Energy Conversion and Management 247 (2021) 114669

Contents lists available at ScienceDirect

Energy Conversion and Management

journal homepage: www.elsevier.com/locate/enconman

Thermoeconomic analysis and optimization of a hybrid solar-thermal

power plant using a genetic algorithm
Duarte Pereira de Araújo Coutinho a, Jorge Augusto Maciel Milhomem Rodrigues a, Aires José
Pinto dos Santos b, Viriato Sérgio de Almeida Semiao c, *
a
Mechanical Engineering Department, Instituto Superior Técnico, Universidade de Lisboa, Av. Rovisco Pais 1, 1049-001 Lisboa, Portugal
b
MARETEC/DEM – Marine, Environment and Technology Centre, LARSyS, Instituto Superior Técnico, Universidade de Lisboa, Av. Rovisco Pais, 1, 1049-001 Lisboa,
Portugal
c
LAETA, IDMEC, Mechanical Engineering Department, Instituto Superior Técnico, Universidade de Lisboa, Av. Rovisco Pais 1, 1049-001 Lisboa, Portugal

A R T I C L E I N F O A B S T R A C T

Keyword: The hybridization of solar-thermal and combustion technologies for power generation is an emerging concept
Thermoeconomics that brings the possibility to mitigate carbon dioxide emissions while maintaining a firm supply. This work
Concentrating solar power presents a thermoeconomic model for a distributed-scale (<100 kWe) hybrid solar-thermal power plant,
Multi-objective optimization
developed to study its performance under different operating conditions in a region with a yearly direct normal
Evolutionary algorithm
irradiance of around 1600 kWh/m2 yr. The proposed system consists of a combined Rankine-Brayton cycle with a
solar receiver and natural gas combustor working in series as heat sources to the topping cycle. A genetic al­
gorithm was employed to perform a multi-objective optimization of such system, and the result was a set of
Pareto-optimal designs, which were compared to a pre-defined reference design. Resulting optimized solutions
yield levelized electricity costs as low as 0.179 USD/kWh, as opposed to the 0.237 USD/kWh associated with the
reference design. Average 1st and 2nd law efficiencies of up to 27.97 % and 33.53 % were achieved, respectively,
which represent increases of up to 7.71 % and 7.31 %. Finally, average solar shares of up to 65 % are possible for
optimized designs versus the 58.4 % yielded by the base design.

1. Introduction being used in CSP power plants worldwide. Another great advantage is
the possibility to directly integrate CSP in standard operational power
Solar technologies for electricity generation can be currently divided cycles either by totally replacing the existent combustor/boiler by the
into two categories: solar photovoltaic (PV) and concentrating solar CSP unit or by combining both (hybrid solar-thermal and fossil fuel
power (CSP). Historically, the investment in solar PV systems has been power plants). In hybrid power plants, the environmental advantages of
far greater than that of CSP. As a result, PV technologies are more renewable solar energy couple with the reliability of fossil fuels to
mature and, therefore, the correspondent levelized electricity cost (LEC) deliver a quite versatile solution that meets the advantages of both ap­
is currently lower, with a global average of 0.068 USD/kWh versus proaches. Continuous generation of electricity during solar downtime is
0.182 USD/kWh for CSP in 2019 [1]. According to the International assured by the combustor, whose short start-up time assures flexibility of
Energy Agency (IEA), the global electricity generation of solar renew­ supply. In the short-term, hybridization of CSP with common fossil fuels
ables was 604 TWh in 2018, out of which only 12 TWh (roughly 2 %) presents an economic solution for carbon dioxide (CO2 ) emissions
account for CSP systems [2]. However, CSP technologies offer some mitigation while maintaining a continuous supply. In the long-term, the
clear advantages to PV and other renewable energy solutions. Since solar integration of these hybrids with CO2 capture technologies, such as pre/
energy is being collected in the form of heat, thermal energy storage post combustion sequestration, oxy-fuel combustion or chemical looping
(TES) systems are easily integrated in CSP power plants, which trans­ combustion, as well as the utilization of so-called green fuels such as
lates into a greater energy storage capacity with lower associated costs biodiesel, could eventually lead to carbon–neutral or even carbon-
[3]. González-Roubaud et al. [4] and Pelay et al. [5] presented thorough negative systems [3].
reviews and technical assessments of the TES systems that are currently Hybrid solutions for utility-scale power generation have been deeply

* Corresponding author.
E-mail address: [email protected] (V.S.A. Semiao).

https://doi.org/10.1016/j.enconman.2021.114669
Received 9 March 2021; Accepted 18 August 2021
Available online 9 September 2021
0196-8904/© 2021 Elsevier Ltd. All rights reserved.
D.P.A. Coutinho et al. Energy Conversion and Management 247 (2021) 114669

studied and discussed in the literature. In their works, Jin and Hong [6], distributed-scale generation. This system was idealized in a previous
Powell et al. [7], and Nathan et al. [3] presented detailed reviews of the master thesis (see Rodrigues [20]), and adapted from one of the models
most common configurations for hybrid power plants and proposed presented by Dunham and Lipiński [21]. The scope of this study is to
methodologies for their classification. Technical studies of operational evaluate the relevance of retrofitting CSP units to combined cycle power
hybrids have also been reported by several authors. Baghernejad and blocks for distributed-scale operation, as opposed to more common so­
Yaghoubi conducted exergoeconomic analyses and optimizations of two lutions that couple with ORC power blocks.
different systems located in Iran. One consists of a 400 MWe Integrated
Solar Combined Cycle (ISCC) [8,9], while the other is a standard 500 2. Modelling
kWe steam turbine cycle with solar-thermal hybridization [10]. The
authors used genetic algorithms to conduct single and multi-objective The proposed system consists of a Brayton-Rankine combined cycle
optimizations, resulting in considerable increases of exergetic effi­ with a solar receiver and a natural gas combustor working in series as
ciency and reductions of generation costs. Similarly, Pihl et al. [11] heat sources to the topping cycle. Considering the cost and maturity of
conducted a thermoeconomic optimization analysis in order to evaluate currently available concentrating solar thermal technologies, a para­
the feasibility of retrofitting a set of parabolic trough collectors (PTC) to bolic trough collector (PTC) was chosen as the most suitable for this
an existent 400 MWe combined cycle power plant, turning the system system, and the receiver operating parameters were set accordingly (see
into an ISCC. Results revealed that the integration of TES is not an Table 1).
attractive solution for the given design, and annual solar shares are The efficiency of the solar receiver is dependent on the operating
limited to 1.2 % (4 % nominal share). However, the minimum achieved temperature and incident solar flux, so it fluctuates with time. As a
costs of generated electricity are close to 0.10 €/kWh, as opposed to the reference, for an optimal operating temperature and a sunny Summer
0.17–0.19 €/kWh achieved by an equivalent stand-alone solar power day, the receiver efficiency of a PTC setup with these characteristics is
plant. Other authors have studied the feasibility of retrofitting CSP around 72%. As a simplifying assumption of the heat transfer process
systems to gas turbine cycles (see [12–15]). In his PhD thesis, Spelling between the solar receiver and the working fluid, it was considered that
[14] conducted a thorough thermoeconomic analysis on a variety of the temperature at the outlet (T4 ) is 50 K lower than the optimal receiver
different possible configurations for CSP retrofitting to a Brayton cycle. temperature for the given conditions. The proposed layout considers the
Obtained results favour the use of combined cycle configurations with addition of a regenerator after the compressor to recover part of the heat
integrated TES. The incorporation of TES allowed for a higher annual discharged by the gas turbine (see Fig. 1). The heat exchanger (HEX) and
solar share and consequently reductions on CO2 emissions of up to 34 %, condenser are responsible for discharging heat from the system, and
and the addition of a bottoming cycle resulted in a decrease on elec­ water at 15 ◦ C and 1 bar is chosen as the cold fluid for both. The model
tricity generation cost of up to 22 % comparing to equivalent conven­ assumes that this water is captured from a nearby river, and thus its cost
tional power generation technologies. is neglected. Pressure drops were defined in accordance with the works
When it comes to distributed-scale electricity generation, the solar of Dunham and Lipiński [21] and Rodrigues [20]: 5 % inside the solar
energy market is vastly dominated by PV solutions, since solar thermal receiver, an estimation based on the work of Pye et al. [22], and 2 % for
alternatives are often considered to be less cost effective. However, CSP each heat exchanger as well as the combustor. Loss of pressure inside the
microgeneration units offer some clear advantages such as their natural transmission pipes is neglected. The minimum temperature and pressure
fit with TES systems and the possibility for combined heat and power of the topping cycle are set to T1 = 308 K and p1 = 1 bar. At the bot­
generation. Currently, the majority of CSP microgeneration units are toming cycle, the working fluid temperature at the pump inlet, T12 , is
based on organic Rankine cycles (ORC), and the potentials of these fixed to 308 K, and its pressure, p12 , is either the saturation or ambient
systems for domestic applications have been reported in the literature. pressure, whichever is greater, assuring that it is either in saturated
Markides [16] presented a detailed review with technical considerations liquid or compressed liquid phase. At the pump outlet, the fluid pressure
on existent non-concentrated and low-concentration solar thermal sys­ is set to the value between p12 and 20 bar that yields the highest net work
tems for distributed power generation based on ORCs. Orosz et al. [17] output for the Rankine cycle, considering the imposed thermodynamic
proposed a holistic approach to evaluate the performance and costs of restrictions. The design hypothesis are listed on Table 2.
solar-powered ORCs, that he called SORCE model. When applying this The thermoeconomic model of the system was simulated using a
model to Spain’s climate, Orosz reported LECs of around 0.15 USD/kWh MatLab® algorithm that was developed for that purpose. This compu­
for a 10 kWe system. More recently, Freeman et al. [18] investigated the tational routine can be divided in two parts: the first is responsible for
potential of small-scale combined solar heat and power (CSHP) systems computing the energy streams flowing through each component, while
based on ORC for domestic use in the UK. Annual simulations revealed the second part calculates the flows of exergy and the system associated
that for a 15 m2 rooftop-mounted array, an average generation of 80–90 costs. It starts off by setting specific thermodynamic properties to key
We can be achieved with associated costs per unit of installed power points (or states) of the system. Most of these properties are inherent to
within the range of 53–68 USD/We. According to the authors, “a simi­ the model itself and were previously described, while others are user-
larly sized (15 m2) c-Si PV system (…) can be expected to output 200 We defined inputs that are summarized in Table 3. It then goes on calcu­
in the same climate, at an installed cost of around 59 USD/We”, which lating additional properties for each state by means of the balances of
means that the CSHP unit seems to be competitive in terms of cost per mass and energy for each component (Eqs. (1) and (2)) coupled with
installed power. Additionally, it “demonstrated a potential for produc­ thermodynamic computational software. For this purpose, the program
ing up to 86% of the required hot water for household consumption”. employs polynomial functions presented in the work of McBride et al.
The levelized electricity costs (LEC), however, do not seem to be very [23] for calculations referring to the topping cycle, and the C++ library
competitive for such small-scale CSHP systems. The authors reported of thermodynamic properties CoolProp for calculations referring to the
values within the range of 0.67–0.87 USD/kWh that tend to fall short
when compared to LECs of standard rooftop-mounted residential PV
arrays (0.184–0.3 USD/kWh in 2015 [19]). Even though the potentials Table 1
of solar-powered ORC systems for low-to-medium power applications Solar receiver design parameters.
have been reported in the literature, there is a clear lack of scientific Absorptivity of the receiver’s tubes surface α = 0.9
research on different alternatives such as solar-powered Brayton cycles Emissivity of the receiver’s tubes surface ε = 0.2
or hybrid systems. Overall heat transfer coefficient
[ ]
Uconv = 10 W/m2 K
The current work presents a thermoeconomic analysis and optimi­
Concentration ratio C = 70
zation of a hybrid CSP combined cycle power plant (CCPP) for

2
D.P.A. Coutinho et al. Energy Conversion and Management 247 (2021) 114669

→
10 Regenerator 9 symbols ρ, V ,Q̇,Ẇ,q̇, σ̇,s,Tamb , Tb ,Ėxd , and exf stand for the density, ve­
locity, rate of energy exchange in the form of heat, rate of energy or
exergy exchange in the form of work, rate of heat transfer per unit of
Receiver Comb.
surface area, rate of entropy generation, specific entropy, ambient
2 3 4
temperature, heat transfer boundary temperature, rate of exergy
5
destruction, and specific flow of exergy, respectively. All integral re­
lations assume steady state, neglect the contributions of the variations of
potential and kinetic energy, and were implemented in the MatLab®
algorithm as simplified single inlet–outlet balances.
Comp. Gas
∫∫
Turb. →
ρ V ∙→n dA = 0 (1)
1 6
Heat Recovery ∫∫
HEX →
8 Steam Generator 7 Q̇ − Ẇ = hρ V ∙→
n dA (2)
11

∫∫ ( ) ∫∫
q̇ →
dA + σ̇ = sρ V ∙→
n dA (3)
13 14 T b
∫∫ ( ) ∫∫
Tamb →
q̇ 1 − dA − Ẇ − Ėxd = exf ρ V ∙→
n dA (4)
Tb
Pump Steam
Turb. Then, the thermodynamic feasibility of the simulated solution is
checked by a routine that applies a set of restrictions, and all unfeasible
solutions are discarded. This step ensures that the results of the simu­
12 15 lations do not violate the laws of Thermodynamics. Finally, the costing
Cond. model is applied to define system costs related to operation and main­
tenance, fuel usage, land acquisition, components purchase, etc. Statis­
Fig. 1. System layout. tical correlations which are present in the literature were used to
approximate these costs (see [11,24–27]). Table 4 shows the costing
equations considered for the system’s components. The Chemical En­
Table 2 gineering Plant Cost Index (CEPCI) method was applied in order to ac­
System design hypothesis.
count for monetary inflation and market fluctuations.
Cycle parameter Design point condition Reference The chosen location for the generation unit was Évora, Portugal
Tw,in 288.15 K assumed (38.57◦ N, 7.91◦ W), given the high insolation values of this southern
pw 1 bar assumed region. Typical values for the direct normal irradiance (DNI) and
Δprec 5% [22] ambient temperature of the site were taken from a typical meteorolog­
ΔpHEX , Δpcond , Δpcomb 2% [21] ical year (TMY) data file, available in EnergyPlus™ database. According
T1 308 K assumed to the file, the yearly DNI of the region is close to 1600 kWh/m2 ⋅yr,
p1 1 bar assumed which is a significant value, although not as high as the values of more
T12 308 K assumed
p12 1 bar orpsat [T12 ] [21]
Table 4
System components costing equations.
Component Capital cost of the equipment, Ck [USD] Ref.
Table 3
( )( ) ( )
MatLab program inputs. Compressor m˙1 p2 p [24]
71.1 ln 2
0.9 − η p1 p1
Input variable Symbol Regenerator*
( )C
[25]
2143 A0.514
reg
[ ]
Nominal solar influx G0 W/m2 Solar 295Acol [11]
Ambient temperature Tamb [K] Receiver**
⎛ ⎞
Compression ratio pr = p2 /p1 Combustor [24]
⎜ m˙1 ⎟
Mass flow rate ratio in the regenerator r=m9 /m1 46.08⎜
⎝ p ⎠(1 +exp(0.018T5 − 26.4) )
⎟
0.995 − 5
Gas turbine inlet temperature T5 [K] ( p4 ) ( )
[ ] Gas turbine m˙1 p [24]
Regenerator heat transfer area Areg m2 479.34 ln 5 (1 + exp(0.036T5 − 54.4))
⎡⎛ 0.93 − ηGT⎞ p6
⎞0.8 ⎤
Steam turbine inlet temperature T14 [K] HRSG 0.8 ⎛ [24]
[ ] ⎢⎜Q̇economizer ⎟ Q̇
⎜ evaporator ⎟ ⎥
Solar collector area Acol m2 6570⎢⎣⎝ ΔTlm,ec ⎠ +⎝ ΔTlm,ev ⎠ ⎦ + 21276ṁb +
⎥
Compressor isentropic efficiency ηcomp [%]
Gas turbine isentropic efficiency ηGT [%] 1184.4m˙7 1.2
Steam turbine isentropic efficiency ηST [%] Heat 2143(A0.514
HEX )
[25]
Pump isentropic efficiency ηpump [%] exchanger*
( )0.05 ( )0.9
Steam Turbine T14 ηST [26]
538 × 103 × 1.9781ṁb p15 − 0.75
p14 1 − ηST
Condenser Q̇cond [26]
bottoming cycle. 430 × 0.582
ΔTlm
Δp−t 0.01 Δp−s 0.1
Once all the thermodynamic states and energy streams are fully Pump
(
ηp
)1.05
[26]
32 × 0.435ṁb 0.55 ΔP0.55
defined, the program employs entropy and exergy balances (Eqs. (3) and 1 − ηp
(4)) to compute the specific exergy at each state as well as the exergy *
Shell and Tube (CS)-CS Heat Exchanger.
destruction inside each component. Note that in Eqs. (1)–(4), the **
In accordance with the price of the EuroTrough models.

3
D.P.A. Coutinho et al. Energy Conversion and Management 247 (2021) 114669

than 2500 kWh/m2 ⋅yr common in the Sahara desert, for example. The
model was simulated for the following days, representative of each
season: 12th of July (Summer), 5th of October (Fall), 26th of January
(Winter) and 26th of March (Spring). For the sake of coherence, four days
with similar daily clearness indexes (Kt ) were chosen, and an effort was
made to meet the clear sky day condition (0.7 < Kt < 0.9) [28] when
possible. Table 5 presents the daily average values for the DNI and Tamb
with respect to each of the representative days. The objective of this
work was to provide reference values for the annual thermoeconomic
performance of a system of this kind and optimize it. Thus, as a
simplification, concerns regarding demand profiles were not considered
and all generated electricity was assumed to meet energy needs at any
given time.
An initial analysis was conducted in order to define a pair of working
fluids for the Brayton and Rankine cycles. The goal was to find the most Fig. 2. Algorithm flowchart and interaction with the MOO.
suitable combination to pursue the optimization analysis. A total of 25
combinations were considered: CO2 , air, molecular nitrogen (N2 ), he­ 1) pr ∈ [1, 20]
lium (He) and molecular hydrogen (H2 ) for the topping cycle, and water, 2) r ∈ [0.01, 0.99]
toluene (C7 H8 ), cyclopentane (C5 H10 ), octamethyltrisiloxane 3) T5 = 825K
(C8 H24 O2 Si3 ) and hexamethyldisilane (C6 H18 OSi2 ) for the bottoming 4) Areg = 6m2
cycle. During this analysis, the MatLab® model was simulated for all the 5) T14 = 550K
combinations and incremental values of compression ratio 6) Acol = 200 m2
(pr ∈ [1 : 20]), mass flow ratio in the regenerator (r ∈ [0.01, 0.99]) and 7) ηcomp = 79.6%
HRSG outlet temperature (T14 ∈ [500 K : 575 K]). The simulations 8) ηGT = 85.8%
assumed a typical summer day, and all remaining inputs were fixed to 9) ηST = 68.0%
standard values based on the work of Rodrigues [20]: 10) ηpump = 60.0%
T5 = 825 K,
As a simplification, it was assumed that the solar collectors are al­
Areg = 6 m2 ,
Acol = 200 m2 , ways normal to the sun rays, thus the incident nominal solar influx (G0 )
ηcomp = 79.6 %, is equal to the DNI. Accordingly, the nominal solar influx and ambient
ηGT = 85.8 %, temperature (Tamb ) were set as the respective daily average values for
ηST = 68.0 %, each of the representative days (see Table 5). Results of these pre­
ηpump = 60.0 %. liminary simulations are presented in Table 6, where ηglobal , εglobal , Ẇnet ,
Ėxd,system , Q̇input and fsol stand for the systems 1st and 2nd law efficiencies,
The best thermodynamic outcome was obtained with CO2 and cyclo­
net work output, global exergy destruction rate, total heat input rate and
pentane, i.e., greatest 1st and 2nd law efficiencies and net work outputs.
solar share, respectively. The solar share is defined as the fraction of the
Thus, this pair was chosen for the optimization analysis that followed.
total heat input that comes from the solar receiver (fsol = Q̇rec /Q̇input ).
The algorithm was developed in such a way that the 1st law efficiency is
3. Multi-objective optimization
directly constrained by the user-defined gas turbine inlet temperature
(T5 ) and compression ratio (pr ), so in the colder months when the gas
A multi-objective optimization (MOO) was conducted in order to
temperature at the receiver outlet is lower, the system will compensate
find the Pareto-optimal designs of the system under study. An evolu­
by burning more fuel in order to achieve the desired T5 . This is the
tionary algorithm denominated “gamultiobj” was employed for this
reason why ηglobal does not change much with the seasons. On the other
matter. It is a computational routine available in MatLab® optimization
toolbox, and a variant of the state-of-the-art NSGA-II controlled elitist hand, exergy efficiency peaks during Summer due to a higher ambient
genetic algorithm. The developed algorithm and the “gamultiobj” MOO temperature, while the net work output peaks during Winter. The
tool interact with each other as illustrated in Fig. 2. behaviour of this last property is also explained by the system operating
strategy adopted in the algorithm: a colder day results in a lower solar
receiver outlet temperature, T4 , which in turn results in higher mass
3.1. Reference case definition flow rates for the topping and bottoming cycle fluids, and consequently,
higher power outputs. This analysis revealed that for the given set of
Initially, the developed algorithm was used independently in order to input variables, the system maximum 1st and 2nd law efficiencies were
define a reference design case for comparison with the optimized de­ achieved for a mass flow rate ratio (r) of 0.01, the minimum considered
signs. During these preliminary simulations, most input variables were value, while the system maximum work output was achieved for mass
fixed to certain values of interest, while a set of iterative compression flow rate ratios between 0.33 and 0.36, depending on the season.
ratios (pr ) and mass flow rate ratios in the regenerator (r) with in­ Additionally, the r = 0.01 scenario is the most profitable, as it yields a
crements of 0.01 were considered, as follows: higher net present value (NPV = 13,420 €), while the maximum work
output scenario is not profitable at all, yielding a negative NPV (− 9718
Table 5 €). This was the first indication that the regenerator might be dispens­
Daily average DNI and Tamb in Évora of representative days considering solar able, depending on the system objective. Consequently, the MOO anal­
hours only. ysis was applied not only to the initially proposed model (with
regenerator), but also to the same model without regenerator or stream
Season Summer Fall Winter Spring
splitting valve in order to compare the performance of both designs. The
Representative day 12/07 05/10 26/01 26/03
maximum efficiencies scenario was considered as the reference/base
DNI [W/m2 ] 672.2 645.5 618.4 650.7
design, as it is the most profitable one, so the considered compression
Tamb [K] 300.7 295.9 282.0 286.3
ratio and mass flow rate ratio were 6.6 and 0.01, respectively.

4
D.P.A. Coutinho et al. Energy Conversion and Management 247 (2021) 114669

Table 6
Results of the preliminary simulations – reference/base design case.
Season Summer Fall Winter Spring

Maximization objective Max. Max.Ẇnet Max. Max.Ẇnet Max. Max.Ẇnet Max. Max.Ẇnet
η&ε η&ε η&ε η&ε
pr 6.6 6.6 6.6 6.6 6.6 6.6 6.6 6.6
r 0.01 0.33 0.01 0.34 0.01 0.36 0.01 0.34
ηglobal [%] 20.26 19.08 20.26 19.00 20.26 18.81 20.26 18.98
εglobal [%] 26.22 23.85 25.70 23.21 24.71 22.00 25.22 22.75
Ẇnet [kW] 31.32 32.59 31.74 33.54 33.80 36.82 33.30 35.49
Ėx d,system [kW] 81.15 96.61 82.54 100.76 86.68 111.33 84.77 104.71
Q̇input [kW] 154.62 170.83 156.68 176.48 166.85 195.67 164.37 186.98
fsol [%] 62.95 56.97 59.55 52.87 53.70 45.79 57.51 50.56

3.2. Objective functions The yearly revenue (R) was computed under the assumption that the
system would supply a small community, working as an auto-
The main targets of the MOO were to maximize the system ther­ consumption unit that may or may not be connected to the main grid.
modynamic performance, increasing revenue, while minimizing asso­ Therefore, it was calculated as the savings that such a community would
ciated costs. Therefore, total investment cost (Cinv ) minimization and net sustain if all its electricity came free of charge directly from the system
present value (NPV) maximization were chosen as objective functions. instead of the main grid, as expressed by Eq. (9). The increasing trend on
To calculate the investment cost, Eq. (5) was employed, which considers the specific cost of electricity (Celectricity ) was also assumed to be 0.3 % for
three contributions: costs associated with equipment purchase (Ceqp ); each year of operation, taking as initial value 215 €/MWh (average cost
land acquisition (Cland ); and power electronics, construction and civil of electricity for domestic users in 2019, accounting for all service ex­
engineering (Ccivil ). This approach was based on the methods used by penses such as transmission, distribution, etc. [29]). The computation of
Pihl et al. [11]. The direct capital cost of purchase of each equipment the electrical power output, Ṗel ,was based on the net work output and
was approximated by costing equations (see Table 4), and the resultant considered a typical electrical generator efficiency of 96 %.
values were converted into 2019 US dollars using the chemical engi­
neering plant cost index (CEPCI). The land purchase and civil engi­ R = Celectricity Ṗel H [€/y] (9)
neering related costs were estimated based on simplifying correlations
[27]. 3.3. Decision variables
Cinv = Ceqp + Cland + Ccivil [€] (5)
The program inputs exhibited on Table 3 were considered as decision
For the calculation of the net present value, Eq. (6) was employed, variables, with the exception of those that determine the ambient con­
∑n ditions, G0 and Tamb , which assumed fixed yearly average values during
R − CO&M
NPV = − Cinv + t [€ ] (6) the MOO (G0 = 646.7 W/m2 and Tamb = 291.23 K). Logically, when
t=1 (1 + i)
analyzing the model without regenerator, only 8 optimization variables
were considered, excluding the mass flow rate ratio and regenerator
where R stands for the yearly revenue of the power plant, i and n
area. For each decision variable, the following ranges of values were
represent the interest rate and expected lifetime of the plant, which were
considered:
fixed at 8 % and 25 years, respectively, and CO&M corresponds to the
yearly operation and maintenance costs, which were divided into two
1) pr ∈ [1, 20]
contributions – equipment associated costs (Co&meqp ) and fuel cost
2) r ∈ [0.01, 0.99]
(Co&mfuel ) - and were calculated as follows:
3) T5 ∈ [800K, 900K]
Co&meqp = Ceqp f (φ − 1) [€/y] (7) 4) Areg ∈ [2m2 , 10m2 ]
5) T14 ∈ [500K, 575K]
Co&mfuel = ṁfuel Cfuel LHVfuel H [€/y] (8) 6) Acol ∈ [200m2 , 400m2 ]
7) ηcomp ∈ [75%, 90%]
In Eq. (7), the maintenance factor (φ) was assumed to be 1.06, and
8) ηGT ∈ [75%, 90%]
the annuity factor (f) was calculated for the interest rate and expected
9) ηST ∈ [60%, 75%]
lifetime that was considered in Eq. (6). In Eq. (8), ṁfuel and Cfuel represent
10) ηpump ∈ [60%, 75%]
the mass flow rate (kg/s) and specific cost (€/kWh) of the natural gas
burnt in the combustor, respectively. The parameter H accounts for the
yearly operating time of the system that was assumed to be 4015 h/yr 3.4. Performance indicators
according to the number of solar hours throughout the year. For the
specific cost of natural gas, an increasing trend of 0.3 % (average For each Pareto-optimal solution, the algorithm computes a set of
inflation in Portugal, 2019) was assumed for each year of operation, performance indicators to be used as references for the measurement of
taking as initial value 40.1 €/MWh (average cost of natural gas for in­ its performance. The global cycle 1st (ηglobal ) and 2nd law efficiencies
dustrial users in 2019, accounting for all service expenses such as (εglobal ), electrical power output (Ṗel ), total heat input (Q̇input ), and exergy
transmission, distribution, etc.). These two values were taken from
destruction rate (Ėxd,system ) quantify the thermodynamic performance of
statistics published by PORDATA on their webpage [29]. Natural gas is
the system, while the payback period (PBP), the internal rate of return
mainly composed of Methane (CH4 ), with typical molar fractions of over
(IRR) and the levelized electricity cost (LEC) quantify its economic
90 % for this hydrocarbon. For this reason, the lower heating value of
performance. The calculation of the PBP is straightforward and requires
the fuel (LHV fuel ) was assumed to be equal to that of CH4 , i.e., 50,020 kJ/
no explanation, the IRR is computed using a built-in MatLab® function,
kg [30].
and the LEC is calculated in a simplifying manner as follows:

5
D.P.A. Coutinho et al. Energy Conversion and Management 247 (2021) 114669

f Cinv + CO&M 100%

LEC = [€/kWh] (10)
Ṗel H
80%

Normalized ranges
where f, Cinv , CO&M ,Ṗel , and H stand for the previously defined annuity
factor, investment cost, operation and maintenance costs, electrical
60%
power output, and yearly operating time, respectively. Finally, the solar
( )
share fsol as well as mass flow rate of CO2 emissions and savings
40%
(ṁCO2 emitted and ṁCO2 saved ) quantify the environmental performance of the
system. The carbon dioxide emissions savings correspond to the addi­
tional mass of CO2 that would be released into the atmosphere per unit 20%
of time if the same system was entirely powered by the combustor. These
last performance indicators are calculated as follows: 0%
1 3 5 6 7 8 9 10
MCO2
ṁCO2 emitted = ṁfuel [kg/s] (11) Decision variable / gene
MCH4
Fig. 4. Genetic distribution of Pareto-optimal solutions based on the previously
fsol × ṁCO2 emitted defined ranges, without regenerator.
ṁCO2 saved = [kg/s] (12)
1 − fsol

where MCO2 and MCH4 stand for the molecular weights of carbon dioxide local derivative of the plot keeps changing
and methane, respectively. This equation assumes stoichiometric com­ (f ’ (i) = (NPVi − NPVi− 1 )/(Cinvi − Cinvi− 1 ) ), and the second region,
bustion of the fuel, which is considered to be methane (CH4 ) as an composed of the last 43 subjects, where such derivative assumes a more
approximation. steady close to unit value (as can be seen in Fig. 3). The transition be­
tween regions is marked with a dashed line in Figs. 3, 5–7, and 9–12.
4. Results and discussion The NPV increases with Cinv throughout the Pareto front, thus a
higher initial investment results in a higher profit in the future. The
The results of the MOOs were a pair of Pareto fronts, one corre­ fitness values of Pareto-optimal solutions go from investment costs of
sponding to the simulation with the regenerator and the other without it, 118,264 € with near zero NPV values, to 328,706 € worth of investment
composed of 70 individuals (or subjects) each, as illustrated in Fig. 3. for a net present value of 267,373 €. Similar conclusions can be drawn
Looking at the two Pareto fronts, it is possible to conclude that the from the payback period and levelized electricity cost, which decrease
results from the MOO of the model without regenerator dominate those with Cinv , and the internal rate of return, which increases with Cinv (see
of the model with regenerator throughout most of the considered Fig. 5). Optimal solutions boast levelized electricity costs in the order of
domain, with the exception of the very low NPV subjects, which are not 0.159–0.221 €/kWh, which corresponds to 0.179–0.248 USD/kWh ac­
very interesting from an economic perspective. This is a strong evidence cording to the considered conversion rate. These results show that the
that the system thermoeconomic performance is indeed stronger if the optimized system can yield similar or even lower costs of electricity
regenerator is eliminated. In fact, without this component the model generation comparing to conventional stand-alone CSP systems, which
exergy destruction decreases, and a larger amount of heat is transferred take on global average LEC values of 0.182 USD/kWh according to
to the bottoming cycle, increasing, therefore, its power output. Addi­ IRENA [31]. Additionally, the solution appears to be competitive with
tionally, it is possible to have a greater solar collector area as well as typical PV systems, which achieved reported LECs of 0.151–0.242 USD/
more efficient components for the same investment cost. Hence, the kWh for domestic scale and 0.075–0.154 USD/kWh for Commercial &
further analysis refers to the case without regenerator, which was Industrial scale applications in 2019 [32]. The simulations conducted in
considered to be a more interesting design. this work do not consider the possibility to deliver residual heat in
The Pareto front is made up of Pareto-optimal designs that assume a addition to electricity, which would boost the competitiveness of the
wide range of values for the decision variables/genes (see Fig. 4) and system under consideration even further.
performance indicators. It can be conveniently divided into two distinct Both efficiencies (1st and 2nd law) follow an increasing trend along
regions: the first region, composed of the first 27 subjects, where the
0.23 18%
300
Levelized Electricity Cost [€/kWh]

0.22 16%

14%
Internal Rate of Return

250 0.21
12%
0.20
200
10%
NPV [k€]

0.19
8%
150
0.18
6%

100 0.17
4%
Model with regenerator
0.16 2%
50 Model without regenerator
0.15 0%
110 150 190 230 270 310 350
0
Cinv [k€]
110 150 190 230 270 310 350
LEC IRR
Cinv [k€]

Fig. 5. Levelized electricity costs and internal rates of return of Pareto-optimal

Fig. 3. Resultant Pareto front. solutions, without regenerator.

6
D.P.A. Coutinho et al. Energy Conversion and Management 247 (2021) 114669

40% 120 4.0E-04

35% 100 3.5E-04

#36
3.0E-04
30% 80
ηglobal or εglobal

Pel / Cinv [kW/€]

Pel [kW]
2.5E-04
25% 60

2.0E-04
20% 40
1.5E-04
15% 20
1.0E-04

10% 0
110 150 190 230 270 310 350 5.0E-05
Cinv [k€]
η ε Pel 0.0E+00
110 150 190 230 270 310 350

Cinv [k€]
Fig. 6. Global 1st and 2nd law efficiencies and electrical power outputs of
Pareto-optimal solutions, without regenerator.
Fig. 9. Relative electrical power output of Pareto-optimal solutions as a func­
tion of the investment cost, without regenerator.
70% 160

140
65% 0.3
120
60%
100 0.25 #36
mCO2 [t/y]
fsol [%]

55% 80
Pel / Acol [kW/m2]

0.2
60
50%
40
45% 0.15
20

40% 0
110 150 190 230 270 310 350 0.1
Cinv [k€]
fsol mCO2 saved mCO2 emitted 0.05

Fig. 7. Environmental performance indicators of Pareto-optimal solutions,

without regenerator. 0
200 250 300 350 400
Acol [m2]
2

Fig. 10. Relative electrical power output of Pareto-optimal solutions as a

function of the collector field area, without regenerator.

1 [33]) but quite lower when compared to standard natural-gas-fired

combined cycle power plants (CCPPs, 50–60 % [34]). This is due to
the constraints imposed to the gas turbine inlet temperature and
compression ratio, which were set to maximums of 900 K and 20:1,
respectively, quite low values when compared to standard CCPPs that
0 often yield values of over 1500 K and 35:1 [34]. However, without these
Cinv NPV LEC ηglobal εglobal Pel fsol limitative constraints, the solar shares would greatly decrease, and the
cost of the system would increase. Moreover, these CCPP reference
Base design Scenario #1 Scenario #2 Scenario #3
values refer to large-scale power plants, and so this comparison is not the
Fig. 8. Comparison of different scenarios regarding normalized fitness values most adequate. Overall, the LECs and efficiencies of Pareto-optimal
and performance indicators of interest. designs are quite interesting for this type of microgeneration unit. On
the other hand, the system net electrical power output increases almost
the first region, and then begin to stabilize after the transition to the linearly throughout both regions. This proves that the second is the
second region at around 26.9 % and 32.6 %, respectively (see Fig. 6). optimal efficiency region, where the main variation from one subject to
These parameters fall within the range of 16.46 %–27.97 % and 21.49 another is the increase of the collector field area accompanied by an
%–33.53 %, representing a valuable improvement to the reference increase on the system size and power output (economy of scale), as
design, which boasts maximum 1st and 2nd law efficiencies of 20.26 % illustrated in Fig. 6.
and 26.22 % respectively (see Table 6). Resultant 1st law efficiencies are As it can be observed in Fig. 7, the solar share (fsol ) is the highest for
higher than those of typical stand-alone PTC power plants (10–16 % the first subject (~65 %), and then assumes a decreasing trend until the

7
D.P.A. Coutinho et al. Energy Conversion and Management 247 (2021) 114669

7.E-04 hurting the other, so it is the designer’s task to consider all the solutions
and chose those that better fit the purpose of the system. With this in
#23
6.E-04 mind, three distinct decision-making scenarios were idealized, resulting
in three optimized designs. These scenarios are based on different
decision-making criteria: the first consists of a solution based on eco­
5.E-04
nomic criteria, the second focuses on thermodynamic criteria, and the
mCO2 saved / Cinv [t·€-1·y-1]

third emphasizes environmental criteria, prioritizing a greener solution.

4.E-04
Fig. 8 and Table 7 summarize the most important results of each sce­
nario. To facilitate comparison, these parameters have been normalized
3.E-04 in Fig. 8: the lowest and highest values of each parameter are set to 1 and
2, respectively, and the remaining values are normalized according to
2.E-04 this range. The thermodynamic indicators, which change seasonally,
have been averaged over the four representative days.
1.E-04

4.1. Scenario #1 – economic criteria

0.E+00
110 150 190 230 270 310 350
Although the LEC and IRR of the project follow a decreasing and
Cinv [k€] increasing trend with Cinv , respectively, the differences from one subject
to another (ΔLEC and ΔIRR) after the transition to the second region are
almost neglectable, as clearly seen in Fig. 5. Therefore, it was considered
that the most interesting solution from a strictly economic point of view
Fig. 11. Relative annual CO2 emissions savings of Pareto-optimal solutions as a
function of the investment cost, without regenerator.
would be the first individual of the second region (# 28), as it yields
close to peak LEC and IRR values for a relatively low Cinv . This solution
yields a NPV of 111,585 €, which comparing to the base design repre­
sents an increase of 98,165 € (731.5 %), for an initial investment that is
160
0.2 % lower (152,858 €). Furthermore, the project becomes profitable
140 after 7 years, which is 3 years earlier than the base design, the IRR is
6.67 % higher and the LEC assumes the competitive value of 0.161
120 €/kWh. The system 1st and 2nd law efficiencies are increased by 5.5 %
and 6.58 %, respectively, on an annual average basis. Even with a
100 significantly higher power output (+38.94 %), the total exergy
mCO2 [t/y]

#23 destruction of the system remains virtually unchanged. The biggest

80
downside of this solution is the lower solar share (1.83 % reduction on
an annual average basis) and consequential higher CO2 emissions
60
(annual average increase of 14.01 %) comparing to the base case.
40
4.2. Scenario #2 – thermodynamic criteria
20

Small-scale power plants are usually limited by two key factors: in­
0
0 40 80 120 160 200 240 280 320 vestment budget and available space (especially for the solar collector
NPV [k€] field in this case). Accordingly, the current scenario represents the
probable choice of an investor whose main objective is to maximize
mCO2 saved mCO2 emitted
generation power (Ṗel ) for interesting values of Cinv and Acol .
The generation power relative to investment cost and collector field
Fig. 12. Annual CO2 emissions of Pareto-optimal solutions with respect to the
area (Ṗel /Cinv and Ṗel /Acol ) follow a trend that is quite similar to that of
NPV, without regenerator.
the 1st and 2nd law efficiencies (see Figs. 9 and 10). Both variables
rapidly increase in value along the initial subjects and then stabilize at
36th subject, where it stabilizes at around 51 %. It is important to note
subject # 36. This behaviour proves that, once the system gets to the
that the employed algorithm invalidates any solution with a solar share
maximum efficiency region (# 36–# 70), every additional Euro invested
under 50 %, so this is the minimum acceptable value. As a reference,
results in a fixed increase in generated power. Thus, it was considered
looking at the Pareto-optimal solution with the closest NPV to that of the
that subject # 36 is the most interesting solution for the given scenario,
reference design (13,420 €) and comparing the solar shares of both so­
as it is the individual within the “stabilization region” that yields the
lutions, it can be seen that the optimized design yields an average solar
lowest Cinv and Acol . As it can be seen in Table 7, comparing to the
share of almost 65 %, which represents an increase of roughly 6.5 % on
reference design, the resultant investment cost is almost 19 % higher
an annual average basis, for an investment that is 20 % lower. These
(182,221 €), but the NPV and generated power are also much greater,
results reveal the clear superiority of some optimized designs from an
with respective percentual increases of 946.8 % (140,484 €) and 76.6 %
environmental point of view. Since the system power output and input
(55.16 kW) on an annual average basis. The economic performance of
are increasing with investment cost, it is expected that the annual mass
this design is generally good, but the system exergy destruction rate is
of CO2 emissions saved due to hybridization also increases with Cinv ,
23.55 % greater (on an annual average basis) than that of the base case.
even though the same trend does not apply to the solar share (see Fig. 7).
This is mostly due to the higher scale of the system, with its remarkably
For a design with the maximum considered solar collector area of 400
higher compression ratio, component isentropic efficiencies, gas turbine
m2, the amount of CO2 that is not discharged to the environment due to
temperature inlet, and solar collector area, which also explains the much
solar hybridization is over 150 tons per year.
greater electrical power output. At the same time, it assumes a quite
It is important to point out that all 70 subjects are optimal solutions,
poor environmental performance, with an annual average decrease of
since it is impossible to improve one of the objective functions without
7.7 % for the solar share and increase of 46.5 % for CO2 emissions.

8
D.P.A. Coutinho et al. Energy Conversion and Management 247 (2021) 114669

Table 7
Genes, fitness values and performance indicators with respect to each scenario.
Optimized designs

Base design Scenario #1 Scenario #2 Scenario #3

(Economic criteria) (Thermodynamic criteria) (Environmental criteria)

Decision variables/Genes
pr 6.6 6.58 7.97 6.44
r 0.01 X
T5 [K] 825 843.72 867.04 802.44
Areg [m2 ] 6 X
T14 [K] 550 575 575 575
Acol [m2 ] 200 211.78 229.36 208.04
ηcomp [%] 79.6 85.34 85.39 83.64
ηGT [%] 85.8 88.53 88.72 87.98
ηST [%] 68 74.66 74.88 74.47
ηpump [%] 60 67.56 70.05 63.12

Objective functions/Fitness values

Cinv [€] 153,192.7 152,858.1 (− 334.6) 182,221 (+29,028.3) 131,102.5 (− 22,090.2)
NPV [€] 13,420.4 111,584.9 (+98,164.5) 140,483.9 (+127,063.5) 75,690.7 (+62,270.3)

Performance indicators
PBP [y] 10 7 (− 3) 7 (− 3) 7 (− 3)
IRR [%] 8.99 15.66 (+6.67) 16.06 (+7.07) 14.14 (+5.15)
LEC [€/kWh] 0.211 0.161 (− 0.05) 0.162 (− 0.05) 0.168 (− 0.04)
ηglobal [%] * 20.26 25.76 (+5.5) 27.08 (+6.82) 22.63 (+2.37)
εglobal [%] * 25.46 32.04 (+6.58) 32.61 (+7.15) 29.42 (+3.96)
Ṗel [kW] * 31.24 43.40 (+12.16) 55.16 (+23.92) 33.00 (+1.76)
Q̇input [kW] * 160.63 175.53 (+14.9) 212.18 (+51.55) 151.91 (− 8.72)
Ėx d,system [kW] * 83.79 83.68 (− 0.11) 103.52 (+19.73) 72.51 (− 11.28)
fsol [%] * 58.43 56.60 (− 1.83) 50.73 (− 7.7) 64.25 (+5.82)
ṁco2 emitted [kg/h] * 13.53 15.42 (+1.89) 21.17 (+7.64) 11.01 (− 2.52)
ṁco2 saved [kg/h] * 18.93 20.05 (+1.12) 21.71 (+2.78) 19.70 (+0.77)

*
Annual average basis.

4.3. Scenario #3 – environmental criteria was developed in a MatLab® environment and used to perform a multi-
objective optimization of the system. The MOO revealed that there is a
As previously described, the system solar share assumes a steep great opportunity for improvement of the system thermodynamic, eco­
decreasing trend for the initial subjects (see Fig. 7) down to the mini­ nomic, and environmental performance by eliminating the regenerator
mum value of 50 %, where it stabilizes since solutions with lower fsol are from the model and varying certain operating parameters. The resultant
automatically discarded by the algorithm. Similarly, the behaviour of Pareto front is composed of 70 optimal solutions, which yield 1st and 2nd
mco2 saved /Cinv reveals that the first individuals yield a larger yearly law efficiencies within the range of 16.46 %–27.97 % and 21.49 %–
amount of CO2 savings relatively to its investment cost (see Fig. 11). It is 33.53 %, respectively, representing increases of up to 7.7 % and 7.3 %
interesting to realize that for the first 24 subjects, the Net Present Value comparing to the reference design. As expected, efficiencies are greater
greatly increases while the amount of CO2 emitted and respective sav­ than those achieved by stand-alone PTC power plants (10–16 % [33]),
ings do not change much (see Fig. 12). Looking at Figs. 7, 11 and 12, but still quite low when compared to standard natural-gas-fired com­
subject # 23 was selected as the optimized design for this scenario, given bined cycle power plants (CCPPs, 50–60 % [34]). Resultant levelized
that it boasts a great environmental performance for an interesting net electricity costs fall within the range of 0.159–0.221 €/kWh, evidencing
present value and investment cost. the competitiveness of such a system when confronted with more
From Table 7 and Fig. 8, one may conclude that the third scenario is common distributed-scale generation solutions such as solar PV or solar-
the only one that outperforms the base design in every considered powered ORC units.
parameter. The net present value of 75,691 €, representing an increase During a decision-making process, three specific designs were
of 464 %, is achieved for a lower investment cost of 131,103 € (− 14.4 selected, each of them corresponding to the most suitable solution for
%). The gas turbine inlet temperature is minimized, greatly reducing the three different scenarios which focus on economic, thermodynamic and
mass of burnt fuel and thus achieving annual average solar shares of environmental performance. The third scenario (environmental), is the
64.25 %. A reduction of 18.6 % in CO2 emissions is achieved with this only one that outperforms the reference design in every considered
design, evidencing its superior environmental performance. Lower heat parameter, with a net present value that is 464 % (75,690.7 €) higher for
generation inside the combustor is compensated by a larger area of solar an initial investment cost that is 14.4 % lower (131,102.5 €) and a
collectors, as well as higher steam turbine inlet temperature and competitive LEC of 0.168 €/kWh, as well as reduced CO2 emissions
component isentropic efficiencies, ultimately resulting in an even (− 18.6 %).
greater electrical power output. Globally, the system exergy destruction Results suggest that there is a potential for integrating CSP tech­
rate is reduced by 13.46 % on an annual average basis. nology in natural-gas-fired combined cycle units for distributed-scale
electricity generation. This solution appears to be competitive with
5. Conclusions standard PV residential units as well as solar-powered ORC systems. A
more detailed analysis of the model should be conducted in order to
A thermoeconomic model for a distributed-scale hybrid power plant support conclusions, including transient hourly simulations and

9
D.P.A. Coutinho et al. Energy Conversion and Management 247 (2021) 114669

considering the possibility to incorporate thermal energy storage as well [10] Baghernejad A, Yaghoubi M. Thermoeconomic methodology for analysis and
optimization of a hybrid solar thermal power plant. Int J Green Energy 2013;10(6):
as combined heat and power generation.
588–609.
[11] Pihl E, Spelling J, Johnsson F. Thermo-economic optimization of hybridization
CRediT authorship contribution statement options for solar retrofitting of combined-cycle power plants. J Sol Energy Eng
Trans ASME 2014;136:1–9.
[12] Buck R, Bräuning T, Denk T, Pfänder M, Schwarzbözl P, Tellez F. Solar-hybrid gas
Duarte Pereira Araújo Coutinho: Methodology, Software, Investi­ turbine-based power tower systems (REFOS). J Sol Energy Eng Trans ASME 2002;
gation, Validation, Writing – original draft. Jorge Augusto Maciel 124:2–9.
[13] Stein WH, Buck R. Advanced power cycles for concentrated solar power. Sol
Milhomem Rodrigues: Methodology, Software, Investigation, Valida­
Energy 2017;152:91–105.
tion. Aires José Pinto Santos: Conceptualization, Methodology, [14] Spelling J. Hybrid Solar Gas-Turbine Power Plants - A Thermoeconomic Analysis.
Writing - review & editing, Supervision. Viriato Sérgio Almeida Phd. thesis. Stockholm, Sweden: Dept. En. Tech KTH Royal Institute of Technology;
Semiao: Conceptualization, Methodology, Writing - review & editing, 2013.
[15] Livshits M, Kribus A. Performance and water consumption of the solar steam-
Supervision. injection gas turbine cycle. J Sol Energy Eng Trans ASME 2013;135:1–7.
[16] Markides CN. Low-concentration solar-power systems based on organic rankine
cycles for distributed-scale applications: Overview and further developments. Front
Declaration of Competing Interest Energy Res 2015;3.
[17] Orosz M, Hemond H, Quoilin S. Sorce: A Design Tool for Solar Organic Rankine
Cycle Systems in Distributed Generation Applications; 2016. p. 1–8.
The authors declare that they have no known competing financial [18] Freeman J, Hellgardt K, Markides CN. An assessment of solar-powered organic
interests or personal relationships that could have appeared to influence Rankine cycle systems for combined heating and power in UK domestic
the work reported in this paper. applications. Appl Energy 2015;138:605–20.
[19] Lazard. Lazard’s levelized cost of energy analysis - version 9.0; 2015.
[20] Rodrigues JAM. Hybrid Solar Thermal Power Plants. Msc. thesis. Lisbon, Portugal:
Acknowledgements Dept. Mech. Eng., Instituto Superior Técnico; 2019.
[21] Dunham MT, Lipiński W. Thermodynamic analyses of single brayton and combined
brayton-rankine cycles for distributed solar thermal power generation. J Sol
This work was financially supported by FCT, through IDMEC, under Energy Eng Trans ASME 2013;135.
LAETA, project UIDB/50022/2020. [22] Pye J, Morrison G, Behnia M. Pressure drops for direct steam generation in line-
focus solar thermal systems. Aust. New Zeal. Sol. Energy Soc. Conf.; 2006.
[23] McBride BJ, Zehe MJ, Gordon S. NASA Glenn coefficients for calculating
References thermodynamic properties of individual species. NASA/TP—2002-211556.
Cleveland OH, USA: Glenn Research Center; 2002.
[1] International Renewable Energy Agency. Renewable Power Generation Costs in [24] Bejan A, Tsatsaronis G, Moran M. Thermal Design and Optimization. John Wiley &
2019, Abu Dhabi; 2020. Sons; 1996.
[2] International Energy Agency. World Energy Outlook 2019; 2019. [25] Guo-Yan Z, En W, Shan-Tung T. Techno-economic study on compact heat
[3] Nathan GJ, Jafarian M, Dally BB, Saw WL, Ashman PJ, Hu E, et al. Solar thermal exchangers. Int J Energy Res 2008;32:1119–27.
hybrids for combustion power plant: a growing opportunity. Prog Energy Combust [26] El-Sayed YM. Designing desalination systems for higher productivity. Desalination
Sci 2018;64:4–28. 2001;134(1-3):129–58.
[4] González-Roubaud E, Pérez-Osorio D, Prieto C. Review of commercial thermal [27] Turchi C. Parabolic Trough Reference Plant for Cost Modeling with the Solar
energy storage in concentrated solar power plants: Steam vs. molten salts. Advisor Model (SAM). NREL/TP-550-47605. Golden CO, USA: National Renewable
Renewable Sustainable Energy Rev 2017;80:133–48. Energy Laboratory (NREL); 2010.
[5] Pelay U, Luo L, Fan Y, Stitou D, Rood M. Thermal energy storage systems for [28] Duffie JA, Beckman WA, Blair N. Solar Engineering of Thermal Processes.
concentrated solar power plants. Renewable Sustainable Energy Rev 2017;79: Photovoltaics and Wind. 2020.
82–100. [29] PORDATA. Base de Dados Portugal Contemporâneo. <https://www.pordata.pt>;
[6] Jin HG, Hong H. Hybridization of concentrating solar power (CSP) with fossil fuel n.d. [Accessed July 3, 2020].
power plants. In: Concentrating Solar Power Technology - Principles, [30] Michael MBB, Moran J, Shapiro HN, Boettner Daisie D. Fundamentals of
Developments and Applications. Woodhead publishing; 2012. p. 395–420. ch. 12. engineering thermodynamics. John Wiley & Sons; 2014.
[7] Powell KM, Rashid K, Ellingwood K, Tuttle J, Iverson BD. Hybrid concentrated [31] IRENA. Solar Energy. <https://www.irena.org/solar>; n.d. [accessed June 24,
solar thermal power systems: a review. Renewable Sustainable Energy Rev 2017; 2020].
80:215–37. [32] Lazard. Lazard’s levelized cost of energy analysis - version 13.0; 2019.
[8] Baghernejad A, Yaghoubi M. Exergoeconomic analysis and optimization of an [33] Moya EZ. Parabolic-trough concentrating solar power (CSP) systems. In:
Integrated Solar Combined Cycle System (ISCCS) using genetic algorithm. Energy Concentrating Solar Power Technology - Principles, Developments and
Convers Manag 2011;52(5):2193–203. Application. Woodhead publishing; 2012. p. 197–239. ch. 7.
[9] Baghernejad A, Yaghoubi M. Multi-objective exergoeconomic optimization of an [34] Boyce MP. “Combined cycle power plants. ch. 1. In: Combined Cycle Systems for
Integrated Solar Combined Cycle System using evolutionary algorithms. Int J near-Zero Emission Power Generation. Woodhead publishing; 2012. p. 1–43.
Energy Res 2011;35(7):601–15.

10