A Global Optimization Approach to the Design of Solar

Power Plants
E. Carrizosaa,b , C. Domı́nguez-Bravob,∗, E. Fernández-Carab,c , M. Querod
a
Departamento de Estadı́stica e Investigación Operativa, Universidad de Sevilla, Spain
b
Instituto de Matemáticas de la Universidad de Sevilla, Universidad de Sevilla, Spain
c
Departamento de Ecuaciones Diferenciales y Análisis Numérico, Universidad de Sevilla,
Spain
d
Abengoa Solar New Technologies, Sevilla, Spain

Abstract

A method for optimizing a Solar Power Tower system is proposed, in which

both the location of the heliostats and the characteristics of the solar tower
are simultaneously considered. Maximizing the efficiency of the plant, i.e.,
optimizing the energy generated per unit cost, leads to a difficult high di-
mensional global optimization problem with an objective function hard to
compute and nonconvex constraints as well. An alternating greedy-based
heuristic is suggested. Our method is compared against other state-of-the-
art proposals in the literature of solar power plants optimization.
Keywords: solar thermal power, heliostat field layout, thermo-economic
optimization, global optimization, greedy algorithm

∗
Corresponding author. Tel. +34954556336.
Email address: carmenanadb@us.es (C. Domı́nguez-Bravo)

Preprint submitted to Optimization Online May 8, 2014

1. Introduction

Solar Power Tower (SPT) systems are known as one of the most promis-
ing technologies for producing solar electricity due to the high temperatures
reached, resulting in high thermodynamic performances; some reviews on
solar thermal electricity technology are Mills (2004), Romero et al. (2002)
and Stine & Harrigan (2001). A SPT system is here considered to consist of
two elements: a tower and a field of (hundreds or thousands) of heliostats.
Direct solar radiation is reflected and concentrated by the heliostat field onto
a receiver placed at the top of the tower. In the receiver, this thermal energy
at a high temperature is then transferred to the heat transfer fluid to produce
electricity through a conventional thermodynamic cycle.
The heliostat field is a group of mirrors having two-axis movement to
reflect the direct light from the sun to the target point on the receiver aper-
ture. The heliostat locations take into account the typical solar radiation at
the site. In the North hemisphere, the amount of solar energy per heliostat
is larger in the North area to the tower, because the so-called cosine factor
losses are reduced.
The optimal design of a SPT system consists of determining the tower
height, the shape and dimensions of the receiver aperture in the tower (tower
optimization) and the location of the heliostats (field optimization) so as to
optimize the annual energy produced and the plant cost. From the mathe-
matical point of view, we want to simultaneously optimize several economical
and environmental criteria (profits, total investment costs, internal rate of
return and global emissions). These objectives are in conflict to each other,
see Spelling et al. (2012), and they are usually aggregated into a single cri-

2
terion of efficiency, namely, the energy per unit cost, see Ramos & Ramos
(2012) and Spelling et al. (2012).
Two challenging issues are the dimensionality of the field optimization
problem, with (a priori unknown) hundreds or thousands of variables and
nonconvex constraints related to the location of heliostats, and the evalua-
tion of the objective function. This evaluation is implicitly defined by the
subroutine, and due to the nature of the process, is not smooth, may have
many local optima and has no apparent mathematical structure which can
be exploited to help to choose an appropriate optimization algorithm.
In SPT systems, the design of a heliostat field layout is of exceptional
importance. Very frequently, a geometrical pattern is imposed, i.e., the he-
liostats locations are assumed to follow a fixed distribution. It is commonly
assumed that heliostats follow a radial-stagger distribution, originally pro-
posed in Lipps (1981), Lipps & Vant-Hull (1978) and also used in Collado
(2009), Siala & Elayeb (2001), among others. Following a spiral distribu-
tion is also used, see Noone et al. (2012). These distributions are usually
optimized using different parametric approaches. Although these geometric
approaches strongly simplify the layout optimization problem, they may not
be accurate enough under, for instance, time asymmetric weather conditions
or terrain constraints as pointed out in Romero et al. (2002). In this paper,
heliostats location will not be forced to follow a specific geometrical pattern,
and, instead, will be obtained as the result of a pattern-free optimization
strategy.
The number of heliostats to be located is usually not given in advance,
but found in the optimization process: an oversized field (i.e., a field with a

3
sufficiently large number of heliostats), following some geometrical pattern,
is built, and then those heliostats generating less energy are sequentially
removed while guaranteeing that a given power output is attained. This
way, although the optimal parameters for the oversized field were obtained,
there is a high risk that a strong distortion exists between the original and
final field. With the procedure presented in this paper, an initial oversized
field is not needed.
This paper presents a new procedure for the layout optimization of a
SPT plant, including the optimization of both the tower and the receiver,
and the heliostats location. Most papers in the literature focus on the opti-
mization of the field layout, see Collado (2009), Sánchez & Romero (2006),
Siala & Elayeb (2001), or on the tower optimization separately, see Ghobeity
& Mitsos (2012). References to simultaneous optimization of the layout and
tower-receiver are very scarce. Pitz-Paal et al. (2011) and Ramos & Ramos
(2012) address the joint optimization, by using a metaheuristic (genetic algo-
rithm, simplex Nelder-Mead) improved by local searches (Powell algorithm),
always under the assumption of a geometric pattern (radial-stagger) for the
field.
The rest of the paper is organized as follows. In Section 2, we describe the
main ingredients affecting the performance of the SPT system. In Section 3,
our methodology to solve the problem is explained. In Section 4, we apply
the optimization algorithms and analysis tools to a typical plant design, and
finally, in Section 5, our main results are summarized and some perspectives
for further work are presented.

4
2. Problem statement

In this Section, the SPT system considered in the paper is described. We

explain the meaning of the variables used in the optimization process and
the constraints that have to be satisfied. The cost and energy functions, that
are the two criteria involved in the objective function, are also described.
Finally, the optimization problem that we want to solve is presented.

2.1. Decision variables

Two types of decision variables appear, some associated to the dimen-

sionalisation of the tower and the receiver aperture, and the remaining ones
associated to the heliostats locations.
We will assume that the receiver consists of a cylinder pointing to the
North, as can be seen in Figure 1 and is also explained in Behar et al. (2013),
Collado (2008), Stine & Harrigan (2001), among others. The front surface of
the receiver, also known as the aperture, is especially important because it is
here where strong radiative losses occur. In our case, we consider a circular
aperture. For simplicity, only the two most relevant variables associated to
the tower and the receiver design are considered, namely the aperture size,
given by its radius ra and its height h in the tower.
In what concerns the heliostat field, the heliostats locations, given by the
coordinates (x, y) of their centers, are the variables to be used. A heliostat
is characterized by its geometry and its optical properties. All heliostats are
assumed to be rectangular, to have the same dimensions and to be composed
of rectangular facets.
From now on we will denote by Θ the variables related to the receiver

5
 z z
Aperture Receiver

ra

West North

Figure 1: Receiver with circular Aperture

aperture dimension and by S the collection of coordinates of the centers of

the heliostats that define the heliostat field. Both variables are expressed as
follows, where Nhel denotes the total amount of heliostats:

S = (x1 , y 1 ), (x2 , y 2 ), . . . , (xNhel , y Nhel ) .


Θ = (ra , h) ,

Observe that Nhel is also a decision variable.

2.2. Constraints

Let Πt (Θ, S) denote the power (output) at time t for a SPT system with
parameters (Θ, S). Usually, when designing a SPT system, a fixed instant
of time is used to size the plant, as explained in Collado (2008), Sánchez &
Romero (2006) and Sanz-Bermejo et al. (2014). This time instant is known
in the literature as the design point, Td . At Td , a minimal power Π0 has to

6
be achieved, that is:

ΠTd (Θ, S) ≥ Π0 . (1)

Other constraints on the variables related to the receiver are determined

by the operation scheme of the plant, which is in turn influenced by technical
and legal regulations, leading to a compact set Θ as the feasible region for
Θ. There exist minimum and maximum values, rmin and rmax respectively,
for the aperture radius and a maximum value hmax for the tower height. The
feasible region Θ can be written as follows:

Θ ∈ Θ = {(ra , h) : rmin ≤ ra ≤ min(h, rmax ) ≤ hmax } . (2)

The heliostats must be located within a given region S0 . This defines a

constraint on the heliostat field:

S ⊆ S0 . (3)

The heliostats located in the field have to rotate freely avoiding collisions
with other heliostats. Consequently, we have to consider other constraints
forcing the heliostats not to overlap:

||(xi , y i ) − (xj , y j )|| ≥ δ for i 6= j , (4)

for a given parameter δ > 0, called the safety distance.

2.3. Functions
Two criteria are taken into account for the optimization of the SPT sys-
tem: the total investment cost and the generated annual energy.

7
 The investment cost function C takes into account the oulay in solar
power plant equipment (tower, receiver and heliostats). Hence, it depends
on the receiver variables Θ, that is, its radius ra and its height h, and the
number of heliostats |S| in the field, as we can see in (5):

C(Θ, |S|) = Ψ(|S|) + β1 (h + κ)σ + β2 πra2 , (5)

where |S| stands for the cardinality of S. We will assume that β1 and β2 are
empirical constants with values in (0, 1), κ and σ are positive and given by
appropriate physical considerations and Ψ is the linear heliostat cost function.
With this notation, the annual energy function E generated by the plant
takes the form:

Z T
E(Θ, S) = Π̃t (Θ, S) dt − γ1 , (6)
O

where the function Π̃t denotes the polinomial fitting of the power output
reached by the plant at each time instant t and γ1 is a constant that mesures
the fixed energy losses related to the hole system.
The power output values of the system are calculated by adding the values
reached by the heliostats located in the field, with a total of Nhel . The power
output function can be written as follows:

Nhel
X
Πt (Θ, S) = I(t)η(t)fref ϕ(t, xi , y i , S, Θ) − γ2 πra2 . (7)
i=1

Here I(t) is the so-called instantaneous direct solar radiation, η(t) is a mea-
sure of the radiation losses, fref is the heliostat reflectance factor, ϕ is the
product of the efficiency factors (usual in this framework), that is, ϕ =

8
fcos · fsb · fsp · fat and the constant γ2 is related with the energy losses asso-
ciated to the receiver size.
In particular, fcos = fcos (t, x, y, Θ) is the cosine efficiency, see Collado
& Turégano (1989); fsb = fsb (t, x, y, S, Θ) is the shadowing and blocking
efficiency; fsp = fsp (t, x, y, Θ) is the interception efficiency or spillage factor
and, finally, fat = fat (x, y) is the atmospheric efficiency, which takes into
account atmospheric losses between the heliostat and the receiver. We will
compute fsb following Sassi’s algorithm, see Sassi (1983); see also Collado &
Guallar (2012) and Stine & Harrigan (2001). On the other hand, fsp will be
computed according to the model described in Collado & Turégano (1986),
Kiera (1980b), Kiera (1980a). Finally, fat is estimated as in Biggs (1976),
Collado & Turégano (1989).
The annual energy of the plant is computed using the NSPOC procedure,
that is described in Crespo & Ramos (2009). We refer the reader to Biggs
(1976), Collado & Turégano (1986), Collado & Turégano (1989), Stine &
Harrigan (2001), for further details.

2.4. Optimization problem

As mentioned above, the two criteria involved are the total investment
cost and the annual energy produced. No common optimum can be found
for both criteria, so they are aggregated into one single objective, namely,
the maximization of generated energy per unit cost.
Written this way, the optimization problem we are addressing is the fol-
lowing:

9
 

 max F (Θ, S) = E(Θ, S)/C(Θ, |S|)

 Θ,S


 subject to Θ ∈ Θ



(P) ΠTd (Θ, S) ≥ Π0



S ⊆ S0






||(xi , y i ) − (xj , y j )|| ≥ δ for i 6= j


3. An alternating procedure to find the most efficient SPT system

In order to solve (P), an alternating procedure is suggested, in which one

sequentially optimizes the field layout for a given tower design and then, the
tower design is optimized for the previously obtained field layout. In other
words, we alternately solve (PΘ ) and (PS ), that are described below.
Thus, the optimization problem (P) has two blocks of decision variables,
namely those related to the design of the tower Θ, and the coordinates of
the centers of the heliostats S, related to the field layout.
The (PΘ ) subproblem, that is, the optimization of the field for a fixed
tower, can be written as follows:



 max F (Θ, S)


 S

 subject to Π (Θ, S) ≥ Π
Td 0
(PΘ ) Θ fixed



 S ⊆ S0

||(xi , y i ) − (xj , y j )|| ≥ δ

 for i 6= j

Observe that the cardinality of S is not fixed in advance in problems (P)

and (PΘ ), thus we cannot express them as standard optimization problems
in fixed dimension.

10
 The other subproblem (PS ) given below describes the optimization of the
tower and receiver when the field of heliostats S is fixed.



 max F (Θ, S)
 Θ

(PS ) S fixed subject to Θ ∈ Θ



 ΠTd (Θ, S) ≥ Π0
The alternating algorithm used to solve the optimization problem (P) is
described in Algorithm 1. As we have said, this algorithm alternatively solve
the tower and the heliostat field optimization problem. We consider that
the algorithm has performed a complete iteration when a tower problem and
a field problem have been solved. Each time an optimization subproblem
is solved, that is (PS ) or (PΘ ), the highest value obtained for the objective
function and the system design associated to this value are stored in the
variables Υobjective and Υdesign respectively.
When a complete iteration is achieved, the SPT system annual efficiency
is checked. If no better annual performance of the system is found the al-
gorithm stops and returns the highest value obtained. It is also possible to
consider a limited number of iterations for the algorithm so it stops when
the maximum number of iterations, kmax , is reached and returns the highest
value attained so far.

11
Algorithm 1 Alternating algorithm
Require: Θ0 feasible and random
k←0 STEP k = 0
S0 ← solve (PΘ ) given Θ = Θ0
Υdesign ← (Θ0 , S0 )
Υobjective ← F (Θ0 , S0 )
repeat
k ←k+1 STEP k > 0
Θk ← solve (PS ) given S = Sk−1
if Υobjective < F (Θk , Sk−1 ) then
Υdesign ← (Θk , Sk−1 )
Υobjective ← F (Θk , Sk−1 )
end if
Sk ← solve (PΘ ) given Θ = Θk
if Υobjective < F (Θk , Sk ) then
Υdesign ← (Θk , Sk )
Υobjective ← F (Θk , Sk )
end if

k , Sk ) < F (Θk−1 , Sk−1 ) or k = kmax

until F (Θ
objective ←
 Υ best objective function.
return
 Υ
design ← best plant design.

At the starting step, the initial value Θ0 for the variable Θ is set ran-
domly in the feasible region Θ. Once this initial value is calculated, the first
problem, (PΘ ) with Θ = Θ0 is solved.
After the initial step is performed, both problems, (PS ) and (PΘ ) are

12
solved at each iteration in this same order. In what follows, we detail the
steps in the optimization of the field for a fixed tower, that is, the (PΘ )
subproblem, since the other subproblem (PS ) is directly solvable by standard
techniques.
We suggest to optimize Θ by a coordinate local search. No difficulties
are expected, since, in our experience, the function F (·, S) has a unimodal
shape, show in Figure 2 using the example field layout P S10 given in Noone
et al. (2012).

F(⋅, S)

280 0.00 0.00 29.54 27.44 24.65 21.79 19.06 16.55 14.31

252.1 0.00 0.00 33.42 30.66 27.24 23.81 20.61 17.72 15.19

224.2 0.00 0.00 37.58 34.05 29.90 25.84 22.12 18.85 16.03

196.3 0.00 0.00 41.90 37.50 32.53 27.80 23.55 19.88 16.77

168.4 0.00 0.00 46.17 40.84 35.01 29.59 24.82 20.78 17.40
h ∈ [hmin,hmax]

140.5 0.00 0.00 50.14 43.85 37.18 31.10 25.86 21.48 17.88

112.6 0.00 0.00 53.49 46.32 38.90 32.24 26.61 21.96 18.18

84.7 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00

56.8 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00

28.9 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00

1 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00

1 3.375 5.75 8.125 10.5 12.875 15.25 17.625 20

ra ∈ [rmin,rmax]

Figure 2: F (·, S) values fixed S = P S10

13
3.1. Field optimization
We are going to describe in this subsection an algorithm for optimizing
the field when the receiver variables Θ are fixed.
The number of variables (heliostats centers) is not fixed. Even fixing the
number of heliostats, which may be greater than 1, 000 in commercial plants,
the huge number of variables and the high computing time needed to evaluate
the energy function make this problem difficult to solve.
The process we are going to implement with the goal to find an optimal
heliostat field is different from others in the literature in three aspects:

• There is no need of a starting configuration, i.e, no initial field is needed.

• No parametric form is used as a field pattern (e.g. we do not assume

that the field has to possess a radially staggered shape).

• The number of heliostats is selected according to the power output

requirements, avoiding oversizing.

The procedure proposed here resembles the algorithm described in Sánchez

& Romero (2006). This method considers a greedy strategy using a rectan-
gular gridding when the tower height and receiver dimensions are initialized
given a value. In such paper, a grid with 1 × 1 m is used to determine the
heliostats positions in the field. In a first step, the annual energy at every
point in the field is calculated. The first heliostat is located at the maxi-
mum energy grid position. Once the first heliostat is set, annual normalized
shadowing and blocking factors are added. Not with a precise calculation, a
grid with only some representative 16 neighboring positions is used, and then
an interpolation process is performed. The shadowing and blocking factor

14
is calculated without considering overlapping. To prevent future heliostats
from being located in the optical path of those previously placed, a mirror
image of the blocking effect is added. Finally, to compress the heliostat field,
a penalizing weight factor is introduced to favor and stimulate the location
of heliostats closer to the tower.
The procedure used in Buck (2014) produces a heliostat field layout simi-
lar to the final layout presented in this paper. An initial field configuration
is needed as a first step in the proposed algorithm. Then, a field layout re-
finement is done without imposing a radial staggered pattern. The number
of heliostats is fixed in the process. The initial field is slightly modified con-
sidering the geometric constraints to avoid collisions between heliostats. The
heliostats are repositioned one by one for the whole field. This simulation
cycle is repeated many times. Sometimes, local minima are found, and the
heliostat are repositioned manually to continue with the process. The algo-
rithm stops when no annual energy improvement is obtained after a complete
cycle.
Another heliostat field layout generation procedure is used in Wei et al.
(2010), adding a heliostat boundary constrained by the receiver geometrical
aperture and using a radial staggered pattern.
The algorithm presented in this paper follows a greedy procedure: locates
the heliostats one by one at the best feasible position in the field, that is,
the location where the annual energy of the whole field is highest for a given
tower height and receiver size. The process is repeated until no annual en-
ergy improvement is reached. The heliostats are located freely, without any
pre-arranged distribution. Only two geometrical constraints, already men-

15
tioned, have to be taken into account to locate the heliostats: the field shape
constraint (3) and the heliostat center constraints to avoid collisions, see (4).
The annual energy is modified at each step as well as the shading and block-
ing effects that the new heliostat is causing in the field. This is the main
reason of increase of the computing time: Once a new heliostat is located
and the shading and blocking effects are incorporated, the process must be
repeated.
Obviously, the first problem, (PΘ0 ), involves locating the first heliostat
center when only the field shape constraint is considered:

 max
 E(Θ, {(x, y)})
PΘ0

(x,y)
Θ fixed
 subject to (x, y) ∈ S0


This problem has an easy-to-handle objective function, as plotted in Fig-

ure 3(a). In return, when we have already located at least one heliostat in the

field and we have obtained a field S k−1 = (x1 , y 1 ), . . . , (xk−1 , y k−1 ) with


k − 1 heliostats that respects (3) and (4), the problem PΘk described below
is difficult to solve, since nonconvex constraints are involved and the energy
function has a complex behavior due to the shadowing and blocking effects,
see Figure 3(b).

16
 (a) Step k = 0 (b) Step k = 20

Figure 3: Annual Energy

We are going to describe now the k-step problem, PΘk , in the previous

algorithm. Let us introduce the notation S k = S k−1 , (x, y) , where (x, y)


denotes the variable with respect we maximize in problem PΘk :


 k−1


 max E(Θ, S , (x, y) )

 (x,y)

 
subject to S k−1 , (x, y) ∈ S 0

PΘk


Θ fixed
||(xi , y i ) − (x, y)|| ≥ δ0






for (xi , y i ) ∈ S k−1 1 ≤ i ≤ k − 1



As noticed the k-step the problem is equivalent to maximizing the energy

generated by the new heliostat location, because the cost function is fixed at
each step.
In order to solve this problem, since the functions involved are highly
multimodal, and the output strongly depends on the starting points, the
greedy procedure is embedded in a global algorithm consisting on a multistart
process, and is repeated Nsem times. The scheme is described in Algorithm 2.
The initial data required are the fixed variables related with the tower size,

17
Θ, the number of times that the algorithm will be repeated, Nsem , and the
∗
number of initial solution used in the multistart procedure, Nini .
Algorithm 2 has two different steps depending on the number of heliostats
already located in the field. At the first step, when no heliostat is located
in the field, there are not shadowing and blocking effects involving multi-
modality in the objective function. That is why no multistart strategy is
required, and the number of initial solutions is set to one. Once the first
heliostat is located in the field, the multimodality of the problem appears
and a multistart strategy with Nini different feasible initial solution is used.


The heliostats are located solving PΘk using Algorithm 3, with the cor-
responding Nini value. We are going to explain this Greedy Algorithm more
in detail in next paragraphs.
In the second step of Algorithm 2 two phases can be also differentiate,
Phase A and Phase B. The first phase, Phase A, consists of locating heliostats
until the power output requirement, Π0 is reached. After this constraint is
achieved, the location of heliostats continues provided that the system annual
efficiency increases, this phase is called Phase B.
It may happen that there exist some manufacturer requirements on the
power output reached by the system in order to prevent damages to the
receiver, see Buck et al. (2006), for this reason Phase B could stop when a
given upper limit, Π1 , of the power output of the system is reached. The
heliostats field design, when the upper limit Π1 is reached, is stored in the
variable Υdesign−B1 .
However, this field may be improved, according to its efficiency, forgetting
this power output upper limit, the algorithm continuous locating heliostat

18
until the system annual efficiency does not increase. In this case, the solu-
tion are the highest annual efficiency value attained Υobjective and its tower
and field configuration, Υdesign . We consider a maximum time for this last
heliostats location phase, denoted by tlimit , to avoid large processes.

19
Algorithm 2 Algorithm to design the heliostat field layout
∗
Require: Θ, Nsem , Nini
for sem = 1 to Nsem do
test stop ← true STEP k = 0
k←0
Nini ← 1
S 0 ← solve (PΘ0 ) with Algorithm 3
while ΠTd (Θ, S k ) < Π0 do
k ←k+1 STEP k > 0 — PHASE A
∗
Nini ← Nini


S k ← solve PΘk with Algorithm 3
end while
repeat
Υ0objective ← F (Θ, Sk ) PHASE B
k ←k+1


S k ← solve PΘk with Algorithm 3
Υ1objective ← F (Θ, Sk )
if ΠTd (Θ, S k ) > Π1 & test stop then
test stop ← false
Υdesign−B1 ← (Θ, Sk )
end if
until Υ1objective < Υ0objective or tlimit reached
end for 
0
objective ← Υobjective
 Υ
return
 Υ
design ← (Θ, Sk−1 )

20


Now, we focus on the resolution of the problem PΘk trough the greedy
strategy explained in Algorithm 3. It is well known that the annual energy
function is hard to compute, see Sánchez & Romero (2006); that is why we
approximate it by a much simpler function. Thus, instead of computing E
as in (6), the power output (7) at the design point Td is used; this is ΠTd .
More accurate approximations, as those suggested in Spelling et al. (2012)
and Zhang et al. (2007) based on calculating the power output at several
time instants could be used, at the expense of increasing the already high
computational cost.
Using ΠTd as the objective function usually leads to a heliostat field so-
lution more compact than using the annual energy function. This is caused
by the reduction on the shadowing and blocking effects at this time instant.
In order to compensate for this, the safety distance value is modified. The
safety distance δ0 is calculated as the product of the initial safety distance δ
and a new parameter Fsep that can be in the range [1, 2]. This new parameter
Fsep controls the separation between heliostats. It depends on the selected
design point, the heliostat size and the feasible region among others, that is
why it has to be set for each problem.
As we have already said, a multistart procedure is used in Algorithm 3 to
avoid local minima. The algorithm begins with Nini different random feasible
points. The final solution is selected taking into account the power output
at the design point given by each new configuration. The heliostat position
that reaches the best objective value will become part of the heliostat field
solution.

21
Algorithm 3 Greedy algorithm with multistart procedure

Require: Θ and S k−1 = (x1 , y 1 ), . . . , (xk−1 , y k−1 )
for r = 1 to Nini do
(xr , yr ): random feasible initial solution
(x∗r , yr∗ ): solve Pθk

Srk = S k−1 , (x∗r , yr∗ ) : optimal feasible solution



ΠTd (Θ, Srk ): objective function

end for
S k ← max ΠTd (Θ, Srk ) with r = 1, . . . , Nini

r
return S k

4. Results

The algorithm has been implemented in Matlab

c
, using fmincon and
fminbnd as local search routines to solve the continuous optimization sub-
problems (PS ) and (PΘ ) respectively.
The accuracy and feasibility of the code have been confirmed by compar-
ing to the heliostat fields obtained with benchmark procedures. In particular,
three fields are considered, with a common tower design. Following Noone
et al. (2012), such test fields are called here PS10, RPS10 (redefined PS10)
and Spiral. The values of the geographical, physical and geometrical pa-
rameters are given in Table 1.

22
Parameter Default value Reference
Location and Time

Emplacement Sanlúcar la Mayor (Seville) Osuna et al. (2004)

Latitude 37o 260 N Noone et al. (2012)
Longitude 6o 150 W Noone et al. (2012)
Design Point Td March 21 Day 12 Hour assumed
Design direct normal irradiation DNI 823.9 W/m2 assumed
DNI model cloudless skies assumed

Tower and Receiver

Tower optical height h 100.50 m assumed

Aperture radius ra 6.39 m assumed
Aperture slope ξ 12.5 Noone et al. (2012)
Minimum radius of the field 50 m assumed
Receiver Technology Saturated Steam Osuna et al. (2004)
Nominal Power 11.02MWe Osuna et al. (2004)

Heliostats and Field

Heliostats Sanlcar 120 Noone et al. (2012)

Heliostat width 12.84 m Noone et al. (2012)
Heliostat height 9.45 m Noone et al. (2012)
Heliostat optical height z0 5, 17 m Osuna et al. (2004)
Heliostat minimal safety distance δ heliostat diagonal Buck (2012) and Buck et al. (2012)
σoptical 2.9 mrad Noone et al. (2012)
Total number of heliostats 624 Osuna et al. (2004)
o
Slope of the field 0 assumed
Field shape semicircle assumed
2
Maximum size 1566869.33 m assumed

Table 1: Parameter Values

23
4.1. Heliostat field configurations comparison

PS10 as given in Noone et al. (2012), is a SPT system similar to a

solar commercial plant located in Sanlúcar la Mayor, Spain, and it is one
of the most popular test instances. RPS10 is a variant suggested in Noone
et al. (2012), in which a local search is performed, taking the PS10 layout
as starting point, optimizing the parameters configuration of the system.
Finally, Spiral, as introduced in Noone et al. (2012), is another field layout
obtained when a spiral pattern is given for the heliostats configuration and
the parameters of the spiral are tuned.
In Figure 4, the previous heliostat field layouts and also the GPS10
configurations, computed using our greedy algorithm, are shown with the
same number of heliostats. This GPS10 configuration have been obtained
after performing different experiments varying Fsep with Nsem = 5. The best
configuration that we show is obtained with Fsep = 1, 5.

24
 Heliostat Field Layout Nhel= 624 Heliostat Field Layout Nhel= 624
800 800

700 700

600 600
(South) x coordinate (North)

(South) x coordinate (North)

500 500

400 400

300 300

200 200

100 100

0 0
−400 −300 −200 −100 0 100 200 300 400 −400 −300 −200 −100 0 100 200 300 400
(West) y coordinate (East) (West) y coordinate (East)

(a) PS10 (b) RPS10

Heliostat Field Layout Nhel= 624 Heliostat Field Layout Nhel= 624
800 800

700 700

600 600
(South) x coordinate (North)

(South) x coordinate (North)

500 500

400 400

300 300

200 200

100 100

0 0
−400 −300 −200 −100 0 100 200 300 400 −400 −300 −200 −100 0 100 200 300 400
(West) y coordinate (East) (West) y coordinate (East)

(c) Spiral (d) GPS10

Figure 4: Heliostat Field Layouts

The power output at the design point Td , the annual energy output and
the system global efficiency computed for each heliostat field layout, are
given in Table 2. If we compare the results obtained fixing the same number
of heliostats (Phase A) we can see that our algorithm improves the results

25
at the design point, obviously because this is the optimization criterion in
practice. In view of the computational cost, this is the most adequate choice.
However, the total annual energy obtained with GPS10 is competitive and
can be sustained.
Moreover, it can be seen in the same table that our algorithm reaches
better values in Power output, Annual Energy Output and global efficiency
at the end of Phase B. The global optimization problem that we address in
this article is to design a SPT system with the aim to reach the best efficiency
value. These results show that our final heliostat field layout improves the
efficiency values previously obtained.

Field Nhel Power Output at Td (MWth) Annual Energy Output (GWHth) SPT Efficiency

PS10 624 42.5252 116.4960 50.3699

RPS10 624 42.8439 116.6736 50.4467
Spiral 624 41.8754 115.6748 50.0148
GPS10 Phase A 624 42.7643 116.1386 50.2154
GPS10 Phase B 943 62.2660 168.7543 54.0919

Table 2: Annual Energy, Power Output and STP Efficiency

The RPS10 and GPS10 fields are shown in Figure 5 with the heliostats
colored according to their contribution to the annual energy.

26
 Field Distribution: Annual Energy
900
[7.5568,7.6893)
[7.4243,7.5568)
800 [7.2917,7.4243)
[7.1592,7.2917)
[7.0267,7.1592)
700 [6.8941,7.0267)
[6.7616,6.8941)
[6.6291,6.7616)
(South) x coordinate (North)

600 [6.4966,6.6291)
[6.3641,6.4966)
[6.2315,6.3641)
500 [6.099,6.2315)
[5.9665,6.099)
[5.8339,5.9665)
400 [5.7014,5.8339)
[5.5689,5.7014)
[5.4364,5.5689)
300 [5.3039,5.4364)
[5.1713,5.3039)
[5.0388,5.1713)
200

100

0
−800 −600 −400 −200 0 200 400 600 800
(West) y coordinate (East)

(a) RPS10 with 624 Nhel

Field Distribution: Annual Energy
900
[7.5568,7.6893)
[7.4243,7.5568)
800 [7.2917,7.4243)
[7.1592,7.2917)
[7.0267,7.1592)
700 [6.8941,7.0267)
[6.7616,6.8941)
[6.6291,6.7616)
(South) x coordinate (North)

600 [6.4966,6.6291)
[6.3641,6.4966)
[6.2315,6.3641)
500 [6.099,6.2315)
[5.9665,6.099)
[5.8339,5.9665)
400 [5.7014,5.8339)
[5.5689,5.7014)
[5.4364,5.5689)
300 [5.3039,5.4364)
[5.1713,5.3039)
[5.0388,5.1713)
200

100

0
−800 −600 −400 −200 0 200 400 600 800
(West) y coordinate (East)

(b) GPS10 with 943 Nhel

Figure 5: Annual Energy performance per heliostat

27
 In order to compare the four fields layouts we have sorted the heliostats
of each field in terms of the amount of annual energy produced. Some results
are described below, displaying the annual energy and the cost in terms of
the amount of heliostats. The annual energy performance versus the number
of heliostats located in the field is shown in Figure 6(a); the cost of the SPT
system versus the number of heliostats located in the field is depicted in
Figure 6(b) and in Figure 6(c), the objective function, that is, the annual
energy per unit cost versus the number of heliostat located in the field is
plotted. In these three figures the four fields have the same behavior, but,
when displaying the marginal annual energy added by each heliostat in the
field in Figure 6(d), some differences can be observed due to the distinct
heliostats locations.

28
 Annual Energy Heliostat Cost
6000 3.5

5000 3

4000 2.5
Annual Energy

Heliostat Cost
3000 2

2000 1.5

PS10−624 PS10−624
1000 1
RPS10−624 RPS10−624
Spiral−624 Spiral−624
GPS10−943 GPS10−943
0 0.5
100 200 300 400 500 600 700 800 900 0 100 200 300 400 500 600 700 800 900 1000
Number of Heliostats Number of Heliostats

(a) Annual Energy per heliostat (b) Cost per heliostat

Annual Energy Individual
Annual Efficiency
8
2000

1800

7.5 PS10−624
1600 RPS10−624
Spiral−624
GPS10−943
1400
7
Annual Energy Individual

1200
Annual Efficiency

1000 PS10−624 6.5

RPS10−624
Spiral−624
800
GPS10−943
6
600

400
5.5

200

0 5
0 100 200 300 400 500 600 700 800 900 1000 0 100 200 300 400 500 600 700 800 900 1000
Number of Heliostats Number of Heliostats

(c) Efficiency per heliostat (d) Marginal Annual Energy

Figure 6: Field Layouts Analysis

The multimodality of the problem is rather strong. In Figure 7, the field

layouts obtained using the greedy algorithm without multistart procedure
and with multistart procedure are shown. When the greedy algorithm is
used without multistart procedure the heliostat field layouts obtained are
not so regular. However, the best results are not necessary matched with the
field regularity as shown in Table 3. The use of different values for Nini and
different seeds lead to different results. In this same table it can be seen that
the results obtained are strongly dependant on the parameter Fsep , and thus
this parameter must be carefully tuned.

29
 Heliostat Field Layout Nhel= 624 Heliostat Field Layout Nhel= 624
800 800

700 700

600 600
(South) x coordinate (North)

(South) x coordinate (North)

500 500

400 400

300 300

200 200

100 100

0 0
−400 −300 −200 −100 0 100 200 300 400 −400 −300 −200 −100 0 100 200 300 400
(West) y coordinate (East) (West) y coordinate (East)

(a) Nini = 1 and Fsep = 1.4 (b) Nini = 25 and Fsep = 1.4

Figure 7: Multistart Analysis

30
 Fsep Nini E(Θ, S) (GWHth) ΠTd (MWth)

1 1 111.8218 42.8547
1 5 113.2056 43.8620
1 25 112.5751 43.9750

1.3 1 115.3329 42.8518

1.3 5 115.9215 43.1986
1.3 25 115.5476 43.2284

1.4 1 115.4601 42.6238

1.4 5 115.9705 42.9515
1.4 25 115.8886 42.9672

1.5 1 115.3927 42.4381

1.5 5 116.1223 42.7607
1.5 25 115.9565 42.7550

1.6 1 115.2510 42.2751

1.6 5 115.6514 42.4423
1.6 25 115.7740 42.5365

1.7 1 114.1481 41.8045

1.7 5 115.2854 42.1711
1.7 25 115.3524 42.2298

1.8 1 113.1910 41.3676

1.8 5 114.7537 41.9031
1.8 25 115.0351 42.0264

1.9 1 112.2490 40.9322

1.9 5 114.2163 41.6361
1.9 25 114.4618 41.7183

2.0 1 111.3872 40.5499

2.0 5 113.6485 41.3305
2.0 25 113.9871 41.4547

Table 3: Results for each experimental field design Phase A

31
4.2. Alternating procedure
When the tower variables are also tuned, using our alternating approach
we obtain step-by-step the results shown in Table 4 and Figure 8. Using
random feasible values as initial data for the tower configuration and the
alternating procedure described in Section 3, a more efficient SPT configu-
ration is sought. To make a fair comparison, the power output to be reached
at Td is fixed at 42.5252 MWth, the same as the value obtained with the
PS10 configuration. The parameter Nini was set to 25, since this value gave
the best results in some preliminary tests performed.
From the theoretical point of view the algorithm would have to improve
at each step. However it may not be necessarily the case due to a premature
termination of the algorithm.

Step Problem Θk Figure |Sk | ΠTd (Θk , Sk ) E(Θk , Sk ) F (Θk , Sk ) Fsep

Calculate Random Θ0 (13.0148,30.1646) - 0 - - - -

k=0
S0 : Solve (PΘ0 ) (13.0148,30.1646) 8(a) 477 42.5769 117.3282 30.7187 2.0

Υobjective = 30.7187
Υdesign = (Θ0 , S0 )

Θ1 : Solve (PS0 ) (5.8186,84.2276) - 777 46.9466 122.7171 47.9114 2.0

k=1
S1 : Solve (PΘ1 ) (5.8186,84.2276) 8(b) 633 42.5412 115.5954 52.6136 1.6

Υobjective = 52.6136
Υdesign = (Θ1 , S1 )

Θ2 : Solve (PS1 ) (6.0509,87.0000) - 633 45.1234 116.3778 52.0542 1.6

k=2
S2 : Solve (PΘ2 ) (6.0509,87.0000) 8(c) 631 42.5401 115.6698 51.8551 1.6

Υobjective = 52.6136
Υdesign = (Θ1 , S1 ) with Θ1 = (5.8186, 84.2276)

Table 4: Results obtained using the alternating algorithm

From the results we conclude that the best solution (Θ1 , S1 ), found at the

32
second step of the Alternating Algorithm, reaches a efficiency value higher
than the efficiency values of the three reference SPT systems. Note that
the first heliostat field obtained in this example, see Figure 8(a), is not as
compact as the heliostat fields obtained in next iterations. This effect is due
to the low tower height value and the large receiver radius value obtained as
initial random solution. When the alternating process continues this effect
is corrected by the own algorithm.

33
 Heliostat Field Layout Nhel= 777 Heliostat Field Layout Nhel= 633
800 800

700 700

600 600
(South) x coordinate (North)

(South) x coordinate (North)

500 500

400 400

300 300

200 200

100 100

0 0
−400 −300 −200 −100 0 100 200 300 400 −400 −300 −200 −100 0 100 200 300 400
(West) y coordinate (East) (West) y coordinate (East)

(a) S0 (b) S1
Heliostat Field Layout Nhel= 631
800

700

600
(South) x coordinate (North)

500

400

300

200

100

0
−400 −300 −200 −100 0 100 200 300 400
(West) y coordinate (East)

(c) S2

Figure 8: Field Layouts obtained using Algorithm 1

4.3. Different feasible regions

The methodology presented in this paper does not make use of the shape
of the region. This is in contrast with used practice in the literature, in which
the heliostat field layout is determined by the radial-stagger distribution, as

34
it is also the case of the PS10 field. In real situations, the region selected
to build the SPT system can have some terrain constraints, associated with
other mechanisms, or terrain irregularities for instance. This means that
some unfeasible zones need to be included in the original feasible region. In
this case, a given pattern can not be followed in an easy way.
Using the greedy algorithm described in Algorithm 2, different feasible
regions can be considered and the heliostat location procedure remains e-
xactly the same, since with this algorithm the heliostat distribution is not
parameterized.
To illustrate the advantages of our pattern-free method against pattern-
based methods, three different possible feasible regions are considered and
both procedures, radial-stagger parameterization and greedy algorithm are
compared in these regions.
The three feasible regions that are considered are: a rectangular region
(R), a perforated region (P) and a valley region (V), where the SPT system
is supposed to be located near a river. They are represented graphically in
Figure 9.

35
 Feasible Region 1 Feasible Region 2
(South) x coordinate (North)

(South) x coordinate (North)

(West) y coordinate (East) (West) y coordinate (East)

(a) Rectangular region (R) (b) Perforated region (P)

Feasible Region 3
(South) x coordinate (North)

(West) y coordinate (East)

(c) Valley region (V)

Figure 9: Feasible Regions

As said before, we design the heliostat field for each feasible region con-
sidering the power output to be reached at Td as Π0 = 42.5252. The field
obtained are calculated using Algorithm 2 and for each feasible region the
different constraints associated are considered.
In order to compare these fields, we have limited the PS10 and RPS10
configurations to the three different regions and evaluated the results.
In Figures 10-12, the layouts resulted of using radial-stagger or pattern-
free configuration for each feasible regions considered are shown. The dif-
ferent phases of the greedy algorithm, Phase A and Phase B, are detailed

36
numerically and graphically. The numerical results are shown in Table 5.
From the results we conclude that for most tests performed the efficiency
value obtained is improved with our greedy based procedure, and in the
case that this value is not improved we propose a close solution. Another
advantage of our algorithm, that can be seen in the figures below, is that
the heliostat fields obtained with a similar number of heliostats (Phase A)
are more compact than those obtained with the radial-stagger distribution.
These fields have a small number of heliostats. However, in commercial plants
a larger number of heliostats is needed. For this reason the compactness of
the heliostat field solution is a profitable characteristic when more heliostats
need to be located.

37
 PS10 RPS10

Reg Field Nhel ΠTd Π1 E(Θ, S) F (Θ, S) Fsep Nhel ΠTd Π1 E(Θ, S) F (Θ, S) Fsep

Orig 420 29.2251 - 79.7184 44.3678 - 446 31.0710 - 84.4838 45.3597 -

Phase A 419 29.2251 30.6863 79.4732 44.2945 1.5 448 31.0712 32.6247 84.4442 45.2177 1.5
R
Phase B1 - - 30.6863 - - 1.5 - - 32.6247 - - 1.5
Phase B 425 29.6463 - 80.5383 44.5109 1.5 452 31.3252 - 85.1351 45.3398 1.5

Orig 611 41.6441 - 114.0742 50.0342 - 611 41.9508 - 114.2621 50.1166 -

Phase A 607 41.6818 43.7263 113.0502 49.8062 1.5 612 42.0029 44.1030 113.9230 49.9136 1.5

38
P
Phase B1 639 43.7296 43.7263 118.5926 50.4487 1.5 645 44.1129 44.1030 119.6093 50.5548 1.5
Phase B 745 50.4087 - 136.5689 52.1474 1.5 824 55.2863 - 149.7042 53.0627 1.5

Orig 565 38.2818 - 105.0733 48.5663 - 565 38.5535 - 105.1897 48.6201 -

Phase A 562 38.3327 40.2465 104.4212 48.4335 1.6 566 38.5885 40.5179 105.1091 48.5248 1.6
V
Phase B1 592 40.2488 40.2465 109.6089 49.1110 1.6 597 40.5670 40.5179 110.4690 49.2174 1.6
Phase B 856 56.4582 - 153.3586 52.8879 1.6 916 59.9543 - 162.7332 53.3294 1.6

Table 5: Energy per unit cost: F (Θ, S) = E(Θ, S)/C(Θ, |S|)

 Heliostat Field Layout Nhel= 420 Heliostat Field Layout Nhel= 419 Heliostat Field Layout Nhel= 425
800 800 800

700 700 700

600 600 600

500 500 500

400 400 400

300 300 300

(South) x coordinate (North)

(South) x coordinate (North)
(South) x coordinate (North)
200 200 200

100 100 100

0 0 0
−300 −200 −100 0 100 200 300 −300 −200 −100 0 100 200 300 −300 −200 −100 0 100 200 300

39
(West) y coordinate (East) (West) y coordinate (East) (West) y coordinate (East)

(a) PS10 (b) Greedy Phase A (c) Greedy Phase B

Heliostat Field Layout Nhel= 446 Heliostat Field Layout Nhel= 448 Heliostat Field Layout Nhel= 452
800 800 800

700 700 700

600 600 600

500 500 500

400 400 400

300 300 300

(South) x coordinate (North)

(South) x coordinate (North)
(South) x coordinate (North)

200 200 200

100 100 100

0 0 0
−300 −200 −100 0 100 200 300 −300 −200 −100 0 100 200 300 −300 −200 −100 0 100 200 300
(West) y coordinate (East) (West) y coordinate (East) (West) y coordinate (East)

(d) RPS10 (e) Greedy Phase A (f) Greedy Phase B

Figure 10: Rectangular region

 Heliostat Field Layout Nhel= 611 Heliostat Field Layout Nhel= 607 Heliostat Field Layout Nhel= 745
800 800 800

700 700 700

600 600 600

500 500 500

400 400 400

300 300 300

(South) x coordinate (North)

(South) x coordinate (North)
(South) x coordinate (North)
200 200 200

100 100 100

0 0 0
−400 −300 −200 −100 0 100 200 300 400 −400 −300 −200 −100 0 100 200 300 400 −400 −300 −200 −100 0 100 200 300 400
(West) y coordinate (East) (West) y coordinate (East) (West) y coordinate (East)

40
(a) PS10 (b) Greedy Phase A (c) Greedy Phase B
Heliostat Field Layout Nhel= 611 Heliostat Field Layout Nhel= 612 Heliostat Field Layout Nhel= 824
800 800 800

700 700 700

600 600 600

500 500 500

400 400 400

300
300 300

(South) x coordinate (North)

(South) x coordinate (North)
200
200 200
(South) x coordinate (North)

100
100 100

0
0 0 −400 −300 −200 −100 0 100 200 300 400
−400 −300 −200 −100 0 100 200 300 400 −400 −300 −200 −100 0 100 200 300 400
(West) y coordinate (East) (West) y coordinate (East) (West) y coordinate (East)

(d) RPS10 (e) Greedy Phase A (f) Greedy Phase B

Figure 11: Perforated Region

 Heliostat Field Layout Nhel= 565 Heliostat Field Layout Nhel= 562 Heliostat Field Layout Nhel= 856
800 800 800

700 700 700

600 600 600

500 500 500

400 400 400

300 300 300

(South) x coordinate (North)

(South) x coordinate (North)
(South) x coordinate (North)
200 200 200

100 100 100

0 0 0
−400 −300 −200 −100 0 100 200 300 400 −400 −300 −200 −100 0 100 200 300 400 −400 −300 −200 −100 0 100 200 300 400
(West) y coordinate (East) (West) y coordinate (East) (West) y coordinate (East)

41
(a) PS10 (b) Greedy Phase A (c) Greedy Phase B
Heliostat Field Layout Nhel= 565 Heliostat Field Layout Nhel= 566 Heliostat Field Layout Nhel= 916
800 800 800

700 700 700

600 600 600

500 500 500

400 400 400

300
300 300

(South) x coordinate (North)

(South) x coordinate (North)
200
200 200
(South) x coordinate (North)

100
100 100

0
0 0 −400 −300 −200 −100 0 100 200 300 400
−400 −300 −200 −100 0 100 200 300 400 −400 −300 −200 −100 0 100 200 300 400
(West) y coordinate (East) (West) y coordinate (East) (West) y coordinate (East)

(d) RPS10 (e) Greedy Phase A (f) Greedy Phase B

Figure 12: Valley region

5. Conclusions

A pattern-free method for optimizing a SPT system has been proposed,

in which both the location of the heliostats and the design of the solar tower
are simultaneously considered. Maximizing the efficiency of the plant, i.e.
optimizing the energy generated per unit cost, leads to a difficult high di-
mensional global optimization problem with an objective function hard to
compute and nonconvex constraints.
An alternating greedy-based heuristic, embedded in a multistart, has been
suggested. Our method provides competitive results against the standard re-
sults in the literature and is more versatile since it is not based on geometrical
patterns which may be valid only under certain physical conditions. More-
over, our strategy can be used to obtain an initial field for other algorithms
(such as the one in Buck et al. (2012)), and can be extended to address many
other situations:

• Ground irregularities in the field.

• The effect of tower shading.

• Variable (stochastic) meteorological data.

• Multiple Receiver and/or Multi-Tower plants.

• Multi-Size heliostat fields (heliostats with different sizes).

The two objectives of maximizing energy produced and minimizing cost

are aggregated in this paper into on single criterion, namely, efficiency max-
imization. A challenging problem would be to design procedures able to

42
approximate the efficient frontier for this biobjective problem. The recent
work Spelling et al. (2012) is a promising first approach.
Another interesting and different problem is the study of the energy stor-
age operations, as in Ghobeity & Mitsos (2012) and Sheu et al. (2012). These
situations will be analyzed in the future.

6. Acknowledgements

The authors thank Noone et al. (2012) for providing the heliostat field
configurations used to compare the results.
This research has been mainly supported by Abengoa Solar N.T. and In-
stitute of Mathematics of University of Seville (IMUS), through the research
contract “CapTorSol”. The authors would also like to acknowledge the sup-
port from the Governments of Spain (Grants MTM2010-15992, MTM2012-
36136), Andalucı́a (Grant P11-FQM-7603) and EU COST Action TD1207.

References

Behar, O., Khellaf, A., & Mohammedi, K. (2013). A review of studies on

central receiver solar thermal power plants. Renewable and Sustainable
Energy Reviews, 23 , 12–39.

Biggs, C., F. y Vittitoe (1976). The HELIOS model for the optical behavior
of reflecting solar concentrators. Technical Report SAND76-0347 Sandia
National Labs. URL: http://www.osti.gov/scitech/biblio/6273705.

Buck, R. (2012). Heliostat Field Layout using Non-Restricted Optimization.

In Proceedings of SolarPACES 2012 .

43
Buck, R. (2014). Heliostat field layout improvement by nonrestricted refine-
ment. Journal of Solar Energy Engineering, 136 , 1–6.

Buck, R., Barth, C., Eck, M., & Steinmann, W. (2006). Dual receiver concept
for solar tower. Solar Energy, 80 , 1249–1254.

Buck, R., Pfahl, A., & Roos, T. H. (2012). Target aligned heliostat field
layout for non-linear flat terrain. In Proceedings of the First Southern
African Solar Energy Conference (SASEC 2012).

Collado, F. J. (2008). Quick evaluation of the annual heliostat field efficiency.

Solar Energy, 82 , 379–384.

Collado, F. J. (2009). Preliminary design of surrounding heliostat fields.

Renewable Energy, 34 , 1359–1363.

Collado, F. J., & Guallar, J. (2012). Campo: Generation of regular heliostat

fields. Renewable Energy, 46 , 49–59.

Collado, F. J., & Turégano, J. (1986). An analytic function for the flux
density due to sunlight reflected from a heliostat. Solar Energy, 37 , 215–
234.

Collado, F. J., & Turégano, J. (1989). Calculation of the annual thermal

energy supplied by a defined heliostat field. Solar Energy, 42 , 149–165.

Crespo, L., & Ramos, F. (2009). NSPOC: A New Powerful Tool for Heliostat
Field Layout and Receiver Geometry Optimizations. In Proceedings of
SolarPACES 2009 .

44
Ghobeity, A., & Mitsos, A. (2012). Optimal Design and Operation of a Solar
Energy Receiver and Storage. Solar Energy Engineering, 134 , 9 pages.

Kiera, M. (1980a). Abbildung der sonne auf die apertur eines son-
nenkrftwerkes. Technical Report imb-BT-a20000-12 GAST project.

Kiera, M. (1980b). Feldoptimierung bei variationen de parameters. Technical

Report imb-BT-320000-05 GAST project.

Lipps, F. (1981). Theory of Cellwise Optimization for Solar Central Receiver

Systems. Technical Report SAND-85-8177 Houston Univ., TX (USA).
Energy Lab. URL: http://www.osti.gov/scitech/biblio/5734792.

Lipps, F. W., & Vant-Hull, L. L. (1978). A cellwise method for the optimiza-
tion of large central receiver systems. Solar Energy, 20 , 505–516.

Mills, D. R. (2004). Advances in solar thermal electricity technology. Solar

Energy, 76 , 19–31.

Noone, C. J., Torrilhon, M., & Mitsos, A. (2012). Heliostat field optimiza-
tion: A new computationally efficient model and biomimetic layout. Solar
Energy, 86 , 792–803.

Osuna, R., Fernández, V., Romero, S., Romero, M., & Sánchez, M. (2004).
PS10: a 11.0-MWe Solar Tower Power Plant with Satured Steam Receiver.
In C. Ramos, & J. Huacuz (Eds.), Proceedings of the 12th SolarPACES In-
ternational Symposium on Concentrated Solar Power and Chemical Energy
Technologies.

45
Pitz-Paal, R., Botero, N. B., & Steinfeld, A. (2011). Heliostat field layout
optimization for high-temperature solar thermochemical processing. Solar
Energy, 85 , 334–343.

Ramos, A., & Ramos, F. (2012). Strategies in Tower Solar Power Plant
optimization. Solar Energy, 86 , 2536–2548.

Romero, M., Buck, R., & Pacheco, J. E. (2002). An update on Solar Central
Receiver Systems, Projects and Technologies. Solar Energy Engineering,
124 , 98–109.

Sánchez, M., & Romero, M. (2006). Methodology for generation of heliostat

field layout in central receiver systems based on yearly normalized energy
surfaces. Solar Energy, 80 , 861–874.

Sanz-Bermejo, J., Gallardo-Natividad, V., Gonzalez-Aguilar, J., & Romero,

M. (2014). Comparative system performance analysis of direct steam gen-
eration central receiver solar thermal power plants in megawatt range.
Journal of Solar Energy Engineering, 136 , 1–9.

Sassi, G. (1983). Some notes on shadow and blockage effects. Solar Energy,
31 , 331–333.

Sheu, E. J., Mitsos, A. A., A. Eter, Mokheimer, E. M. A., Habib, M. A., & Al-
Qutub, A. (2012). A Review of Hybrid Solar-Fossil Fuel Power Generation
Systems and Performance Metrics. Journal of Solar Energy Engineering,
134 , 17 Pages.

Siala, F. M. F., & Elayeb, M. E. (2001). Mathematical formulation of a

46
 graphical method for a no blocking heliostat field layout. Renewable En-
ergy, 23 , 77–92.

Spelling, J., Favrat, D., Martin, A., & Augsburger, G. (2012). Thermoeco-
nomic optimization of a combined-cycle solar tower power plant. Energy,
41 , 113–120.

Stine, W. B., & Harrigan, R. W. (2001). Power From The Sun. John Wiley
and Sons. URL: http://www.powerfromthesun.net/book.htm.

Wei, X., Lu, Z., Wang, Z., Yu, W., Zhang, H., & Yao, Z. (2010). A new
method for the design of the heliostat field layout for solar tower power
plant. Renewable Energy, 35 , 1970–1975.

Zhang, H., Juchi, I., Favrat, D., & Pelet, X. (2007). Multi-objective ther-
moeconomic optimisation of the design of heliostat field of solar tower
power plants. In Proceedings of the Engineering for sustainable Energy in
developing countries.

47