIE 8124,
nonimaging optics: efcient design for illumination and solar concentration
VIII 2011. http://dx.doi.org/10.1117/12.893390 [no. 81240E].
[24] Leutz R, Suzuki A, Akisawa A, Kashiwagi A. Flux uniformity and spectral
reproduction in solar concentrators using secondary optics. In: ISES solar
world congress, 2001, Adelaide.
[25] Baig H, Heasman KC, Mallick TK. Non-uniform illumination in concentrating
solar cells. Renew Sustain Energy Rev 2012;16(8):5890909. http://dx.doi.org/
10.1016/j.rser.2012.06.020.
[26] Salemi A, Eccher M, Miotello A, Brusa RS. Dense array connections for
photovoltaic systems. Prog Photovolt: Res Appl 2011;19(4):37990. http://
dx.doi.org/10.1002/pip.1040.
[27] Loeckenhoff R, Kubera T, Rasch KD. Water cooled TJ dense array modules for
parabolic dishes. In: AIP conference proceedings, vol. 1277; 2010. p. 436.
http://dx.doi.org/10.1063/1.3509229.
[28] Vivar M, Antn I, Sala G. Radial CPV receiver. Prog Photovolt: Res Appl
2010;18(5):35362. http://dx.doi.org/10.1002/pip.921.
[29] Chong KK, Wong CW, Siaw FL, Yew TK. Optical characterization of nonimaging
planar concentrator for the application in concentrator photovoltaic system. J
Sol Energy Eng 2010;132(1). http://dx.doi.org/10.1115/1.4000355 [no.
011011].
[30] Siaw FL, Chong KK, Wong CW. A comprehensive study of dense-array
concentrator photovoltaic system using non-imaging planar concentrator.
Renew Energy 2014;62:54255. http://dx.doi.org/10.1016/j.renene.2013.08.014.
[31] Riveros-Rosas D, Snchez-Gonzlez M, Arancibia-Bulnes CA, Estrada CA.
Inuence of the size of facets on point focus solar concentrators. Renew
Energy 2011;36(3):96670. http://dx.doi.org/10.1016/j.renene.2010.08.038.
[32] Tan MH, Chong KK, Wong CW. Optical characterization of nonimaging dish
concentrator for the application of dense-array concentrator photovoltaic
system.
Appl
Opt
2014;53(3):47586.
http://dx.doi.org/10.1364/
AO.53.000475.
[33] Bianchi M, Diolaiti E, Giannuzzi A, Marano B, Melino F. Energetic and economic
analysis of a new concept of solar concentrator for residential application. In:
Energy procedia, Proc. of the 7th international conference on applied energy,
Abu Dhabi (UAE): ICAE; 2015.

222

A. Giannuzzi et al. / Applied Energy 145 (2015) 211222

[34] Wilson RN, Franza F, Noethe L. Active optics I: A system for optimizing the
optical quality and reducing the costs of large telescopes. J Modern Opt
1987;34(4):485509.
[35] Noethe L, Franza F, Giordano P, Wilson RN, et al. Active optics II. Results of an
experiment with a thin 1 m test mirror. J Modern Opt 1988;35(9):142757.
http://dx.doi.org/10.1080/09500348814551591.
[36] Biasi R, Gallieni D, Salinari P, Riccardi A, Mantegazza P. Contactless thin
adaptive mirror technology: past, present, and future. In: Proc. SPIE adaptive
optics systems II, 2010. p. 7736. http://dx.doi.org/10.1117/12.858816.
[37] Angel R, Connors T, Davison W, Olbert B, Sivanandam S. New architecture for
utility-scale electricity from concentrator photovoltaics. In: Proc. of SPIE the
international society for optical engineering, 2010. p. 7769.
[38] Steeves J, Pellegrino S. Ultra-Thin Highly Deformable Composite Mirrors. In:
54th AIAA/ASME/ASCE/AHS/ASC structures, structural dynamics, and
materials conference, 2013.
[39] Ning X, Pellegrino S. Design of lightweight structural components for direct
digital manufacturing. In: 53rd AIAA/ASME/ASCE/AHS/ASC structures,
structural dynamics and materials conference, 2012.

[40] Irschik H. A review on static and dynamic shape control of structures by

piezoelectric actuation. Eng Struct 2002;24(1):511. http://dx.doi.org/
10.1016/S0141-0296(01)00081-5.
[41] Noll R. Zernike polynomials and atmospheric turbulence. J Opt Soc Am
1976;66(3):20711. http://dx.doi.org/10.1364/JOSA.66.000207.
[42] Domnguez C, Antn I, Sala G. Multijunction solar cell model for translating IV
characteristics as a function of irradiance, spectrum, and cell temperature.
Prog Photovolt: Res Appl 2010;8(4):27284. http://dx.doi.org/10.1002/
pip.965.
[43] Green MA. Solar cell ll factors: general graph and empirical expressions.
Solid-State Electron 1981;24(8):7889. http://dx.doi.org/10.1016/00381101(81)90062-9.
[44] http://www.azurspace.com.
[45] Fontani D, Sansoni P, Francini F, Jafrancesco D, et al. Pointing sensors and sun
tracking techniques. Int J Photoenergy 2011. http://dx.doi.org/10.1155/2011/
806518 [Article ID: 806518].
[46] Shack RV, Platt BC. Production and use of a lenticular Hartmann screen. J Opt
Soc Am 1971;61(5):65660.