Transient analysis of an open-cell foam volumetric receiver ...
-
Upload
khangminh22 -
Category
Documents
-
view
1 -
download
0
Transcript of Transient analysis of an open-cell foam volumetric receiver ...
Transient analysis of an open-cell foam volumetric receiver
Miguel Ferreira Mora
Thesis to obtain the Master of Science Degree in
Engineering Physics
Supervisors: Prof. Luís Filipe Moreira MendesEng. João Pereira Cardoso
Examination Committee
Chairperson: Prof. Ilídio Pereira LopesSupervisor: Prof. Luís Filipe Moreira Mendes
Member of the Committee: Dr. Filipe Alexandre Ereira Mendes Marques
November 2019
Acknowledgments
First, I would like to thank my supervisors, Professor Luıs Filipe Mendes and Engineer Joao Pereira
Cardoso, for their dedication. Their expertise and advises throughout this thesis have constantly chal-
lenged me to improve my work. Additionally, I would like to thank Professors Pedro Coelho and Joao
Fareleira, who were available to discuss the physical and mathematical model in its early stages.
Moreover, I would like to express my gratitude to my girlfriend Ana Almeida, who has kept me focused
and motivated during this process. Without her love, I wouldn’t achieved these results. To my family who
supported me and made my life easier, thank you for everything.
iii
Resumo
A energia solar com concentracao tem vindo a destacar-se como uma alternativa para producao de
eletricidade. Neste setor, as torres solares demonstram maior potencial, devido a grande capacidade
de armazenamento e concentracao. Contudo, esta tecnologia implica investimentos elevados, o que
torna crucial a analise e melhoramento do seu desempenho.
Este trabalho tem como principal objectivo a optimizacao do funcionamento de um recetor volumetrico,
com espuma ceramica de celula aberta, atraves de simulacoes numericas. Para tal, desenvolveu-se um
modelo de transferencia de calor que permite a simulacao do comportamento do recetor no regime tran-
siente. De referir que foi assumida uma representacao unidimensional do modelo, pois as dimensoes
transversais sao muito maiores que a dimensao longitudinal do fluxo. Relativamente ao metodo radia-
tivo, foi aplicada a aproximacao de dois fluxos para simplificar a equacao de transferencia radiativa.
Com a otimizacao, foi possıvel estimar uma eficiencia maxima de 39.2 % para uma central eletrica,
com uma irradiancia incidente de 1000 kW.m−2 e um caudal massico de 0.71 kg.s−1. Adicionalmente,
constatou-se que das caracterısticas inerentes ao recetor, a porosidade apresenta o maior impacto no
perfil de temperaturas. Quanto a evolucao da temperatura em regime transiente, foi realizada uma
analise do tempo necessario para se atingir o equilıbrio termico. As simulacoes revelaram que o tempo
de equilıbrio e altamente dependente do caudal massico. Por fim, na analise do regime transiente,
foram identificadas tensoes mecanicas de origem termica que podem causar danos ao recetor. Assim,
estrategias para prevenir estes danos sao discutidas.
Palavras-chave: Energia solar com concentracao; recetor volumetrico; espuma ceramica
de celula aberta; modelo de transferencia de calor; aproximacao de dois fluxos; regime transiente.
v
Abstract
Concentrating solar power interest in the power generation market has risen over the years. Particu-
larly, central receiver systems (CRS) show great potential in this sector, due to their high concentration
ratios and storage capacity. Since the capital cost of CRS plants is still elevated, there is a need to
enhance the performance of these systems.
This thesis’s primary goals are to analyse and optimize the ceramic open-cell foam (OCF) volumet-
ric receiver thermal performance in an open loop system, through numerical simulations. To do that,
a heat transfer model was developed to predict the transient behaviour of the OCF receiver. As the
cross-section dimensions of OCF panels are usually much larger than the flow direction length, a unidi-
mensional representation is assumed, and the two-flux approximation was used to simplify the radiative
transfer equation.
A maximum system efficiency of 39.2 % is estimated for an incident irradiance of 1000 kW.m−2 and a
mass flow rate of 0.71 kg.s−1. In addition, simulation results show that porosity is the absorber’s intrinsic
property with the greatest impact on both receiver and air temperature profile. Moreover, a study of the
transient evolution of the receiver’s temperature was developed, mainly focusing on the time required
for the system to reach thermal equilibrium after a perturbation. Simulations have demonstrated that
equilibrium time is highly dependent on the mass flow rate. Finally, sources of thermal stress have
been identified during the transient analysis, which can damage the OCF absorber. Hence, strategies
to prevent this effect are presented.
Keywords: Concentrating solar power; volumetric receiver; ceramic open-cell foam; heat trans-
fer model; two-flux approximation; transient analysis.
vii
Contents
Acknowledgments . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . iii
Resumo . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . v
Abstract . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . vii
List of Tables . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . xiii
List of Figures . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . xv
Nomenclature . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . xvii
1 Introduction 1
1.1 Motivation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1
1.2 Objectives . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4
1.3 Methodology . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4
1.4 Thesis Outline . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5
2 Literature review 7
2.1 Receivers in Solar Towers . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7
2.1.1 Solid particle receivers . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7
2.1.2 Liquid receivers . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8
2.1.3 Gas receivers . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9
2.2 Volumetric receivers . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10
2.2.1 Metallic absorbers . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11
2.2.2 Ceramic absorbers: Monolithic honeycomb . . . . . . . . . . . . . . . . . . . . . . 12
2.2.3 Ceramic absorbers: Open-cell foam . . . . . . . . . . . . . . . . . . . . . . . . . . 15
2.3 Radiative models . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 16
3 Heat transfer model 19
3.1 Geometry of an OCF absorber . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 19
3.2 Transient heat transfer equations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 20
3.2.1 Internal nodes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 21
3.2.2 Boundary conditions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 22
3.2.3 Thermal evolution of the receiver . . . . . . . . . . . . . . . . . . . . . . . . . . . . 24
3.3 Radiative heat transfer . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 25
3.3.1 Radiative transfer equation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 25
ix
3.3.2 Two-flux approximation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 26
3.3.3 Collimated radiation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 29
3.3.4 Application of the two-flux approximation . . . . . . . . . . . . . . . . . . . . . . . 30
3.4 Parameters of the model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 32
3.4.1 Convective heat transfer coefficient . . . . . . . . . . . . . . . . . . . . . . . . . . . 32
3.4.2 Optical correlations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 34
3.4.3 Temperature correlations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 34
3.4.4 Fluid pressure . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 35
3.5 Courant-Friedrichs-Lewy condition . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 37
3.6 Algorithm . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 38
3.7 Absorber performance . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 39
4 Simulation results 41
4.1 Model validation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 41
4.2 Model comparison with a similar work . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 43
4.3 Steady state analysis: Absorber’s properties . . . . . . . . . . . . . . . . . . . . . . . . . 45
4.3.1 Thermal conductivity . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 45
4.3.2 Particle diameter . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 46
4.3.3 Porosity . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 48
4.3.4 Dispertion ratio . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 49
4.4 Steady state analysis: External conditions . . . . . . . . . . . . . . . . . . . . . . . . . . . 50
4.4.1 Thermal behaviour . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 50
4.4.2 Efficiency . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 52
4.4.3 Equilibrium time . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 54
4.5 Transient analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 55
4.5.1 Start-up and shutdown response . . . . . . . . . . . . . . . . . . . . . . . . . . . . 55
4.5.2 Cloudy weather response . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 56
4.5.3 Thermal stress . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 60
5 Conclusions 61
5.1 Future Work . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 62
Bibliography 63
A Finite-divided-difference formulas 71
A.1 Forward finite-divided-difference formulas . . . . . . . . . . . . . . . . . . . . . . . . . . . 71
A.2 Backward finite-divided-difference formulas . . . . . . . . . . . . . . . . . . . . . . . . . . 71
A.3 Centered finite-divided-difference formulas . . . . . . . . . . . . . . . . . . . . . . . . . . 72
x
B Auxiliary calculations 73
B.1 Transient heat transfer: Interior nodes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 73
B.2 Transient heat transfer: Boundary conditions . . . . . . . . . . . . . . . . . . . . . . . . . 74
B.2.1 Inlet . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 74
B.2.2 Outlet . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 75
B.3 Radiative heat transfer: Boundary conditions . . . . . . . . . . . . . . . . . . . . . . . . . 76
C Thomas algorithm 77
D Numerical implementation of the algorithm 79
E Detailed results of the simulation 91
xi
List of Tables
1.1 Characteristics of the CSP technology families. . . . . . . . . . . . . . . . . . . . . . . . . 2
3.1 Values of j for each cell geometry. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 33
4.1 Conditions for the model validation. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 41
4.2 Standard conditions of simulations. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 44
E.1 Detailed results of simulations performed in the validation. . . . . . . . . . . . . . . . . . . 92
E.2 Energy balance of simulations performed in the validation. . . . . . . . . . . . . . . . . . . 93
E.3 Detailed results of steady state simulations. . . . . . . . . . . . . . . . . . . . . . . . . . . 94
E.4 Energy balance of steady state simulations. . . . . . . . . . . . . . . . . . . . . . . . . . . 95
E.5 Summary of transient simulation results. . . . . . . . . . . . . . . . . . . . . . . . . . . . . 96
xiii
List of Figures
1.1 Flow diagram of a CSP plant. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1
1.2 Main CSP technology families (adapted from [15]). . . . . . . . . . . . . . . . . . . . . . . 3
2.1 Free-falling particle receiver solar tower with embedded storage and heat exchanger. . . . 8
2.2 Falling-film cavity receivers (adapted from [21]). . . . . . . . . . . . . . . . . . . . . . . . . 9
2.3 Volumetric effect compared with the tubular receiver response. . . . . . . . . . . . . . . . 10
2.4 CRS plant with an open volumetric gas receiver. . . . . . . . . . . . . . . . . . . . . . . . 11
2.5 Open loop metallic absorber (from [37]) and closed loop metallic absorber application on
the REFOS project (from [38]). . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12
2.6 Monolithic honeycomb structure (adapted from [40]) and principle of operation (from [10]). 13
2.7 Open-cell foam (from [47]) and its application on the DIAPR 30-50 project (adapted from [48]). 15
2.8 Radiative transfer equation methods. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 17
3.1 OCF receiver schematic. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 19
3.2 Intensity of radiation between parallel plates. . . . . . . . . . . . . . . . . . . . . . . . . . 27
3.3 Incident radiation on several models in a cold medium simulation. . . . . . . . . . . . . . . 29
3.4 Cell geometry. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 32
3.5 Particle’s minimum diameter as a function of the porosity. . . . . . . . . . . . . . . . . . . 37
4.1 Simulation results of the model compared with the experimental data of Pitz-Paal et al. [42]. 42
4.2 Steady state results of the model compared with the work of Kribus et al. [54]. . . . . . . . 44
4.3 Simulation results along the receiver for different solid thermal conductivities. . . . . . . . 46
4.4 Simulation results along the receiver for different particle diameters. . . . . . . . . . . . . 47
4.5 Temperature profile along the absorber for different porosities. . . . . . . . . . . . . . . . . 48
4.6 Simulation results along the absorber for different dispersion ratios. . . . . . . . . . . . . . 49
4.7 Temperature profile along the receiver for different ratios of incident energy per kilogram
of air (Qsup/m) and for different incident irradiances (Ginc). . . . . . . . . . . . . . . . . . 51
4.8 Overall results of the simulations (as a function of the incident energy per kilogram of air). 52
4.9 Overall results of the simulations (as a function of the mass flow rate). . . . . . . . . . . . 53
4.10 Equilibrium time (teq) as a function of the mass flow rate (m) and as a function of the
porosity (ϕ) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 55
4.11 Start-up and shutdown comparison of the receiver response. . . . . . . . . . . . . . . . . 56
xv
4.12 Transient thermal response of the receiver inlet (Ts,0) and fluid outlet (Tf,out) to a tempo-
rary decrease of incident irradiance (Ginc), for different time intervals of the cloud passage
(∆tlo). . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 57
4.13 Transient response for a shutdown procedure, with Ginc,up = 800 kW.m−2 and m =
0.5 kg.s−1. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 58
4.14 Equilibrium time required for the system to recover its normal operation after the cloud
passage. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 59
xvi
Nomenclature
Greek symbols
α Absorptance.
αp Specific surface area [m−1].
αsf Specific interfacial area [m−1].
β Extinction coefficient [m−1].
ε Emittance.
ζ Dispersion ratio.
ηcycle Power cycle efficiency.
ηsystem System overall efficiency.
ηthermal Receiver thermal efficiency.
θ Angle [rad, o].
κ Absorption coefficient [m−1].
λ Fourier number.
µ Dynamic viscosity [Pa.s].
µθ Function of the cosine.
ν Kinematic viscosity [m2.s−1].
ρ Density [kg.m−3].
σ Stefan-Boltzmann constant [W.m−2.K−4].
σsc Scattering coefficient [m−1].
σth Thermal stress [Pa].
τ Optical distance.
Φ Scattering phase function.
xvii
φ Biot number.
ϕ Porosity.
Ω Solid angle [rad].
ω Single scattering albedo.
Latin symbols
A Cross section area [m2].
Ap Particle’s area [m2].
a Coefficient of thermal expansion [K−1].
C Courant number.
Cg,2D Linear-focus geometric concentration ratio.
Cg,3D Point-focus geometric concentration ratio.
c Light speed [m.s−1].
cP Specific heat at constant pressure [J.kg−1.K−1].
ct1, ct2 Time counters (of the algorithm).
dh Pore hydraulic diameter [m].
dp Particle’s diameter [m].
E Young’s modulus [Pa].
G Incident radiation [W.m−2].
Ginc Incident irradiance [W.m−2].
h Volumetric heat transfer coefficient [W.m−3.K−1].
hsf Interfacial heat transfer coefficient [W.m−2.K−1].
I Intensity of radiation [W.m−2].
Ib Blackbody intensity [W.m−2].
J Radiosity [W.m−2].
k Thermal conductivity [W.m−1.K−1].
K1 Viscous permeability coefficient [m2].
K2 Inertial permeability coefficient [m].
Lr Receiver’s length [m].
xviii
Loss Thermal loss factor.
m Mass flow rate [kg.s−1].
n Last node of the receiver.
nt1, nt2 Time counter limit (of the algorithm).
Nu Nusselt number.
Pr Prandtl number.
p Pressure [Pa].
Q Power [W].
q Heat flux [W.m2].
Re Reynolds number.
Rf Specific gas constant [J.kg−1.K−1].
s Arbitrary distance [m].
T Temperature [K].
t Time [s].
u,v Darcy velocity [m.s−1].
Vp Particle’s volume [m3].
x Length [m].
y Width [m].
z Height [m].
Subscripts
amb Ambient air.
backsc Backscattering.
C Low temperature reservoir.
c Collimated.
cell Within a cell.
conv Convective.
d Diffuse.
eq Equilibrium.
xix
f Fluid (air).
H High temperature reservoir.
h Heliostat field.
i Internal node.
in Receiver’s inlet.
k Arbitrary direction.
L Receiver’s end.
l Local.
lo Lower value.
max Maximum.
min Minimum.
n Last node.
out Receiver’s outlet.
pr Printed.
R Radiative.
rad Emitted.
refl Reflected.
s Solid (receiver).
sky Sky.
sun Sun.
sup Supplied.
up Upper value.
v Volume.
w Wall (in front of the receiver’s outlet).
Superscripts
+ Forward.
− Backward.
t Time instant.
xx
Abbreviations
BC Brayton cycle.
CC Combined cycle.
CFD Computational Fluid Dynamics.
CFL Courant-Friedrichs-Lewy.
CPC Compound parabolic concentrator.
CRS Central receiver system.
CSP Concentrating solar power.
DAR Direct absorption receiver.
DLR German Aerospace Center.
DNI Direct normal irradiance.
DOM Discrete ordinate method.
HiTRec High temperature receiver.
HRSG Heat recovery steam generator.
HTF Heat transfer fluid.
LFR Linear Fresnel reflectors.
MH Monolithic honeycomb.
OCF Open-cell foam.
RC Rankine cycle.
RPC Reticulated porous ceramics.
reSiC Recrystallized silicon carbide.
RTE Radiative transfer equation.
sCO2 Supercritical carbon dioxide.
SiC Silicon carbide.
SiSiC Siliconized silicon carbide.
TES Thermal energy storage.
TSA Technology Program Solar Air Receiver.
VA Volumetric absorber.
WIS Weizmann Institute of Science.
xxi
Chapter 1
Introduction
In this chapter, the concentrating solar power interest is presented in the current context of decar-
bonization of the global economy, due to climate changes. The advantages and principle of operation
of this technology are briefly described. Moreover, the four known families of this kind are presented,
where the central receiver system is highlighted as the most promising technology.
1.1 Motivation
Climate changes occur due greenhouse gas emissions, which mainly results from burning fossil fuels
for electricity, thermal energy and transportation. Since demand for energy is expected to increase and
policies to reduce these emissions are being undertaken, the expansion of renewable energy influence
has been observed over the years in the power generation market [2]. One technology whose interest
has risen over the years is concentrating solar power (CSP), mainly due to the flexibility of its transformed
energy. In fact, from 2012 to 2013, its installed capacity had the fastest growth rate of all renewable tech-
nologies [3, 4]. CSP plants can store heat in a thermal energy storage (TES) system to later convert it
into electricity, which can be dispatched to the grid at any time [5]. In these systems, only the beam solar
radiation can be concentrated. Typically, a CSP plant project requires at least 2000 kWh.m−2 of direct
Figure 1.1: Flow diagram of a CSP plant (from [1]).
1
normal irradiance (DNI) per year to be economic viable currently. With lower DNI, solar photovoltaic
technology becomes a better alternative, since it uses both direct and diffuse irradiance [5, 6].
Commercial CSP plants were first deployed in 1984 in the United States. However, from 1991 to
2005, no plants were built due to a drop in oil and gas prices, and it was only in 2006 that the market
re-emerged [3, 7]. By the end of 2018, global CSP installed capacity reached 6069 MW [8].
The principle of operation of a CSP plant is presented in Figure 1.1. A set of mirrors, which track
the movement of the Sun, focus the beam solar radiation on the receiver. Those rays can be focused
on a receiver, where the Sun’s radiation is converted to thermal energy [5]. The heat attained from the
concentrated radiation in the receiver is then delivered to a heat transfer fluid (HTF). For commercial
applications, synthetic oils, molten salts, steam and air are the most common HTF used, but other fluids
are being studied [1, 9]. The heated fluid can be directly applied in a power generation cycle, or can
go to a TES system to be used later. The fluid heated in the solar receiver can transfer its energy to
the power cycle’s working fluid (in a heat exchanger), or can be directly used as the working fluid. The
latter statement depends on the CSP technology used, the selected HTF and power cycle [1, 3, 10]. If
necessary, a fossil-fuel back-up system can be added to the CSP plant to guarantee capacity and thus
enhancing performance [3, 5].
With the operation explained, the CSP types can now be analysed. The heliostat field can focus
the solar radiation linearly or in a single point. Concentration ratios for a linear-focus (Cg,2D) and for a
point-focus (Cg,3D) concentrators are given by equation (1.1):
Cg,2D =sin θhelsin θsun
, (1.1a)
Cg,3D =sin2 θhel
sin2 θsun, (1.1b)
where θsun = 4.65 mrad is the half-angle of the sun and θhel is the angular spread of the concentrated
radiation incident on the receiver, which ideally is 90 o [11]. Additionally, the solar receiver can either
be a stationary device, independent of the mirror field (fixed receiver), or can move together with the
focusing system (mobile receiver) [5]. Figure 1.2 and Table 1.1 show the main four CSP technologies,
which result from the possible combinations of the focus type with the receiver’s mobility [1, 5, 6, 12]:
• Parabolic troughs consists on parallel rows of reflectors curved in one dimension (typically 100 m
Table 1.1: Characteristics of the CSP technology families (from [3, 5, 6, 13, 14]).
CSP technology Parabolic trough LFR Parabolic dish CRS
Solar Collector Linear Focus Linear Focus Point Focus Point FocusSolar Receiver Mobile Fixed Mobile FixedConcentration ratio 70− 80 > 60 > 1300 > 1000
Typical capacity [MW] 10− 300 10− 200 0.01− 0.025 10− 200
Current efficiency [%] 15− 16 8− 10 20− 25 16− 17
2
(a) Parabolic trough (b) Linear Fresnel reflector
(c) Parabolic dish (d) Central receiver system
Figure 1.2: Main CSP technology families (adapted from [15]).
long or more, with 5 to 6 m of curved aperture). The linear focus property of this system grants
a simpler tracking of the sun (usually a horizontal N-S axis with E-W tracking), and the mobile
receiver feature allows the system to collect more energy (when compared to a fixed receiver). It
has been commercially proven using synthetic oils as HTF, and plants with molten salts are under
demonstration;
• Linear Fresnel reflectors (LFR) uses long rows of flat (or slightly curved) mirrors that reflect the
sun’s rays onto a downward-facing tubular receiver. The existent fixed receiver enables the trans-
port of collected heat to the power block more easily. LFRs are best suited to use direct steam
generation, making storage capacity more difficult to implement. Plants with molten salts are also
under demonstration;
• Parabolic dish (or Stirling dish) is a parabolic dish-shaped concentrator that reflects solar irradiance
onto a receiver at the focal point of the dish, which have an independent Stirling engine embedded
(or a micro-turbine). The point focus receiver allows the use of a two-axis tracking, and hence, a
higher temperature on the receiver. The low available capacity of each dish is the downside of this
technology;
• Central receiver system (CRS), also known as solar tower, consists on a field of heliostats (computer-
3
controlled mirrors with a two-axis tracking system) that focus the sun’s rays in a fixed receiver
placed at the top of a tower. Most of the current commercial plants use molten salts as the HTF,
but other fluids with higher operating temperatures are being considered in future plants.
Of all the CSP types presented, the central receiver system has the greatest potential to become the
preferred technology [3, 6]. The high concentration ratio results in higher temperatures (up to 1000 oC)
and thus in a better efficiency of the power block. This technology has also the flexibility to be imple-
mented in existing fossil fuel plants for hybrid operation, and offers the best TES capacity of all CSP
types [5, 6, 13]. Additionally, the cost of all the central receiver subsystems (heliostat field, solar re-
ceiver and power block) is expected to reduce over the years due to technology innovations. All of
these factors have the potential to make electricity from solar towers cheaper than other CSP technolo-
gies [1, 3, 6, 13]. The numerous developments that are taking place in CRS plants have created an
interest in current innovations on this sector, as well as provided an opportunity to enhance knowledge
about CSP technologies. Those reasons have motivated the presented work.
1.2 Objectives
Since the capital cost of a solar tower is still high, numerical simulations have been increasingly used
as a tool for planning and optimizing CRS plants. With this approach, each subsystem can be tested
and improved individually, so that the power plant overall efficiency is enhanced [16, 17]. In light of this
trend, the purpose of this thesis is to create an one dimensional heat transfer model of an open-cell
foam (OCF) absorber in transient regime. This absorber is a type of volumetric receiver that can be
installed in open loop solar tower power plants, using air as the HTF. So far the literature doesn’t report
any one-dimensional transient model applied to OCF receivers. The development of this model is fully
justified due to the particular geometry of this type of receiver and constitutes a new tool, suitable for the
transient behaviour analysis of OCF absorbers [16].
With this work, the thermal performance of the receiver in steady state can be improved, whether it
is on its intrinsic properties or on the imposed external conditions. Additionally, the absorber’s transient
behaviour is evaluated in order to assess possible problems that may arise during its operation. Regard-
less of the analysis, the model must be flexible enough so that a wide range of these receivers (with
different compositions and properties) can be simulated.
1.3 Methodology
The developed work presents a numerical model for an open-cell foam absorber’s energy distribution
in time and space. This receiver consists of a porous structure, where an air flow is imposed in the lon-
gitudinal direction. Due to the volumetric effect, local thermal non-equilibrium is assumed in this system.
This means that the solid structure and the fluid are treated separately, where the energy expressions
of both media are coupled by a convective term [16]. Moreover, one knows that the dimensions of the
4
cross section of the receiver’s panels are typically much larger than the flow direction length. For this
reason, it is possible to apply a unidimensional representation to the developed work, along with the
two-flux approximation as the radiative method.
The receiver is then discretized in the direction of the fluid flow, where each node contains detailed
information about its thermal properties. This implies that every thermal related quantity of both media
must be discretized as well. Regarding evolution in time, the explicit method was first considered.
However, it was soon rejected, since a complex stability criteria was required for the non-linear equations
of the model [18]. Therefore, the implicit method was used to advance in time. This method is more
difficult to implement than the explicit form, but it has the advantage of being unconditionally stable
numerically [19]. Nevertheless, physical instabilities can arise if the Courant-Friedrichs-Lewy condition
is ignored [20].
The application of the transient heat transfer model is then divided in three steps. First, the model is
validated against experimental data and also compared with a simulation work. Then, an optimization of
the receiver’s performance is carried out by varying the intrinsic properties of the absorber, as well as by
changing the receiver’s surrounding conditions. Lastly, a study more focused on the transient evolution
of the open-cell foam absorber is accomplished. With all those simulations performed, a full assessment
of the model is completed.
1.4 Thesis Outline
This thesis is divided in five parts. The current chapter shows an overview of concentrating solar
power technology, where the principle of operation of each type is explained. Chapter 2 provides a re-
vision on the state-of-the-art receivers for solar towers, with special emphasis on volumetric receivers.
In addition, a literature review on radiative models is also presented. In the third chapter, the heat trans-
fer model is presented. A detailed discussion of all simulation results is given in Chapter 4, where the
validation of the model with experimental data is performed as well. Finally, the fifth chapter summa-
rizes the advantages and limitations of the proposed model. Additionally, this last chapter shows some
suggestions for future studies using the developed model.
5
Chapter 2
Literature review
In this chapter, a revision on documented state-of-the-art absorbers of CRS is presented, in which
the receivers are categorized in respect to the state of the heat transfer substance. A special attention is
given to volumetric receivers, where its principle of operation and implementation is evaluated in detail,
both in metallic and in ceramic absorbers. Finally, a brief discussion on radiative models (applied to
volumetric absorbers) is presented, which have highlighted some works where radiative methods are
used on open-cell foam absorbers.
2.1 Receivers in Solar Towers
All of the solar tower’s subsystems are currently being optimized in R&D activities [3]. Regarding
solar receivers, several designs of this component have been tested over the years in order to improve
its performance. Ho and Iverson [21] have suggested dividing these state-of the-art designs in three
categories, according to the state of the heat transfer substance.
2.1.1 Solid particle receivers
Solid particle receivers can work in a wider range of temperatures when compared to other receivers.
During the 1980s, the free-falling particle receiver was the first proposed absorber of this kind. As shown
in Figure 2.1, sand-like ceramic particles are lifted by an elevator to the top of the tower and released
through an opening. This creates a thin curtain of particles that fall through a cavity receiver, which is
directly irradiated by concentrated sunlight. The desired outlet temperature is achieved by controlling
the particles’ mass flow with a slide gate. These high temperature particles are then stored to later
heat a secondary working fluid for the power cycle [21–23]. Although these particles are cheaper than
molten salts, they present some disadvantages, such as higher convective heat losses (due to entrained
airflow), low residence time on the receiver and particle spillage. A thermal efficiency of 50 % was
reached by this system in 2010 [24].
Other designs of solid particle receivers have been studied to solve some of the issues stated above.
For instance, the obstructed falling particle receiver has a porous structure or obstacles which slows the
7
downward velocity of particles, thus increasing their residence time in the receiver. Another example is
the centrifugal receiver for solid particles which uses the rotational velocity of the kiln to achieve a better
control of their residence time. Indirect methods for heating solid particles, such as the gravity-driven
particle flow, avoid particle loss and reduce convective loss. Particles in a enclosure fall around an array
of tubes, whose interior surfaces are irradiated by concentrated sunlight. Fluidization of solid particles
has also been considered [23].
2.1.2 Liquid receivers
Most of the CRS commercial plants use liquid-based receivers, due to their high heat-transfer rates
and high specific heat. Molten salts are currently the most common HTF used, but fluids with higher
boiling points and lower melting points are being studied as well [3, 23, 25].
One type of liquid receivers is the falling-film receiver, in which a film of HTF flows down an inclined
wall (due to gravity only) while it is heated by concentrated sunlight. Figure 2.2 presents two methods
that have been proposed for this kind of receivers. The first one is the direct absorption receiver (DAR),
where the curtain of fluid is irradiated directly at an exposed external wall or at an internal cavity. With
this approach, the outlet fluid temperature reaches the highest value of all liquid receivers, but the ex-
posed HTF will eventually pick up some impurities. Alternatively, one can use indirect-exposure receiver
designs, where the irradiated external walls transfer the heat to the film of fluid that slides inside the
internal cavity walls. In both presented methods, issues with the film stability have yet to be further in-
vestigated [21]. Theoretical predictions show a thermal efficiency of 94 % for the DAR, while the indirect
approach can potentially reach values greater than 80 % with design improvements [26, 27].
Another type is the tubular liquid receiver, which has already been commercially proven. Like the
name suggests, it consists on a set of thin hollow tubes whose external walls are exposed to concen-
trated radiation. The heat is then transferred to the interior of each tube, where the working fluid flows.
The tubular design can either be an external receiver or a cavity absorber. In addition, pipes can be
Figure 2.1: Free-falling particle receiver solar tower with embedded storage and heat exchanger(from [21]).
8
arranged in order that the HTF passes through the receiver multiple times, thus allowing the absorption
of more energy. Although this system needs to be pre-heated in the morning, it has the advantage of
avoiding any source of HTF spillage, as well as tolerating high pressures of the working fluid [21]. Ex-
perimental results from 2010 show a thermal efficiency of 86.2 % for this kind of receiver [28]. Further
improvements are expected with the research of tubular receiver materials that can withstand higher
temperatures and better resist to the working fluid’s corrosion effect [23].
2.1.3 Gas receivers
Although gases have a low thermal conductivity, receivers of this kind can achieve temperatures
over 1000 oC. Air has been the most chosen working fluid for these systems, but applications using
supercritical carbon dioxide (sCO2) are being developed [21, 23].
One absorber of this kind is the small particle air receiver, where an aerosol of carbon nano particles
and air is heated in a pressurized cavity. The cumulative area of the particles absorbs the concentrated
radiation and transfers the energy to the air, which undergoes a Brayton cycle. This method is able
to reach high air temperatures without damaging the receiver, however, it requires a system that can
maintain the aerosol [21]. Theoretical results show that the efficiency of this receiver can be as high as
90 % [29].
Similar to liquid receivers, tubular receivers were also implemented for gases, where the heated
HTF can be directly used in the power block (in a Brayton cycle). Considering air as the working fluid,
liquid-metal heat pipes were proposed due to their high heat transfer coefficient, which can provide
a more compact heat exchange of the concentrated radiation to the gas. High air temperatures and
pressures can be attained for this kind of receiver, but large convective and radiative heat losses are
also expected. For this particular design, a thermal efficiency of 85 % was theoretically obtained, with an
outlet HTF temperature of 815 oC [21, 30]. Other designs of tubular gas receivers were developed using
(a) Direct-exposure absorption receiver. (b) Indirect-exposure absorption receiver.
Figure 2.2: Falling-film cavity receivers (adapted from [21]).
9
sCO2, such as positioning the tubes in a blade structure, to increase light traping. Procedures using this
HTF are able of reaching thermal efficiencies of 50 % using higher temperatures. The downside is that
materials which accommodate the high pressure sCO2 are more expensive [23].
An innovative type of technology is the microchannel tube receiver, that consists on a set of panels
with small tubes where the working fluid passes through. The main advantage of this system is that it
has a higher surface area between the receiver and the HTF when compared to other receivers, thus
increasing heat transfer to the gas. Modules of this receiver can be added to suit higher capacities,
although construction difficulties and increase in cost may arise. Thermal efficiencies of 90 % were
reported for this system [23].
2.2 Volumetric receivers
The gas receiver system left to address is the volumetric air absorber (VA), which first appeared
in the 1980s as a cheaper alternative that had better thermal properties than tubular receivers. Its
working principle is quite simple. A porous material is placed in a volume inside the receiver, that is
irradiated by concentrated sunlight. The radiation is absorbed throughout the porous structure, heating
this material. The attained heat is then transferred to the flowing air. Finally, a blower in this system
makes the air flow through the porous media, where the latter transfers its heat to the passing fluid by
forced convection [10].
The ideal temperature distribution of this receiver is shown on the bottom right of Figure 2.3. Cold air
enters the structure from the irradiated side and cools the receiver in that zone, thus reducing radiative
heat losses. The air gets heated as it crosses the structure, reaching the end of the receiver theoretically
Figure 2.3: Volumetric effect (right) compared with the tubular receiver response (left) (from [31]).
10
with no temperature difference between them. The so called volumetric effect is attained when the inlet
receiver temperature is lower than the outlet temperature. This is an advantage compared to tubular
receivers, whose outer walls must be hotter to provide the same outlet air temperature, which leads to
higher radiative losses. Additionally, volumetric receivers can withstand larger solar fluxes than tubular
receivers, meaning that VAs can achieve the same power with a smaller aperture area [31]. However,
flow instabilities can arise when high solar fluxes (above 2000 kW.m−2) are applied to the volumetric
receiver, which will shorten the material’s durability [32].
CRS plants using volumetric air receivers have been under development over the last years. Up to
date, two possible designs for a CRS plant using volumetric air receivers have been created: closed loop
pressurized system and open loop atmospheric system [10, 23]. Closed loop volumetric receivers use
pressurized air as the working fluid for the power block, which can either be a Brayton Cycle (BC) or a
Combined Cycle (CC). For the latter, cycle hybridization has been proposed, where the solar contribution
would preheat the air that flowed from the compressor to the combustion chamber of the gas turbine.
This application grants a higher efficiency for this cycle, as well as a wider power range [9, 33, 34]. On
the other hand, open loop receivers use atmospheric air to heat a secondary working fluid. The heated
air generates steam in a heat recovery steam generator (HRSG) to power a Rankine Cycle (RC), as
seen in Figure 2.4. Additionally, an air return system can be incorporated within this system, as a way
to recycle colder air that left the heat exchanger and to cool the receiver as well [9, 35, 36].
Volumetric receivers are typically employed when working fluid temperatures above 800 oC are re-
quired, so the porous structure must endure those high temperatures as well. Metallic porous receivers
are used in applications where the maximum outlet air temperature is below 1000 oC. If higher temper-
atures are needed, one must use ceramic absorbers instead. The most common ceramic materials are
silicon carbide (SiC) and siliconized silicon carbide (SiSiC), which can heat the air up to 1500 oC and
1200 oC respectively [10, 21, 23]. The following sections provide a description of each VA family, as well
as its performance on implemented projects.
2.2.1 Metallic absorbers
In this kind of receivers, the passing air is heated by a metal wire mesh. These absorbers have been
applied in several projects which require lower outlet temperatures. In some cases, they were used to
Figure 2.4: CRS plant with an open volumetric gas receiver (from [31]).
11
test the power plant operation before scaling the project to higher temperatures (which require ceramic
absorbers) [10].
Metallic receivers have been applied in open and closed loop power plants. Starting with the former,
Phoebus-TSA (Technology Program Solar Air Receiver) is the reference project of this type. Designed
by the Swiss SOTEL Consortium, the 2.5 MWth metallic receiver was able to reach in 1993 an efficiency
of 85 % at 700 oC. The heated air was then capable to produce steam at 480− 540 oC and 35− 150 bar in
the HRSG. Although this absorber was considered a success, a larger plant using the TSA concept was
never funded, due to the outlet temperature restriction [9]. Nevertheless, Capuano et al. [37] presented in
a recent study an innovative design of a titanium-aluminium alloy receiver, as exhibited in Figure 2.5(a).
With manufacturing improvements, this technology has the potential to surpass the current state-of-the-
art receiver for open loop cycles.
Regarding absorbers in closed loop cycles, REFOS is the reference project, which was developed
by DLR (German Aerospace Center). Figure 2.5(b) shows a representation of the REFOS absorber’s
operation, where the incident radiation is not only focused by the heliostat field but also by a secondary
concentrator. The Incoloy 600 receiver is placed on the convex side of the domed quartz window, where
pressurized air flows through [38]. Design conditions predicted an outlet temperature of 800 oC, an
operating pressure of 15 bar and a thermal efficiency of 80 % for a 350 kWth thermal power module. Tests
in 1996 yield a thermal efficiency of 67 % at 800 oC, which was subsequently raised due to improvements
on the secondary concentrator [9, 38]. The REFOS receiver scheme was later applied in the first phase
of the SOLGATE project [9, 10, 39].
2.2.2 Ceramic absorbers: Monolithic honeycomb
Monolithic honeycomb (MH) receivers are a group of ceramic absorbers that are typically employed
in open loop systems [10]. In this structure, atmospheric air gets heated as it crosses the array of
parallel channels, as presented in Figure 2.6(a). Two main manufacturing processes are known for
(a) Metallic absorber concept. (b) REFOS receiver scheme.
Figure 2.5: Open loop metallic absorber (from [37]) and closed loop metallic absorber application on theREFOS project (from [38]).
12
these receivers: by infiltration, that forms siliconized silicon carbide; or by recrystallization, which gives
recrystallized silicon carbide (reSiC) [41].
Most of the first monolithic honeycomb absorbers showed cracks on its structure after being tested.
One of their main problems was the tensile stress caused by flow instabilities. This issue was first
observed for VAs in 1995, during the Catrec 2 tests. In light of these events, the High Temperature
Receiver (HiTRec) technology was designed to minimize these instabilities, using techniques learned
from tests of former VA [36, 42]. Its principle of operation is revealed in Figure 2.6(b). A set of ceramic
absorbers modules, with less than 0.25 m of diameter each, are mounted on top of a stainless steel
frame. The ceramic absorber is set on a porous cup, making possible for the absorber to move or expand
freely while operating the receiver. Gaps between modules allow the return air of the system to escape
(or be reused) while cooling the stainless steel frame. The gaps also provide an easy replacement of
individual modules, either due to maintenance reasons or to install a new kind of absorber, allowing a
simple way to implement technology innovations on this receiver [36].
In 1998, the 200 kWth HiTRec I was tested, where the goal was to reach a mean outlet temperature
higher than 1000 oC. This receiver consisted on 37 hexagonal modules, making a total aperture area
of 0.491 m2. Each module had a SiSiC cup and a reSiC structure as the absorber material, with an
open porosity of 49.5 %. To cool the structure below the modules, a inner duct using only ambient air
was planned. Tests results showed a maximum outlet temperature of 980 oC associated to an efficiency
of 68 %. Higher temperatures couldn’t be attained due to limitations of the test bed. For an outlet
temperature of 800 oC, thermal efficiencies between 75 % and 80 % were attained. Additionally, a low
temperature difference all over the receiver led to no hot spots. Despite its apparent good operability,
some disadvantages were found. A design error combined with a poor control of the return air caused
deformations the stainless steel, which is unacceptable for larger receivers. In addition, modules had
too much freedom due to an inadequate fixation of those components to the retainer device [36, 43].
In the following years, the 200 KWth HiTRec II was developed to solve the identified issues of its
predecessor. The receiver had 32 hexagonal modules, making a total aperture area of 0.41 m2, The
modules were mass produced using a simplification of the prior receiver, and improvements of the
(a) Monolithic honeycomb samples. (b) HiTRec operating scheme.
Figure 2.6: Monolithic honeycomb structure (adapted from [40]) and principle of operation (from [10]).
13
retainer device were made. In addition, the frame of the receiver was substituted by a steel-nickel alloy
(Incoloy 800) well suited to withstand high temperatures, and whose expansion coefficient is similar to
the SiSiC. A cooling system was also planned with the return air of the heat exchanger. Tests results
showed not only that the structure issue was solved, but also that the return air cool it to temperatures
lower than 500 oC, which meant that cheaper materials could be used in the future for this component [36,
44]. A mean efficiency of 72 % has achieved for a 800 oC outlet air temperature, and 76 % has observed
for a 700 oC outlet air temperature. Efficiencies for HiTRec I in the same conditions were slightly higher,
due to higher radiative losses on the receiver’s sides in the current tests. Nevertheless, HiTRec II solved
the stated problems of its previous version and shows potential for improvements as well. This MH
receiver is still considered to be the current state of the art for open volumetric air receivers in pre-
commercial industrial scale [35–37, 44].
In 2001, a project of two stages was launched: the SOLAIR project, which directly used the HiTRec
technology. The first stage (SOLAIR 200) aimed to design a highly efficient receiver for outlet tempera-
tures over 800 oC in high-flux conditions. The 200 KWth receiver consisted on 36 square-shaped SiSiC
cups, that created an aperture area of 0.62 m2. The modules are mounted on a refractory steel frame,
due to the successful results of HiTRec II [10, 45]. Tests started in 2002 using reSiC as the absorber ma-
terial (configuration 1). Signs of oxidation on the receiver led to the substitution of the top 16 absorbers,
where the material was replaced by SiSiC absorbers (configuration 2). For an outlet air temperature of
800 oC, efficiencies of 75 % and 74 % were achieved respectively for configuration 1 and 2. However, for
a 700 oC outlet temperature, an efficiency of 81 % and 83 % was attained respectively for configuration 1
and 2, meaning that the latter arrangement is well suited for applications below 750 oC. A temperature
difference below 100 oC was observed across the whole absorber for an outlet temperature of 750 oC,
which indicated an easy operability [45].
In 2003, the second stage of the SOLAIR project (SOLAIR 3000) was initiated, where the purpose
was to raise the receiver useful power up to 3 MWth. The 270 square-shaped SiSiC cups with a reSiC
absorber needed to generate stable mean outlet temperatures between 680 oC and 800 oC, as well as
endure temperatures of 1000 oC. Test results showed that this receiver endured an incident power as
high as 2950 kWth. For a solar flux between 370 kW/m2 and 520 kW/m2, and with an outlet temperature
of 750 oC, efficiencies varied from 70 % to 75 %. For the same outlet temperature, a temperature spread
as high as 450 oC was recorded across the whole absorber [10].
With the success of the SOLAIR 3000, a solar thermal power tower plant was planned in 2006. The
solar tower Julich was built to be the first pre-commercial power plant using the open volumetric receiver
concept. The 55 m tower uses the HiTRec technology, where 1080 modules form an overall aperture
area of 23 m2. Combined with a thermal energy storage system, an outlet air temperature of 700 oC
produces steam at 500 oC and 100 bar, in order to drive a 1.5 MWe turbine generator. Solar electricity
is being delivered to the grid since 2009, but research programs are still being done to promote this
receiver to a fully commercial application [10, 31, 46].
14
2.2.3 Ceramic absorbers: Open-cell foam
Open-cell foams (OCF), also known as reticulated porous ceramics (RPC), are another family of
ceramic absorbers. Instead of having an array of thin parallel tubes, RPC are composed by open cells
(with various dimensions) randomly distributed, which gives it a foam shape as seen in Figure 2.7(a). To
build a SiC RPC, the polymer foam replication is the most common method applied [41, 49].
These open-cell foams were mostly employed in projects with closed loop power plants. Particularly,
DIAPR (Directly Irradiated Annular Pressurized Receiver) is the benchmark absorber for closed loop
systems with a OCF, which was designed by the Weizmann Institute of Science (WIS) and Rotem Indus-
tries [10]. Figure 2.7(b) presents a scheme of the implemented RPC receiver on the DIAPR 30-50. In
this design, the Pythagoras alumina–silica OCF is arranged in a configuration called the Porcupine. This
absorber arrangement has been tested individually and within the DIAPR in a broad range of tempera-
tures and irradiances, which the absorber endured without presenting any sort of damage. In addition,
DIAPR 30-50 contains a fused-silica window that withstood pressures above 50 bar. Tests also revealed
that this window isn’t sensitive to contamination agents (such as dirt), since it hadn’t generated hot
spots on this component. Lastly, a compound parabolic concentrator (CPC) was also been included in
the DIAPR design as a secondary concentrator [10, 48].
From 1994 to 1996, the DIAPR was tested for power inputs between 30 KWth to 50 KWth. For solar
irradiances between 3600 kW.m−2 and 5300 kW.m−2 and for air pressures from 17 bar to 20 bar, it was
attained an outlet temperature as high as 1200 oC, which was the largest temperature ever recorded
for a VA at the time. Under these conditions, absorber efficiencies from 70 % to 80 % were reported as
well[10, 48].
Due to the accomplishments of the DIAPR 30-50, some design improvements were suggested to
optimize the absorber performance. As a result, the DIAPR multistage concept was proposed in 1996 to
reduce thermal losses. The new DIAPR design took into account the incident irradiance distribution, by
(a) Open-cell foam sample. (b) DIAPR 30-50 receiver scheme.
Figure 2.7: Open-cell foam (from [47]) and its application on the DIAPR 30-50 project (adaptedfrom [48]).
15
installing preheaters around the main high temperature module. In addition, the main receiver (presented
in Figure 2.7(b)) has suffered some modifications. One significant change was the inclusion of a hot air
inlet that comes from the preheaters, which allowed the air to be gradually heated in two stages [10,
34, 48]. Tests in 1998 have estimated a solar flux between 2500 kW.m−2 and 4000 kW.m−2 in the main
receiver module, as well as predicted an incident irradiance between 850 kW.m−2 and 1400 kW.m−2
in the preheaters. A maximum outlet temperature of 1000 oC was achieved, which has supplied power
between 30 kWth to 60 kWth, with operating pressures from 16 bar to 19 bar. Additionally, the preheaters
delivered air to the main module, with temperatures ranging from 650 oC to 750 oC. Unfortunately, the
receiver efficiency wasn’t determined for this project, since the inlet flux detector wasn’t functional during
these tests [10, 34].
Reticulated porous ceramics were also used in the SOLGATE pilot plant. In this project, a solar-
hybrid power system was built in order to drive a gas turbine in a Brayton cycle. To that end, the
REFOS design (seen in Figure 2.5(b)) was selected as the solar receiver to heat the pressurized air.
However, the second phase of the SOLGATE project required outlet air temperatures of 1000 oC, which
was unattainable with the REFOS metallic receiver. Therefore, that component was replaced by a SiC
OCF absorber in order to achieve the demanded high temperatures. In addition, a window cooling
technology had to be implemented to prevent glass overheating. Tests results in 2003 showed, for a
770 W.m−2 DNI, a 960 oC outlet air temperature and an absolute operating pressure of 5 bar. In these
conditions, efficiencies ranging from 75 % to 90 % were recorded, which leads to a solar fraction of 70 %
in the combustion chamber of the solar-hybrid power plant [10, 39].
Open-cell foam applications were performed at laboratory scale as well. Fend and Pitz-Paal et al.
have tested small RPC and MH samples in a open loop systems, to assess their thermal perfor-
mance [31, 42, 50]. Results have shown that MH receivers can reach higher outlet air temperatures
than OCF absorbers under the same conditions, and thus larger efficiencies are attained for the former
family of ceramic absorbers [41]. Additionally, pressure drops in RPCs have been studied in recent
works [49, 51].
2.3 Radiative models
In volumetric receivers, the incident radiation is absorbed throughout the absorber [31]. The ra-
diative intensity along a porous medium can be described by the radiative transfer equation (RTE). If
one chooses to simulate the thermal behaviour of VAs, the RTE must be included in the overall energy
conservation calculation. Nowadays, solving the analytical RTE for the stated purpose is still a time-
consuming process, even if supercomputers are used [52]. Instead of taking an analytical approach,
works in literature use radiative models to approximate the RTE to a more computable expression.
Avila-Marin et al. [16] has grouped the radiative methods (used to simulate the VA performance)
in several categories, as shown in Figure 2.8. Regarding porous mediums, two simulations types are
considered: one-dimensional models and Computational Fluid Dynamics (CFD) models. The former
model only analyses property variations along the flow direction, which greatly reduces the computation
16
time. On the other hand, CFD models evaluates those property changes on a two-dimensional or three-
dimensional scale, where the flow direction is included. Information about the receiver’s cross section
behaviour is only given by the latter representation, so CFD models are more used than one-dimensional
models [16].
Regarding radiative methods, many have been used in the literature to describe the radiative propa-
gation in open-cell foams. Mey et al. [53] has compared the Rosseland approximation, P1 method and
the two-flux approximation with the Monte Carlo method, which is considered to be the most accurate
approach. In an one-dimensional representation, the two-flux approximation yielded the best results.
Kribus et al. [54] has chosen this approximation to perform a parametric study for OCF absorbers. In the
following year, the same authors have extended this study by using the Discrete ordinate method (DOM),
which is a generalization of the two-flux approximation [55]. Finally, Wu et al. [56] has also performed a
similar investigation on a two-dimensional model, using the P1 method. With the same simulation model
and radiative method, a transient analysis was accomplished as well [57].
Figure 2.8: Radiative transfer equation methods (from [16]).
17
Chapter 3
Heat transfer model
In this chapter, the one dimensional transient heat transfer model of an open-cell foam absorber is
presented. One starts by describing the energy equations along the absorber, together with its boundary
conditions. The thermal evolution in time of the system is obtained after the discretization of the model.
Moreover, the general RTE is attained step by step, as well as the two-flux approximation that is applied
to the model. The remaining related thermal parameters of the receiver and fluid are included, and a brief
explanation on physical instabilities is incorporated. Finally, the implemented algorithm is presented,
followed by an analysis on the treatment of the absorber’s performance.
3.1 Geometry of an OCF absorber
The proposed model describes the thermal behaviour of an open-cell foam, which has a porous
structure. Therefore, two different media must be considered: the solid structure itself (SiC) and the fluid
inside the receiver (air) that flows at a given velocity v imposed by a blower placed after the receiver’s
outlet. The porosity ϕ defines the fraction of the absorber’s volume that is occupied by void space [58].
In order to simplify the model, the receiver is defined as a rectangular prism with cross section A
and length Lr, as shown in Figure 3.1. This absorber can be divided in n + 1 nodes, where the nodes
0 and n are at the boundaries of the receiver, while the remaining are located inside the absorber. As
suggested in Figure 3.1, each node is separated by ∆x = Lrn . Note that the presented dimensions of
this figure are distorted, since the length Lr (x direction) is typically much smaller than the cross section
dimensions (y and z directions).
Figure 3.1: OCF Receiver schematic.
19
3.2 Transient heat transfer equations
In the absence of local thermal equilibrium, the general heat transfer equations are presented for the
solid and fluid phase respectively [52, 58]:
(1− ϕ)(ρcP )s∂Ts∂t
= (1− ϕ)∇.(ks∇Ts)−∇qR,s + h(Tf − Ts) , (3.1a)
ϕ(ρcP )f∂Tf∂t
+ (ρcP )fvf .∇Tf = ϕ∇.(kf∇Tf )−∇qR,f + h(Ts − Tf ) , (3.1b)
where ρ is the density of the substance, cP the specific heat at constant pressure, T the temperature, k
the thermal conductivity, qR the radiative heat flux and h the volumetric heat transfer coefficient. Note
that ∂∂t is the partial derivative in respect to time and ∇ the divergence (or the gradient). The indexes s
and f refer to the solid and fluid media respectively.
The right side of expression (3.1a) contains a conduction term, a radiative term and a convective
term, respectively. On the left side, it only has an energy storage term. In expression (3.1b), the same
terms are applied for the fluid. However, it appears an additional term in the left side, which represents
the advection imposed to the fluid.
Some assumptions are made in the model. First, one considers that the foam is homogeneous
throughout the absorber. Furthermore, the y and z directions are considered to be much larger when
compared to the x direction. Hence, the heat transfer model can have a unidimensional representation
in the direction of the fluid flow (x direction). The radiative heat transfer in the fluid phase has a negli-
gible effect, so it can be discarded from equation (3.1b). Finally, the gradient of thermal conductivity is
considered to be negligible in the fluid phase and null in the solid phase. Therefore, the partial derivative
will only affect the temperature gradient. With this proposed assumptions, the general equations are
rewritten:
(1− ϕ)(ρcP )s∂Ts∂t
= (1− ϕ)ks∂2Ts∂x2
− dqRdx
+ h(Tf − Ts) , (3.2a)
ϕ(ρcP )f∂Tf∂t
+ (ρcP )fuf∂Tf∂x
= ϕkf∂2Tf∂x2
+ h(Ts − Tf ) . (3.2b)
where qR and uf are respectively the radiative heat flux and the fluid velocity in the x direction. Ex-
pression (3.2) determines the solid and fluid thermal behaviour at a given time and space. Since these
equations have partial derivatives, one way to solve them is by using the finite difference method.
The purpose of the finite differences is to approximate a given partial derivative to a determined
finite-divided-difference formula. These formulas are derived from the Taylor series expansion, and it’s
accuracy increases with the number of terms used. There are 3 types of approximations available: a
forward difference, a backward difference and a centered difference. Appendix A shows two formulas
for each approximation type (the simplest and one with an additional term of the Taylor series) and for
each derivative order. Note that, for the same order and formula, the centered difference has double the
20
order of accuracy of the forward and backward differences [59].
The finite difference method is used in the model to solve partial derivatives in time and space.
Therefore, one needs to choose between an explicit or implicit form. In the explicit form, an unknown
temperature of a node at a time t+∆t is determined only by known nodal temperatures at the preceding
time t. On the other hand, in the implicit form, an unknown temperature at a time t + ∆t is partially (or
fully) determined by nodal temperatures at a time t + ∆t. The latter form requires that the equations of
all nodes at a time t+ ∆t are solved simultaneously, while the former doesn’t [19].
The explicit form was first considered for this model, due to its simplicity. However, this method was
soon discarded, because of the stability criteria. For a given ∆x, there is a maximum time step (∆t)
that guarantees stability, and it must hold for all nodes [19]. There are some methods to find a stability
condition, such as the Von Neumann analysis, but they only work for linear systems [60, 61]. Due
to the emission terms present in the model, linearity can’t be achieved, so those methods are invalid.
Since methods to non-linear systems are too complex to apply, the explicit form becomes unreliable [18].
Therefore, the implicit finite difference method is applied to this model. Although it’s more complex to
solve the system of equations, this form is unconditionally stable numerically, which means that ∆x and
∆t can be chosen independently [19]. However, one must note that physical instabilities may occur if
the Courant-Friedrichs-Lewy condition is ignored [20].
3.2.1 Internal nodes
The presented model applies the implicit finite differences method. As seen in Figure 3.1, the re-
ceiver is divided in discrete nodes (separated by a distance ∆x between them), where the thermal
properties are calculated. Time also has discrete values, which means that the properties of each node
are computed in a finite number of time instants (that are separated between each other by a time step
∆t).
Equation (3.2) is assigned to each internal node (from 1 to n − 1), and using the finite differences,
the temperature of the respective node can be predicted for the next time instant. To do so, one must
first choose which approximation to use for each partial derivative:
• The energy storage terms contain a partial derivative with respect to time. As previously stated,
the equations intend to determine the temperature of the node at the next time step. Therefore, the
appropriate approximation for this partial derivative is the forward finite-divided-difference formula;
• The conduction terms have a spacial second order partial derivative. Conduction implies an in-
teraction between adjacent nodes, so the applied approximation is the centered finite-divided-
difference formula;
• The term with the imposed advection contains a spacial first order partial derivative. In this phe-
nomena, the fluid is carried at a velocity u, and flows from node ’0’ to node ’n’. The air that reaches
a node at each time step comes from a previous node, so the backward finite-divided-difference
formula should be used to approximate this partial derivative.
21
The finite differences can now be applied to expression (3.2) using the proper approximations:
(1− ϕ)(ρcP )s,iT t+1s,i − T ts,i
∆t= (1− ϕ)ks,i
T t+1s,i+1 − 2T t+1
s,i + T t+1s,i−1
∆x2− dqR
dx
∣∣∣∣i
+ hi(T tf,i − T ts,i
), (3.3a)
ϕ(ρcP )f,iT t+1f,i − T tf,i
∆t+ (ρcP )f,iuf,i
T t+1f,i − T
t+1f,i−1
∆x= ϕkf,i
T t+1f,i+1 − 2T t+1
f,i + T t+1f,i−1
∆x2+ hi
(T ts,i − T tf,i
),
(3.3b)
where the subscript i denotes the number of the current node in study and the superscript t the given
time instant. After manipulating expression (3.3), one gets:
−λs,iT t+1s,i−1 +
(1 + 2λs,i
)T t+1s,i −λs,iT
t+1s,i+1 =
(1−λs,i
φs,i1− ϕ
)T ts,i−λs,i
∆x2
(1− ϕ)ks,i
dqRdx
∣∣∣∣i
+λs,iφs,i
1− ϕT tf,i .
(3.4a)
−(λf,i+uf,i
∆t
ϕ∆x
)T t+1f,i−1 +
(1+2λf,i+uf,i
∆t
ϕ∆x
)T t+1f,i −λf,iT
t+1f,i+1 =
(1−λf,i
φf,iϕ
)T tf,i+λf,i
φf,iϕT ts,i .
(3.4b)
where λ and φ are the Fourier and Biot numbers, respectively. The calculations can be seen step-by-step
in Appendix B.1.
3.2.2 Boundary conditions
In order to close the sets of equations, one must find the appropriate energy balance for each bound-
ary of the receiver. Starting with the entrance, the energy balance of the solid phase is shown in expres-
sion (3.5):
(1− ϕ)αGinc − (1− ϕ)qrad,in − (1− ϕ)qconv,in = −(1− ϕ)ks,0∂T
∂x
∣∣∣∣0
, (3.5)
where α is the absorptance of the receiver and Ginc the incident irradiance on the aperture. qconv,in and
qrad,in are, respectively, the radiative and convective heat fluxes at the receiver’s inlet surface:
qconv,in = hin(T t+1s,0 − T
t+1f,in
), (3.6a)
qrad,in = εσ
[(T ts,0
)4 − (T tsky)4] , (3.6b)
where hin is the inlet convective heat transfer coefficient, Tf,in the temperature of the inlet air close to
the receiver’s surface, ε is the receiver’s emittance, σ the Stefan-Boltzmann constant and Tsky the sky
temperature [19, 62]. Note that in expression (3.6b) the sky is considered to be a blackbody [62].
The left side of equation (3.5) includes an irradiance absorption term, as well as a radiative and a
convective loss terms. The right side includes only a conduction term. The conduction at the boundary
22
occurs in one side only, since there is no solid material at the other one.
Regarding the fluid phase, the energy balance at the receiver’s inlet is given by expression (3.7):
− (cP )f,0 m(T t+1f,0 − T
t+1f,in
)−Aϕkf,0
∂T
∂x
∣∣∣∣0−
= −Aϕkf,0∂T
∂x
∣∣∣∣0+
, (3.7)
where m is the mass flow rate of the fluid [63]:
m = ρf,iAuf,i . (3.8)
The left side of equation (3.7) includes an advection term and a conduction term, and the right side
a conduction term as well. Note that, unlike the solid phase, conduction at the boundary in the fluid
phase occurs in both sides (just like in interior nodes). This happens because air exists not only inside
the material but outside too.
One can now look at the receiver’s outlet. The energy balance of the solid phase is shown in expres-
sion (3.9):
− (1− ϕ)ks,n∂T
∂x
∣∣∣∣n
= (1− ϕ)qconv,out + (1− ϕ)qrad,out . (3.9)
As before, qconv,out and qrad,out are respectively the convective and radiative heat fluxes at the receiver’s
outlet surface:
qconv,out = hout(T t+1s,n − T t+1
f,out
), (3.10a)
qrad,out = εσ
[(T ts,n
)4 − (T t+1w
)4], (3.10b)
where hout is the outlet convective heat transfer coefficient, Tf,out the temperature of the outlet air and
Tw the temperature of the wall in front of the receiver’s outlet surface.
Similar to the entrance, the right side of equation (3.9) contains a radiative and convective loss terms,
and the left side a conduction term. Note that there isn’t irradiance coming out of the receiver outlet,
since it is considered to extinguish before reaching the end of the absorber.
Finally, expression (3.11) gives the energy balance of the fluid phase at the receiver’s outlet. This
equation contains an advection term and a conduction term on both sides of the equation:
− (cP )f,n m(T t+1f,n − T
t+1f,n−1
)−Aϕkf,n
∂T
∂x
∣∣∣∣n−
= −(cP )f,nm(T t+1f,out − T
t+1f,n
)−Aϕkf,n
∂T
∂x
∣∣∣∣n+
. (3.11)
The computation of all the boundary conditions can be seen in Appendix B.2.
23
3.2.3 Thermal evolution of the receiver
The receiver’s temperature at the next time step can be calculated using both systems of equations
formed, where each phase has an associated matrix. Since the matrices are tridiagonal, they can be
solved using the Thomas algorithm (see Appendix C) [59, 60, 64].
The expressions (3.4a), (B.7) and (B.13) found for the solid phase form the matrix equation (3.12):
hin +ks,0∆x −ks,0∆x 0
−λs,1 1 + 2λs,1 −λs,1 0
. . . . . . . . .
0 −λs,n−1 1 + 2λs,n−1 −λs,n−1
0 −ks,n∆xks,n∆x + hout
·
T t+1s,0
T t+1s,1
...
T t+1s,n−1
T t+1s,n
=
S0
S1
...
Sn−1
Sn
, (3.12)
where
S0 = αGinc + hinTt+1f,in + εσ
[(T tsky
)4 − (T ts,0)4] , (3.13a)
Si =
(1− λs,i
φs,i1− ϕ
)T ts,i − λs,i
∆x2
(1− ϕ)ks,i
dqRdx
∣∣∣∣i
+ λs,iφs,i
1− ϕT tf,i 1, 2, . . . , n− 1 , (3.13b)
Sn = houtTtf,n + εσ
[(T tf,n
)4 − (T ts,n)4] . (3.13c)
The expressions (3.4b), (B.10) and (B.16) found for the fluid phase form the matrix equation (3.14):
2FA,0 − ϕkf,0∆x −ϕkf,0∆x 0
−FB,1 FC,1 −λf,1 0
. . . . . . . . .
0 −FB,n−1 FC,n−1 −λf,n−1
0 −FA,n 2FA,n
·
T t+1f,0
T t+1f,1
...
T t+1f,n−1
T t+1f,n
=
F0
F1
...
Fn−1
Fn
, (3.14)
where
FA,i = (ρcP )f,iuf,i +ϕkf,i∆x
, (3.15a)
FB,i = λf,i + uf,i∆t
ϕ∆x, (3.15b)
FC,i = 1 + 2λf,i + uf,i∆t
ϕ∆x, (3.15c)
F0 = FA,0Tt+1f,in , (3.15d)
Fi =
(1− λf,i
φf,iϕ
)T tf,i + λf,i
φf,iϕT ts,i 1, 2, . . . , n− 1 , (3.15e)
Fn = FA,nTtf,n . (3.15f)
24
3.3 Radiative heat transfer
In equation (3.2a), the radiative term (−dqRdx ) represents the amount of radiative energy that is stored
inside an unit volume of the receiver. Due to its importance in the heat transfer model, a detailed
explanation on how to attain this quantity is presented. Note that this demonstration is mostly based
on the information gathered by Modest [52]. Nevertheless, some approximations were added to this
section. All of the following quantities are considered to be independent of the radiation wavelength.
Additionally, the porous medium is considered to be a gray body, meaning that α = ε.
3.3.1 Radiative transfer equation
Consider a light beam (with intensity of radiation I) travelling at a given direction s through a partici-
pating medium. This beam will gain and lose energy along its trajectory, at a rate given by the medium’s
properties.
Starting with the losses, there are two phenomena that diminish the intensity of the beam. One of
them is the absorption of radiation, where a fraction of the beam’s energy is absorbed by the medium
itself. The magnitude of absorbed energy, (dI)abs, is proportional to the distance travelled by the beam
(ds) as well as to its incident intensity of radiation:
(dI)abs = −κIds , (3.16)
where κ is the absorption coefficient. The other effect is the out-scattering of radiation. This occurs
when the medium can redirect a portion of the incident radiation to another direction sk, thus diminishing
the intensity along s. The magnitude of scattered intensity, (dI)abs, is proportional to the same quantities
as equation (3.16):
(dI)sca = −σscIds , (3.17)
where σsc is the scattering coefficient. Note that the coefficients in equations (3.16) and (3.17) define
the extinction coefficient and optical distance (β and τ , respectively):
β = κ+ σsc . (3.18a)
τ =
∫ s
0
βds . (3.18b)
As for the gains, there are two factors that increase the intensity of the beam. The first one is when
the beam can receive energy due to emission from the participating medium. If local thermal equilibrium
is assumed, the emitted intensity is equal to the blackbody intensity (Ib). The emitted intensity per unit
length to the beam is given by expression (3.19):
(dI)emi = κIbds . (3.19)
25
The second effect is the in-scattering of radiation. In this case, the medium is able to redirect a portion
of radiation from sk to s, which is the opposite effect of out-scattering. The magnitude of in-scattering
intensity is shown in equation (3.20):
(dI)sca(s) = dsσsc4π
∫4π
I(sk)Φ(sk, s)dΩk , (3.20)
where Ω is the solid angle and Φ the scattering phase function, that is, the probability of a ray deflect
from sk to s. If the amount of energy is equally scattered to all directions, one gets isotropic scattering
(given by Φ = 1 in equation (3.20)).
Using the effects from equations (3.16), (3.17), (3.19) and (3.20), an expression for the radiative
energy balance can be assembled:
I(s+ ds, s, t+ dt)− I(s, s, t) = κIb(s, t)ds− κI(s, s, t)ds− σscI(s, s, t)ds+ dsσsc4π
∫4π
I(sk)Φ(sk, s)dΩk ,
(3.21)
where:
I(s+ ds, s, t+ dt) = I(s, s, t) + dt∂I
∂t+ ds
∂I
∂s. (3.22)
Since the radiation travels at light speed (c), the relation between length and time is ds = cdt. If one
divides expression (3.21) by ds and insert the definition of β, one attains the radiative transfer equation:
1
c
∂I
∂t+∂I
∂s= κIb − βI +
σsc4π
∫4π
I(sk)Φ(sk, s)dΩk . (3.23)
Equation (3.23) can be simplified when one analyses the heat transfer model scope. Since light
speed is much larger when compared to the length and time scales under study, the partial derivative in
respect to time can be neglected. Therefore, expression (3.23) becomes:
dI
ds= s.∇I = κIb − βI +
σsc4π
∫4π
I(sk)Φ(sk, s)dΩk . (3.24)
Equation (3.24) can be also expressed as a function of optical distance:
dI
dτ= (1− ω)Ib − I +
ω
4π
∫4π
I(sk)Φ(sk, s)dΩk , (3.25)
where ω is the single scattering albedo:
ω =σscβ. (3.26)
3.3.2 Two-flux approximation
With the radiative transfer equations deduced, one must find a unidimensional model that solves it.
One starts by assuming the simplest case, which is only solving the propagation of the diffuse radiation
26
in the participating medium.
The model is assembled assuming that the participating medium is bounded by two parallel plates,
as shown in Figure 3.2. The plates A1 and A2 are located, respectively, at τ = 0 and τ = τL, and are
assumed to be isothermal and isotropic. Additionally, it is assumed that the intensity of radiation that
leaves each plate is directionally dependent, but only on θ. Finally, the properties of the medium can
only change along the τ coordinate. With all these considerations, and using the geometric relation
τs = τcos θ , expression (3.25) becomes:
dI
dτs= cos θ
dI
dτ= (1− ω)Ib − I +
ω
2
∫ π
0
I(τ, θk)Φ(θ, θk) sin θkdθk , (3.27)
which can be rewritten as a function of the cosine (µθ = cos θ), so that:
µθdI
dτ= (1− ω)Ib − I +
ω
2
∫ 1
−1
I(τ, µθ,k)Φ(µθ, µθ,k)dµθ,k . (3.28)
The intensities I+ and I− are also shown in Figure 3.2 (a) and (b) respectively, and are applied at any
point of the medium. I+ is limited to all directions available for A1 (that is, is limited to 0 < µθ < 1).
Analogously, I− is limited to all directions available for A2 (in other words, is limited to −1 < µθ < 0).
Based on these quantities, one unidimensional radiative model that can be employed is the Schuster-
Schwarzschild approximation, (also known as the two-flux approximation), which accurately describes
the propagation of diffuse radiation in the participating medium [53]. This model divides the intensity
of radiation in two hemispheres: an upper (or forward), I+; and a lower (or backward), I−. In each
point, the intensity of radiation within each hemisphere is isotropic. However, the intensity between
hemispheres at the same point may differ. The intensity of radiation in the two-flux approximation is then
given by expression (3.29):
I(τ, µθ) =
I−(τ), −1 < µθ < 0
I+(τ), 0 < µθ < 1
. (3.29)
Using equation (3.29), and assuming isotropic scattering (Φ = 1), expression (3.28) simplifies to:
Figure 3.2: Intensity of radiation between parallel plates on the upward direction (a) and downwarddirection (b) (from [52]).
27
µθdI
dτ= (1− ω)Ib(τ)− I(τ, µθ) +
ω
2(I+(τ) + I−(τ)) . (3.30)
Note that, in a porous medium, the internal emission propagates only in the void space (and not through
the solid). Therefore, the blackbody emission term in equation (3.30) must be multiplied by the poros-
ity [65]. By also applying the substitution dτ = βdx, one gets expression (3.31):
µθdI
dx= β(1− ω)ϕIb(x)− βI(x, µθ) + β
ω
2(I+(x) + I−(x)) . (3.31)
Integrating equation (3.31) in each hemisphere gives expression (3.32):
1
2
dI+
dx= β(1− ω)ϕIb − βI+ + β
ω
2(I+ + I−) , (3.32a)
− 1
2
dI−
dx= β(1− ω)ϕIb − βI− + β
ω
2(I+ + I−) . (3.32b)
Two new quantities are now defined. The first is the diffuse incident radiation (Gd), which represents
the total intensity of diffuse radiation on all directions at a given point. The second quantity is the diffuse
radiative heat flux (qd), which is the balance between incident and outgoing diffuse intensity at a given
point. Since intensity I has only dependence in θ, and using equation (3.29), one gets expression (3.33):
Gd(x) = 2π
∫ π
0
I(x, θ) sin θdθ = 2π
∫ 1
−1
I(x, µθ)dµθ = 2π(I+(x) + I−(x)) . (3.33a)
qd(x) = 2π
∫ π
0
I(x, θ) cos θ sin θdθ = 2π
∫ 1
−1
I(x, µθ)µθdµθ = π(I+(x)− I−(x)) . (3.33b)
With equations (3.32) and (3.33), one finally arrives to a set of differential equations, which are only
dependent on Gd and qd:
dGddx
= −4βqd , (3.34a)
dqddx
= κ(4ϕπIb −Gd) . (3.34b)
Expression (3.34) solves the diffuse radiative transfer inside the receiver. In order to close the system
of equations, one must find a set of boundary conditions to the exterior surfaces. A possible solution is
to attain an expression for I+(x) and I−(x) by manipulating equation (3.33):
I+(x) =Gd(x) + 2qd(x)
4π, (3.35a)
I−(x) =Gd(x)− 2qd(x)
4π. (3.35b)
28
By substituting I+(0) and I−(Lr) for a radiosity J in expressions (3.35a) and (3.35b) respectively, one
obtains the boundary conditions for the two-flux approximation:
I+(0) =J1
π↔ Gd(0) + 2qd(0) = 4J1 , (3.36a)
I−(Lr) =J2
π↔ Gd(Lr)− 2qd(Lr) = 4J2 . (3.36b)
3.3.3 Collimated radiation
The analysis in the previous section only considers the presence of diffuse radiation. However, in a
volumetric receiver, one must also account the collimated radiation.
As stated before, the beam solar radiation is focused on the absorber by a field of heliostats. Due
to this process, solar radiation reaches the receiver as collimated radiation. Since the absorber has a
porous structure, some of the collimated radiation propagates through the participating medium. This
kind of behaviour is unique among solar tower receivers, because collimated energy is partially absorbed
inside the receiver (in opposition to be fully absorbed at the inlet wall) [10, 21, 23].
The inclusion of the collimated term in the two-flux approximation makes the radiative model closer
to reality. Figure 3.3 demonstrates the importance of this parameter. In this figure, the incident radiation
is plotted along the receiver for several models, in a set of conditions where the scattering effect is the
dominant phenomenon. It is assumed that the curve closer to reality is the Monte-Carlo simulation, due
to the experimental results of Mey et al. [53]. With that in mind, one can observe that the collimated
two-flux approximation follows the behaviour of Monte-Carlo simulation, while the diffused two-flux ap-
proximation curve doesn’t. The collimated radiation has a strong impact on the model, and thus it should
be added in the radiative model.
Figure 3.3: Incident radiation on several models in a cold medium simulation (from [55]).
29
The diffuse and collimated radiation must be treated separately [55, 57]. Starting with the latter, its
intensity decreases along its trajectory, as described by equation (3.37). Just like the internal emission,
collimated radiation can only propagate in the void space, so it must be multiplied by ϕ.
Gc(x) = ϕ(1− ζ)Gince−βx , (3.37)
where ζ is the dispersion ratio, which is the fraction of the collimated radiation that is scattered at the
receiver’s entrance. Note that in a real solar tower system, the field of heliostats is usually surrounding
the receiver (at an angle θhel). In this scenario, the radiation that reaches the receiver is a spread
of rays (with angle θhel), as opposed to an array of parallel rays [66]. For this reason, ζ is placed in
equation (3.37) as a correction factor, so that one can treat the incoming radiation as collimated.
Regarding the diffuse radiation, one must add in expression (3.31) the contribution of the collimated
radiation that is scattered along its trajectory [55]:
µθdI
dx= β(1− ω)ϕIb − βI + β
ω
2(I+ + I−) +
βωϕ(1− ζ)Gince−βx
4π. (3.38)
By propagating the effect of the collimated term in expression (3.38), one arrives at a new set of differ-
ential equations:
dGddx
= −4βqd , (3.39a)
dqddx
= κ(4ϕπIb −Gd) + σscϕ(1− ζ)Gince−βx . (3.39b)
Note that equation (3.39a) hasn’t suffered any changes when compared to expression (3.34), while
equation (3.39b) has a new collimated component. However, the boundary conditions remain the same,
since they were deduced using only the definitions of Gd and qd.
3.3.4 Application of the two-flux approximation
The two-flux approximation is now applied to the heat transfer model, in order to find an expression fordqRdx
∣∣i. Before proceeding, one has to compute the diffuse radiation along the receiver. Some quantities
still need to be defined, mainly the blackbody radiation (Ib) and the radiosities J1 and J2, which are
presented in expression (3.40):
Ib =σT 4
s
π, (3.40a)
J1 = ϕ
[ζGinc + εσ
(T ts,0
)4], (3.40b)
J2 = ϕεσ(T ts,n
)4. (3.40c)
30
Notice that the fraction of collimated radiation scattered at the entrance becomes diffuse, so it is added
to equation (3.40b). Note also that some effects were neglected from the radiosities, particularly the
backscattering in J2. It was assumed that those had a smaller contribution to the radiosity in comparison
to the collimated and emitted radiation.
The diffuse radiation can now be attained. To do so, equation (3.39b) is differentiated in order to x:
d2Gddx2
= −4βdqddx
= −4βκ(4ϕπIb −Gd)− 4βσscϕ(1− ζ)Gince−βx , (3.41)
where dqdx was replaced by expression (3.39b). Equation (3.41) can now be discretized (as seen in
Figure 3.1), so that the finite difference method can be performed. Just as before, the centered finite-
divided-difference formula is applied to the second order derivative:
Gd,i−1 − 2Gd,i +Gd,i+1
∆x2= −4βκ
[4ϕσ
(T ts,i)4 −Gd,i]− 4βσscGc,i , (3.42)
where Ib is substituted by equation (3.40a). Note that Gc,i is the collimated radiation at a node i:
Gc,i = ϕ(1− ζ)Gince−iβ∆x . (3.43)
Manipulating expression (3.42) leads to:
−Gd,i−1 + (2 + 4βκ∆x2)Gd,i −Gd,i+1 = 16βκ∆x2ϕσ(T ts,i)4
+ 4βσsc∆x2Gc,i . (3.44)
In order to close this set of equations, one can apply the boundary conditions previously defined
by expression (3.36). The auxiliary calculations performed for the boundary conditions can be seen in
Appendix B.3. With equations (3.44) and (B.21), it is possible to build a matrix to solve Gd:
1 + 2β∆x −1 0
−1 2 + 4βκ∆x2 −1 0
. . . . . . . . .
0 −1 2 + 4βκ∆x2 −1
0 −1 1 + 2β∆x
·
Gd,0
Gd,1...
Gd,n−1
Gd,n
=
R0
R1
...
Rn−1
Rn
, (3.45)
where
R0 = 8β∆xϕ
[ζGinc + εσ
(T ts,0
)4], (3.46a)
Ri = 16βκ∆x2ϕσ(T ts,i)4
+ 4βσsc∆x2Gc,i 1, 2, . . . , n− 1 , (3.46b)
Rn = 8β∆xϕεσ(T ts,n
)4. (3.46c)
The matrix equation (3.45) is also tridiagonal, thus Gd can be computed using Thomas algorithm (see
Appendix C) [59, 60, 64]. With the diffuse incident radiation discovered, one can finally attain an expres-
31
sion for the radiative energy that is stored inside an unit volume in each node [57]:
dqRdx
∣∣∣∣i
= κ[4ϕσ
(T ts,i)4 −Gi] , (3.47)
where G is the total incident radiation:
Gi = Gc,i +Gd,i . (3.48)
3.4 Parameters of the model
3.4.1 Convective heat transfer coefficient
Equation (3.1) contains a volumetric heat transfer coefficient, which represents the amount of heat
exchanged between the solid and fluid phases through convection (in an unit volume). In the literature,
this quantity is generally given by the following expression [58, 68]:
hi = hsf,iαsf , (3.49)
where hsf is the interfacial convective heat transfer coefficient and αsf the interfacial area per unit
volume. Starting with the αsf , it is commonly assumed that the solid phase of the receiver is formed by
spherical particles. In this condition, it is considered that the receiver is formed by particles packed in a
bed [58, 67]. As seen in Figure 3.4, three different cell geometries can be attained: a simple cubic (SC),
a body centered cubic (BCC) and a face centered cubic (FCC). The particle’s area and volume (Ap and
Vp respectively) are given by expression (3.50), where the values of j depend on the cell geometry (as
shown in Table 3.1) [67].
Ap = j4π
(dp2
)2
, (3.50a)
Vp = j4
3π
(dp2
)3
. (3.50b)
With equation (3.50), one can attain the specific surface area of the particles [67]:
αp =
(ApVp
)cell
=6
dp, (3.51)
Figure 3.4: Cell geometry (adapted from [67]).
32
Table 3.1: Values of j for each cell geometry.
Cell Geometry SC BCC FCCj 1 2 4
where the subscript cell states that the values are calculated within a cell. However, it is different from
αsf , which is the surface area presented to the fluid when the particles are packed in a bed. To attain
this parameter, the following correction must be performed [68, 69]:
αsf = (1− ϕ)αp . (3.52)
With this quantity attained, one can also define the hydraulic diameter of a pore [63, 67, 69]:
dh =4
αp
ϕ
(1− ϕ)=
4ϕ
αsf. (3.53)
Regarding the interfacial heat transfer coefficient, it can be determined from the definition of the
Nusselt number (Nu) [67]:
hsf,i =kf,iNuv,i
dp. (3.54)
Note that in equation (3.54), dp is considered to be the characteristic length. To compute the Nusselt
number, it is necessary to apply an empirical correlation that suits the flow conditions. In packed beds,
it follows the formulation given by expression (3.55)[67]:
Nuv,i = o1 + o2Pro3i Reo4i
(dpdhϕ
)o4, (3.55)
where o1, o2, o3 and o4 are fitting parameters of the Nusselt correlation. Pr and Re are respectively the
Prandtl and Reynolds number:
Pri =(ρcP )f,iµf,iρf,ikf,i
, (3.56a)
Rei =uf,idhϕνf,i
, (3.56b)
where µ and ν are respectively the dynamic and kinematic viscosity.
Several solutions for the fitting parameters of equation (3.55) have been proposed [58]. One of the
most known correlations is the one presented by Wakao and Kaguei, which also has the advantage to
be independent of ϕ [58, 67, 70]. However, it doesn’t give a good correlation for low Reynolds numbers
(Re < 100) [70]. For this reason, the work of Kuwahara et al. is chosen for the model, since it yields a
good correlation for all Reynolds numbers and for a large range of porosities (0.2 < ϕ < 0.9) [71, 72]:
Nuv,i =
[2 +
12(1− ϕ)
ϕ
]+√
1− ϕ 3√
Pri
(dpdhϕRei
)0.6
, (3.57)
The volumetric heat transfer coefficient defined in equation (3.49) can be used in all of the nodes
33
inside the receiver. However, at its boundaries, one only considers the interactions on the surface (as
opposed to the volume as before). The heat transfer coefficient for the convection interactions between
the receiver and the exterior is presented in expression (3.58):
hsf,in(out) =kf,0(n)Nul,0(n)
dp, (3.58)
where the Nusselt number is given by equation (3.59):
Nul,0(n) = 2.0696ϕ0.38Re0.4380(n) . (3.59)
Note that expression (3.58) is limited to 0.66 < ϕ < 0.93 [73].
3.4.2 Optical correlations
In Section 3.3, the absorption, scattering and extinction coefficients are presented, as well as the sin-
gle scattering albedo. These optical properties depend on the absorptance, porosity and pore diameter,
which are inherent quantities of the absorber material.
For a silicon carbide absorber, there are many proposed correlations for these properties [74, 75].
However, Zhao and Tang state that these correlations have a limited range of application, due to the
limited experimental data in which they were based on. Using the Monte Carlo method, a more accurate
correlation for the extinction coefficient is then employed [76]:
β =12.64(1− ϕ)0.7
d0.79h
. (3.60)
If one considers that absorbers can be treated as opaque and diffuse structures, the absorption and
scattering coefficients are attained by expression (3.61) [41]:
κ = αβ , (3.61a)
σsc = (1− α)β . (3.61b)
Note that equation (3.61) respects the definition of β presented in Section 3.3. With equations (3.60)
and (3.61b), ω can be calculated from its definition (expression (3.26)).
3.4.3 Temperature correlations
The heat transfer model is expected to run on a wide range of temperatures (from ambient tempera-
ture to temperatures above 1000oC). Some parameters in this model are highly dependent on temper-
ature, and in the stated range, can even vary an order of magnitude. Therefore, a set of correlations to
this quantities must be defined to accommodate these temperature changes.
For the solid receiver, one knows that thermal conductivity, specific heat and density vary with the
temperature. However, this information can only be accessed by the manufacturers’ reports, and they
34
have only published a single value for each parameter. Therefore, it is assumed that these quantities
are independent of the temperature, and one will only use constant values that other articles have
considered.
For the fluid phase, multiple correlations can be found for dry air. Thermal conductivity and specific
heat are given by expression (3.62), which is valid in the range of 100 K < Tf < 1600 K [77, 78]:
kf,i(T tf,i)
= −3.9333×10−4 + 1.0184×10−4T tf,i−4.8574×10−8(T tf,i)2
+ 1.5207×10−11(T tf,i)3. (3.62a)
(cP )f,i(T tf,i
)= 1.0575×103−4.4890×10−1T t
f,i+1.1407×10−3(T tf,i
)2−7.9999×10−7(T tf,i
)3+1.9327×10−10(T t
f,i
)4.
(3.62b)
Air density is attained with equation (3.63), which is found by manipulating the ideal gas law:
ρf,i(T tf,i)
=pf,i
RfT tf,i, (3.63)
where pf is the fluid pressure and Rf = 287 J.kg−1.K−1 the specific gas constant [79]. The dynamic
viscosity is obtained by Sutherland’s equation:
µf,i(T tf,i)
=C1
(T tf,i) 3
2
T tf,i + C2, (3.64)
where C1 = 1.458× 10−6 kg.m−1.s−1.K−12 is a constant applied for air and C2 = 110.4 K the Sutherland
constant [80]. Notice that its temperature range of validity goes from 100 K to 1880 K [81].
With these quantities defined, one can also find an expression for the kinematic viscosity and for the
fluid velocity [81]:
νf,i =µf,iρf,i
, (3.65a)
uf,i =m
Aρf,i(3.65b)
3.4.4 Fluid pressure
In a porous structure, the mass flow rate is produced by a blower, which creates a pressure drop.
This pressure drop can be described by Darcy’s law with the Forchheimer extension:
− dpfdx
=µfK1
uf +ρfK2
u2f , (3.66)
where K1 and K2 are respectively the viscous and inertial permeability coefficients [32, 58]. Using
equations (3.63) and (3.65b), expression (3.66) becomes:
− pfdpf = RfTtf,i
[µf,iK1
m
A+
1
K2
(m
A
)2]dx . (3.67)
35
Integrating expression (3.67) leads to:
p2f,i−1 − p2
f,i
2= RfT
tf,i∆x
[µf,iK1
m
A+
1
K2
(m
A
)2], (3.68)
and rearranging equation (3.68), one gets the fluid pressure at each node:
pf,i =
√√√√p2f,i−1 − 2RfT tf,i∆x
[µf,iK1
m
A+
1
K2
(m
A
)2]. (3.69)
As for the boundary condition of expression (3.69), it is assumed that the fluid at node 0 is at atmospheric
pressure (pf,0 = 101325 Pa).
The only parameters left to be found are the permeability coefficients (K1 and K2), which depend
only on the absorber’s geometry [42]. To find these constants, an alternate expression of equation (3.66)
was derived for the unidimensional flow:
− dpfdx
=E1(1− ϕ)2µf
ϕ3d2p
uf +E2(1− ϕ)ρf
ϕ3dpu2f , (3.70)
where E1 and E2 are shape factors [58, 82]. Note that if one considers E1 = 150 and E2 = 1.75,
expression (3.70) becomes the Ergun’s equation, which is widely used in the literature [49, 58, 83, 84].
The viscous and inertial permeability coefficients can be attained by comparing expression (3.66) with
equation (3.70):
K1 =ϕ3d2
p
E1(1− ϕ)2, (3.71a)
K2 =ϕ3dp
E2(1− ϕ). (3.71b)
There is an issue left to address, which is the possible flow instabilities that can happen in the ab-
sorber, due to the permeability coefficients. In a real absorber, these instabilities arise when a given
pressure drop has the ability to generate multiple mass flows. If that happens, different outlet tempera-
tures can happen for the same quadratic pressure difference. This means that the mass flow rate might
not be constant throughout the absorber, creating a wide temperature spread that can damage the re-
ceiver [32]. Although this issue doesn’t affect the performance of the unidimensional analysis, the model
must be assembled in order to avoid it. According to Becker et al. [32], flow instabilities only arise when:
K1
K2< 1.94× 10−6 m . (3.72)
Since K1 and K2 are only dependent on dp and ϕ, Figure 3.5 shows the minimum particle diameter
for every porosity value. Particle diameters below the minimum value create flow instabilities in a real
volumetric receiver.
36
Figure 3.5: Particle’s minimum diameter as a function of the porosity
3.5 Courant-Friedrichs-Lewy condition
The transient heat transfer model is composed by discrete nodes. In this model, each node can only
interact with adjacent ones, and air flows at each node with a certain velocity. When dealing with this
kind of problems, one must pay attention to the Courant-Friedrichs-Lewy condition (also known as the
CFL condition). This condition states that, for a certain spacing ∆x and velocity uf , there is a maximum
∆t that must be respected in order to prevent loss of information [18, 20]. The CFL condition is given by
expression (3.73):
C =uf∆t
∆x≤ Cmax ↔ ∆t ≤ ∆xCmax
uf(3.73)
where C is the Courant number. Since the CFL condition is a physical limitation, it also affects implicit
time-marching schemes, even though they are unconditionally stable (numerically). If ∆t is too high, ef-
fects that should be felt several nodes beyond are only transmitted to the adjacent node, which generates
loss of physical information and possibly creates instabilities [18].
The objective is to run the transient model at a maximum constant value of ∆t without generating
loss of physical information. However, the fluid velocity changes along the receiver, so one must assume
a critical maximum value for this parameter (uf,max), which diminish ∆t. According to equation (3.65b),
maximizing uf implies that one finds a minimum value for the fluid density (ρf,min). By testing different
mass flows in the range of 100 K < Tf < 1600 K, one concluded that ρf,min = 0.21 kg.m−3.
The time step can only be maximized by manipulating the maximum Courant number (in expres-
sion (3.73)) through a simple procedure. One starts by running the transient model at Cmax = 1, which
is the upper limit for the explicit time-marching scheme. If the model doesn’t become physically unstable,
one can try to increase the maximum Courant number. Otherwise, one must decrease the maximum
Courant number until instabilities disappear. For the heat transfer model, instabilities have appeared for
Cmax = 1, so the Courant number had to be decreased. After testing, one discovered that the maximum
time step is attained with Cmax = 0.1.
37
3.6 Algorithm
The heat transfer model is implemented using the following algorithm:
1. Define some initial conditions to the model:
(a) Absorber properties: Lr, A, ϕ, dp, α, ε, (cP )s, ks, ρs;
(b) External conditions: Ginc, m, ζ, Tamb, Tsky;
(c) Internal data: n, nt1, nt2, Cmax, ρf,min.
2. Compute constants of the model: ∆x, uf,max, ∆t, αp, αsf , dh, K1, K2 κ σsc, β, Gc,i.
3. Calculate Gc for every node.
4. Start the model at t = 0 s (with the time counters ct1 = ct2 = 0), where all the nodes of both media
are at ambient temperature(T ts(f) = Tamb
).
5. Compute Gd and dqRdx for every node.
6. Calculate temperature dependent parameters for every node using its respective correlation: (cP )f ,
kf , µf , pf , ρf , uf νf .
7. Determine the remaining parameters, which are dependent on the quantities computed in step 6:
Re, Pr, Nuv,l, hsf , h hin,out, λs,f , φs,f .
8. Calculate the temperature of every cell (in the solid and fluid phase) at the next time instant t+ 1,
using expressions (3.12) and (3.14) respectively.
9. Update the temperatures of time t (of each node in both media) with the ones attained in step 8(T ts(f) = T t+1
s(f)
).
10. Is ct1 = nt1?
(a) If so, print the solid and fluid temperatures (at nodes 0 and n) and ct2 and proceed to step 11.
(b) If not, increment ct1 and return to step 5.
11. Is ct2 = nt2?
(a) If so, go to step 12.
(b) If not, increment ct2, reset ct1 and return to step 5.
12. Print the solid and fluid temperatures for each node, as well as the radiative properties (Gc, Gd,dqRdx ) and fluid velocity.
The numerical implementation of this algorithm (in C) is presented in Appendix D.
38
3.7 Absorber performance
When the model simulates a long period of time, it will eventually arrive at thermal equilibrium. If this
state is reached, one can calculate the thermal efficiency of the absorber:
ηthermal =Qf
Qsup=m∫ Tf,outTf,in
(cP )fdT
AGinc, (3.74)
where Qsup is the power supplied to the absorber and Qf the power transferred to the fluid [54]. Note
that, at thermal equilibrium, Tf,out = Tf,n.
One can also predict the efficiency of the power block (ηcycle), and thus calculate the system overall
efficiency (ηsystem). For the former, one could assume the Carnot cycle, in which the power cycle
efficiency is maximized. However, the latter represents a limiting case where the heat sources are at
thermal equilibrium with the working fluid (in the power block). This means that heat is exchanged in
an infinitely slow process, which drives the generated electrical power to 0 W. Since the purpose of
the power block is to maximize electrical power, this process can’t be considered [85]. Instead, some
approaches to estimate ηcycle have been suggested in the literature, where its value is lower than in the
Carnot cycle but the extracted power is maximized. Nokinov [86] has derived an efficiency expression
which takes into account a temperature difference that occurs between the high temperature reservoir
and the working fluid. Moreover, Rebhan [85] has proposed an analysis which considers the influence
of friction losses on this efficiency. For this work, the approach done by Nokinov has been implemented,
and so ηcycle and ηsystem are given by expression (3.75):
ηcycle = 1−√TCTH
, (3.75a)
ηsystem = ηthermal ηcycle , (3.75b)
where TC and TH are the low and high reservoir temperatures respectively [86]. Note that either Nokinov
and Rebhan approaches yield a ηcycle close to what is observed in real power plants [85].
Finally, the receiver’s losses can be analysed as well. In this model, the loss factors are the remaining
fraction of supplied power that isn’t transferred to the fluid. Lossrad,in, Lossconv,in and Lossrefl,in are
respectively the radiative, convective and reflected loss factors at the receiver’s inlet. At the outlet,
the remaining collimated radiation that isn’t scattered or absorbed is also treated as a loss (Lossc,out).
Additionally, the radiative and convective losses at the outlet are considered (Lossrad+conv,out), even
though their contribution is small. The remaining fraction of power can be used to estimate the radiative
backscattering losses (Lossbacksc). Note that the latter source of losses has a fairly small accuracy, due
to the numerical calculation. All of the loss factors are attained by equation (3.76):
Loss1 = Lossrad,in =qrad,inGinc
, (3.76a)
Loss2 = Lossconv,in =qconv,inGinc
, (3.76b)
39
Loss3 = Lossrefl,in = (1− ϕ)(1− α) , (3.76c)
Loss4 = Lossc,out =Gc,nGinc
, (3.76d)
Loss5 = Lossrad+conv,out =qrad,out + qconv,out
Ginc, (3.76e)
Loss6 = Lossbacksc = 1− ηthermal −5∑i=1
Lossi . (3.76f)
40
Chapter 4
Simulation results
In this chapter, a full evaluation of the assembled heat transfer model is carried out. Simulations can
be divided in three categories. First, the model is validated against experimental data and compared
with another work in literature. In the second stage, the model is run without perturbations until a steady
state condition is attained. Results of the simulations are then presented for this state, where special
focus is given to the optimization of the receiver’s thermal performance. Note that for the performed
simulations, the steady state is always reached in less than two minutes. Finally, the absorber’s response
is evaluated for different perturbations of the incident irradiance. The purpose of these simulations is to
further investigate the transient effect on the receiver, mainly the equilibrium time and thermal stress.
4.1 Model validation
The heat transfer model must be validated with experimental data, in order to access its accuracy.
However, most of the experimental works presented in literature show little information about this kind
of receivers. Therefore, the experimental results given by Pitz-Paal et al. [42] are the only ones chosen,
since that work is the one that provides most information about the absorber parameters.
Table 4.1 shows the selected conditions necessary to run the model, in which most of the absorber’s
Table 4.1: Conditions for the model validation.
External conditions Absorber parameters Simulation options
Lr [m] 0.046
A [m2] 0.016742
ks [W.m−1.K−1] 12.5
Ginc [kW.m−2] 1300 (cP )s [J.kg−1.K−1] 1244 n 100
Tamb [K] 300 ρs [kg.m−3] 3210 nt1 15000
Tsky [K] 273.15 dp [mm] 0.351 nt2 1000
α 0.9
ϕ 0.782
ζ 0.1
41
characteristics were taken from the article of Pitz-Paal et al.. The only exceptions are the assumed
values of Tamb, Tsky, α and ζ, as well as the specific heat and density of the receiver, which are retrieved
from the work of Wu and Wang [57]. Since the results of Pitz-Paal et al. were gathered at thermal
equilibrium, the values of nt1 and nt2 are chosen so that the model reaches a steady state situation.
Two sets of simulations are performed, which differ from one another at the inlet fluid boundary
condition. In expression (3.5), there is a convection term that transfers energy between the absorber
inlet wall and the air outside the receiver. In equation (3.7), there isn’t a convection term, because one
considers that the energy transferred to the air outside the receiver is lost in that medium. However,
some authors state that the energy transfered to the air by convection is gained by the inlet air that
enters the receiver [54]. If that case is assumed, one must substitute equation (3.7) in the model by the
following expression:
A(1− ϕ)qconv,in − (cP )f,0 m(T t+1f,0 − T
t+1f,in
)−Aϕkf,0
∂T
∂x
∣∣∣∣0−
= −Aϕkf,0∂T
∂x
∣∣∣∣0+
, (4.1)
Both sets of simulations are compared with the work of Pitz-Paal et al. in Figure 4.1. Tables E.1 and E.2
(in Appendix E) present more information about these simulations.
In Figure 4.1(a), the outlet temperature in both simulations increases with the increase of the energy
per kilogram of air (Qsup/m), which is also predicted by Pitz-Paal et al. [42]. However, the increase of
temperature is higher for the simulation that uses equation (4.1), which also leads to higher efficiencies
(as seen in Figure 4.1(b)). It is possible to conclude from Table E.2 that the difference between efficien-
cies of both simulations (with the same Qsup/m) is mostly due to the convection, which is considered a
loss when using equation (3.7) and a gain when using equation (4.1).
One can now compare the simulations with the experimental results of Pitz-Paal et al. [42]. For low
Qsup/m, the experimental data is closer to the simulation that uses equation (4.1), but for high Qsup/m,
it approaches the simulation with equation (3.7). The results indicate that there might be two operating
regimes for the heat transfer. This can be explained by the effect of the wind speed and its relation with
the inlet velocity.
(a) Outlet air temperature (b) Thermal efficiency
Figure 4.1: Simulation results of the model compared with the experimental data of Pitz-Paal et al. [42].
42
To understand this effect, consider a mass of air placed close to the receiver’s aperture, that has
gained energy from the absorber (due to convection). Note that, in a real CRS, the absorber is placed
at the top of a tower, which is exposed to the wind. If the inlet velocity and wind speed have the same
order of magnitude, it is probable that the mass of air will enter the receiver, bringing in air warmed up
by the receiver aperture’s convective losses. In this case, the energy gained by the air will increase the
receiver efficiency. However, if the wind speed is much greater than the inlet velocity, the mass of air
will most likely be carried away by the wind. Unlike the other case, the energy transferred by convection
must be considered has a loss.
One can now look at the inlet velocities shown in Table E.1. The results with low Qsup/m have a
higher inlet velocity, so the energy transferred to the air is more likely to reenter the absorber. For this
case, expression (4.1) is more accurate. On the other hand, the results with high Qsup/m have a lower
inlet velocity. Therefore, the energy transferred to the air is probably lost, and equation (3.7) must be
considered.
One can then assume that the two points with lowest Qsup/m follow expression (4.1) and the re-
maining are given by equation (3.7). If this interpretation is followed, the maximum relative deviation for
the outlet temperature and thermal efficiency is 3.5 % and 5.4 % respectively. This maximum deviation
is given by the point with the highest Qsup/m. For the remaining results, the relative deviation of both
quantities never exceeds 2.5 %. One has concluded that the information about wind velocity (which is
absent from the experimental work) is highly relevant to validate the model.
Since error bars aren’t presented for the experimental data, one can’t quantitatively see if the simu-
lations are within the expected measurement error. However, it is known that the experimental incident
irradiance changes greatly. Pitz-Paal et al. [42] reported an average density flux of 1.3 MW.m−2 (which
has been used in the simulations), but also stated a peak density flux of 1.9 MW.m−2. This fluctua-
tion must likely generate a measurement error larger than the attained relative error in the simulations.
Therefore, one can consider that the simulation results follow the experimental data, and thus the model
is validated.
Note that for the remaining simulations of this chapter, equation (3.7) is always used as the fluid inlet
boundary condition, since the wind speed is considered to be much greater than the inlet velocity.
4.2 Model comparison with a similar work
With the heat transfer model validated, it can be compared against other simulations in literature.
The work of Kribus et al. is chosen for this comparison, since it also uses the two-flux approximation
as the mechanism for the radiative heat transfer [54]. Table 4.2 shows the conditions that Kribus et al.
imposed for this simulation, where the exceptions are applied for the same quantities stated previously.
Figure 4.2 compares the attained temperature along the receiver with the results of Kribus et al. [54].
Regardless of the absorber’s thermal conductivity, the heat transfer model reaches higher temperatures
at the receiver outlet than the published work. Additionally, the fluid temperature of the presented model
has a faster increase as it crosses the absorber. The latter results might be due to the conduction term
43
Table 4.2: Standard conditions of simulations.
External conditions Absorber parameters Simulation options
Lr [m] 0.02
A [m2] 1
Ginc [kW.m−2] 600 ks [W.m−1.K−1] 40 n 100
m [kg.s−1] 0.6 (cP )s [J.kg−1.K−1] 1244 nt1 15000
Tamb [K] 300 ρs [kg.m−3] 3210 nt2 2000
Tsky [K] 273.15 dp [mm] 0.8
α 0.9
ϕ 0.8
ζ 0.1
of the fluid, which is neglected by Kribus et al. [54]. Although thermal conductivity of air has a low value,
its temperature gradient certainly doesn’t, especially when it is closer to the inlet boundary. Thus, it is
advisable to keep the fluid conduction term, since it predicts its thermal behaviour with better precision.
It is also noticeable in Figure 4.2 that the inlet fluid temperature of the model is lower than in the
results of Kribus et al. [54]. This is because of the convective energy transferred to the fluid, which
has been discussed previously. The gain of the convective energy by the inlet fluid is considered in
the published work, and that causes an increase on both solid and fluid temperatures. This is also the
reason why the volumetric effect is more evident in the presented model than in that work. Particularly
in Figure 4.2(b), the latter phenomenon is non-existent in the results of Kribus et al. [54].
Nevertheless, the temperature profiles of the compared simulations are quite similar. In Figure 4.2(a),
the temperature of the solid increases slightly along its length. As for the fluid temperature, it increases
along its path, and reaches the absorber’s outlet with the same temperature of the solid. Regarding
Figure 4.2(b), the fluid temperature profile attained is similar to the previous case. However, the solid
temperature of both compared simulations experiences a significant decrease closer to the inlet, which
is due to the receiver low thermal conductivity.
The simulations of Figures 4.2(a) and 4.2(b) yield a thermal efficiency of 77.8 % and 79.4 % respec-
(a) ks = 40 W.m−1.K−1 (b) ks = 1 W.m−1.K−1
Figure 4.2: Steady state results of the model compared with the work of Kribus et al. [54].
44
tively, which is an increase of 5.8 % and 9.1 % when compared with the results of Kribus et al. [54]. As
for the radiative losses, the difference between the compared results is less than 1 percentage point. It
can then be concluded that the compared simulations predict an approximate thermal behaviour for both
media. Aside from some minor differences, the simulation results are in agreement with this published
work.
4.3 Steady state analysis: Absorber’s properties
Since the transient model has been successfully validated and compared, it can now be used under
different conditions. For the remaining simulations of this chapter, the selected standard conditions
chosen to run the model are presented in Table 4.2. For each simulation, these parameters are expected
to remain constant unless they are mentioned otherwise.
One can start by performing a study on the absorber’s intrinsic properties (and on ζ as well). These
parameters are changed individually, so that its contribution to the model is identified. Tables E.3 and E.4
(in Appendix E) show more detailed information about the steady state results of each set of presented
simulations.
4.3.1 Thermal conductivity
The effect of thermal conductivity on the absorber thermal behaviour was identified by Kribus et al.
[54], and it has also manifested in this model (as seen in the model comparison). Therefore, a proper
discussion on this parameter must be executed.
Figure 4.3(a) illustrates the temperature profile for each selected value of thermal conductivity. Sim-
ulations with a high ks present a similar absorber temperature throughout its length. However, when
the thermal conductivity is lowered, the receiver temperature near the inlet boundary decreases. By
diminishing ks, the conduction term in expression (3.2a) becomes less significant when compared to the
convection term, which leads to a lower temperature of the solid phase.
The absorber thermal conductivity also affects the thermal behaviour of air. A lower receiver temper-
ature decreases the convection term in equation (3.2b), so less energy is transferred to the fluid. This is
why the temperature gradient of air for ks = 1 W.m−1.K−1 is lower than in the rest of the simulations.
Notice that the volumetric effect is attained for all simulations, but it is more evident for low ther-
mal conductivities. Additionally, the outlet temperatures increase slightly as the thermal conductivities
decrease. Therefore, lowering the absorber thermal conductivity lead to an increase of the thermal effi-
ciency. Simulations show that if ks is decreased from 80 W.m−1.K−1 to 1 W.m−1.K−1, thermal efficiency
increases 1.8 percentage points. Regarding the losses of the system, they aren’t significantly affected
by the different thermal conductivities.
Although decreasing ks raises thermal efficiency, one must also consider thermal stress:
σth,i = aE∆Ts,i , (4.2)
45
Figure 4.3: Simulation results along the receiver for different solid thermal conductivities: (a) Tempera-ture profile; (b) Receiver temperature difference.
where a is the coefficient of thermal expansion, E the Young’s modulus and ∆Ts,i = Ts,i+1− Ts,i−1 [87].
As seen in Figure 4.3(b), the variation of temperature is negligible for high thermal conductivities. How-
ever, for low ks, it presents a pronounced fluctuation near the inlet, which leads to a larger thermal stress
in certain areas of the absorber. This might pose as an issue to the receiver. If σth surpasses the elastic
range, the absorber will be permanently deformed or even become fractured, which is an undesirable
situation. Therefore, the choice of the absorber thermal conductivity must be performed with caution.
4.3.2 Particle diameter
The following simulations address the variation of the particle diameter that compose the receiver.
Although this quantity is sometimes not disclosed in literature, it has a great impact on the thermal be-
haviour of both media, as one can see in Figure 4.4(a) [31]. One can immediately see that the solid
temperature decreases as dp gets lower. One of the main reasons is the high values for the backscatter-
ing losses recorded for low particle diameters (in Table E.4), which must be discussed in detail. Although
backscattering losses are estimated rather than calculated, one knows that this quantity can be related
to the contribution of the collimated radiation to the diffuse radiation (due to expression (3.39b)). Fig-
ure 4.4(b) shows the gradient of diffuse radiative heat flux along the receiver. One can observe that dqddxis much lower at the inlet when dp = 0.2 mm. This means that the lowest particle diameter simulation
has the highest backscattered radiation near the inlet. This radiation is most likely to escape from the
receiver’s inlet than in the other simulations, which leads to higher backscattering losses. The small par-
ticle diameter is also the main reason why the backscattering losses are so high in the model validation.
Note that a small increase of dqddx at the receiver outlet has been observed, which might indicate forward
46
Figure 4.4: Simulation results along the receiver for different particle diameters: (a) Temperature profile;(b) Gradient of the diffuse radiative heat flux.
scattering losses. Although the backscattering loss estimation includes forward scattering losses as
well, the latter contribution is negligible when compared to the former.
Other effects can be seen in Figure 4.4(a). As stated before, the absorber’s temperature raises with
the increase of the particle diameter. However, the gradient of air temperature (at the inlet region) is
lower for larger dp. Both these effects are explained by the heat transfer coefficients, which are inversely
proportional to the particle diameter (h ∝ d−1.4p and hsf,in ∝ d−0.562
p ). Therefore, for low diameters,
more heat is exchanged through convection, which decreases the receiver temperature and increases
the air temperature gradient. Note that the low convection at the inlet for simulations with high particle
diameters causes an overheating of the absorber at that region. For that reason, the volumetric effect
becomes less evident as the diameter increases and is lost when dp = 1.4 mm.
Due to the referred effects of the particle diameter, the energy balance has considerable changes
between simulations. Raising the particle diameter from 0.2 mm to 1.4 mm increased the radiative losses
by 7.3 percentage points, but decreased the convective losses by 10.1 percentage points. Nevertheless,
thermal efficiency has raised significantly with the increase of the particle diameter. From dp = 0.2 mm
to dp = 1.4 mm, an increase of 40.5 % has been recorded for this parameter. Notice that the loss of
collimated energy at the outlet has slightly increased with the raise of the particle diameters. This is
because the extinction coefficient diminishes when dp gets larger, so more collimated radiation exits the
receiver through the outlet aperture.
47
4.3.3 Porosity
The next parameter that is analysed is the porosity, which is one of the most important features of a
volumetric receiver. Figure 4.5 shows the temperature profile of both media for porosities between 0.7
and 0.9, which almost describes the total range of validity of this parameter in the model. Similar to the
previous case, the air temperature gradient (at the inlet region) lowers with the increase of the porosity.
As porosity raises, both αsf and hsf decrease, which leads to a diminish of the volumetric convection
term.
If the same logic of the previous case is applied, one would expect that the decrease of convection led
to an overheating of the absorber at the inlet. However, it is seen in Figure 4.5 that the volumetric effect
becomes more apparent as porosity increases. In addition, the receiver’s inlet temperature decreases
when porosity goes from 0.85 to 0.9, which indicates that there is a maximum in the absorber’s inlet
temperature between 0.8 < ϕ < 0.9.
By definition, the increase of porosity lowers the volume occupied by the solid material. This has
two implications. On one hand, less incident irradiance is absorbed at the inlet solid boundary. On
the other hand, raising the porosity leads to a lower extinction coefficient, meaning that less energy
is absorbed along the receiver (even though the available collimated radiation increases). Therefore,
as porosity increases, the absorbed radiation becomes so low that the inlet air is capable to cool the
receiver entrance, even with a low heat transfer coefficient.
The outlet air temperature raises with the increase of porosity, and so does the thermal efficiency.
When ϕ rises from 0.7 to 0.9, thermal efficiency has an increase of 17 % . Note that if one chooses
an absorber with a lower length and with a high porosity, the outlet air might not reach the receiver
temperature, and thus ηthermal might decrease. Regarding convective losses, they are expected to
decrease when raising the porosity, since hsf,in ∝ ϕ−0.058. A rise in porosity from 0.7 to 0.9 reduces
Figure 4.5: Temperature profile along the absorber for different porosities.
48
convective losses by 4.9 percentage points. As for radiative losses, the difference between the highest
and lowest loss is 1.3 %, which is fairly low.
The remaining loss factors are related to the incident irradiance. As stated before, raising the porosity
diminishes β and increases the collimated radiation travelling through the receiver, so collimated losses
increase as well. In particular, the simulation with the highest porosity recorded a value of 2.6 %, which
is much larger than any other. Concerning the reflection loss factor, it decreased 2 percentage points in
the considerate range of ϕ. Since the absorptance remains unchanged, the variation of this parameter
is linear. Finally, there is an increase of the backscattering losses when porosity diminishes, and the
reason is the same as before. Backscattering is particularly large for the simulation with lowest ϕ.
Nevertheless, its contribution is minor when compared to the one observed for the particle diameter.
4.3.4 Dispertion ratio
The dispersion ratio was introduced in equation (3.37) as a correction factor for the angular spread
of rays. This parameter has never been used in similar models, so one must make an analysis of this
quantity [53, 55, 57]. Different simulations have changed the dispersion ratio from 0 (fully collimated at
the inlet) until 1 (fully diffused), with increments of 0.2.
Figure 4.6(b) illustrates the total incident radiation along the receiver, which according to expres-
sion (3.48), is the sum of the collimated and diffuse radiation. As the dispersion ratio rises, more
collimated radiation becomes diffuse, and thus the total incident radiation increases at the receiver’s
inlet. However, G diminishes along the receiver, and this decrease is greater for high values of ζ. Re-
gardless of the dispersion ratio, the total incident radiation of all simulations arrive to a constant value
Figure 4.6: Simulation results along the absorber for different dispersion ratios: (a) Temperature profile;(b) Total incident radiation.
49
of 0.2 MW.m−2, which corresponds to the internal emission of the receiver. This is why the thermal
behaviour at the outlet is almost indistinguishable between simulations, as seen in Figure 4.6(a). From
ζ = 0 to ζ = 1, the outlet fluid temperature lowers 11.5 K, which corresponds to a decrease of 1.7 % on
the receiver thermal efficiency.
Due to the high values of G at the inlet, the temperature of the absorber increases slightly in this
region for high dispersion ratios. When ζ increases from 0 to 1, so does the radiative losses by 1
percentage point. Additionally, the volumetric effect is lost for dispersion ratios equal or greater than 0.8.
Regarding the convective heat transfer, the dispersion ratio doesn’t affect the heat transfer coefficient,
so the temperature profile of all simulations is similar. Nevertheless, the small increase of the receiver’s
temperature at the inlet rises the convection losses by 0.5 percentage points (from ζ = 0 to ζ = 1).
Finally, the variation of the backscattering losses in these simulations is considered negligible, meaning
that this quantity depends mostly on the absorber geometry rather than on the heliostat field distribution
and focus.
4.4 Steady state analysis: External conditions
An analysis on the external parameters is now carried, specifically to the incident irradiance and to
the mass flow rate. Unlike the absorber’s properties, these quantities can be easily adjusted during the
simulation, in order to reach a maximum system efficiency.
The steady state results of each simulation are also presented in Tables E.3 and E.4. In this study,
four different values of the incident irradiance are chosen. For each Ginc, the mass flow rates are chosen
so that the incident energy per kilogram of air ranges from 600 kJ.kg−1 to 1600 kJ.kg−1, with constant
increments of 200 kJ.kg−1. Notice that the selected range of Qsup/m is similar to the experimental works
performed to this kind of receivers [41].
4.4.1 Thermal behaviour
The thermal behaviour of both phases is illustrated in Figure 4.7. It can be observed that for the
same incident irradiance, the absorber’s temperature raises with the increase of the incident energy
per kilogram of air. When Ginc is constant, an increase of Qsup/m corresponds to a decrease of the
mass flow rate, meaning that the useful power extracted lowers as well. This effect raises the losses of
the system, thus leading to a higher temperature of the receiver. Additionally, a lower mass flow rate
decreases the heat transfer coefficient, which also contributes to the low useful power extracted.
However, the gradient of fluid temperature (at the inlet region) increases when Qsup/m becomes
higher. Considering the information above, the temperature profile of air may seem atypical, but it
can be explained. A decrease on m means that less mass of air enters the absorber (per unit of
time), which lowers its heat capacity. Hence, the fluid temperature increases faster along the absorber.
Moreover, higher fluid temperatures raise the volumetric heat transfer coefficient, which contributes to
the mentioned growth. For those reasons, the volumetric effect is more accentuated for lower mass flow
50
(a) Ginc = 500 kW.m−2 (b) Ginc = 600 kW.m−2
(c) Ginc = 800 kW.m−2 (d) Ginc = 1000 kW.m−2
Figure 4.7: Temperature profile along the receiver for different ratios of incident energy per kilogram ofair (Qsup/m) and for different incident irradiances (Ginc).
rates. For this set of absorber’s properties, this phenomenon is lost for m ≥ 1 kg.s−1 and it is sustained
for mass flow rates lower than 0.83 kg.s−1. This observation can be confirmed in Table E.4, where the
mass flow rates are shown explicitly.
For the same incident energy per kilogram of air, the temperature of the absorber increases when
Ginc raises. Although it isn’t immediately evident, it’s also possible to observe that the fluid temperature
gradient diminishes with the increase of the incident irradiance. In order to maintain the same Qsup/m,
the mass flow rate must increase when Ginc raises. Therefore, the same explanation previously made
to m can also be applied to this analysis.
One must also address the convective and radiative losses of all simulations. For the same incident
irradiance, these losses increase as Qsup/m becomes higher, since the difference between the receiver
temperature and the outside temperatures (Tamb and Tsky) raises. Notice that the increase of tempera-
ture difference has a greater contribution to the convective losses than the decrease of the heat transfer
coefficient. Finally, when maintaining the same incident energy per kilogram of air, the amount of lost
energy increases with the increasing Ginc. Nevertheless, the energy that is supplied to the system is
also greater, and so the fraction of lost energy is actually diminished.
51
4.4.2 Efficiency
With the analysis of the thermal behaviour completed, the optimization of the system efficiency can
be performed. To that end, Figure 4.8 presents all of the efficiencies attained in each simulation, along-
side the outlet fluid temperature. Similar to the model validation, these quantities appear as a function
of Qsup/m, since it is the preferred form of showing these results in literature [41].
For every performed simulation, the air temperature matches the absorber’s temperature at the outlet
(with a maximum difference of 0.5 K). Therefore, the same behaviour as before is seen in Figure 4.8(a),
where Tf,out raises with the increase of Qsup/m and Ginc. As a result, when Ginc remains constant,
the thermal efficiency (shown in Figure 4.8(b)) decreases for higher Qsup/m . However, for the same
Qsup/m, this efficiency raises for larger incident irradiances. This effect is related to the increase of the
mass flow rate, which will become clearer in the figure below.
However, the difference between fluid temperatures with the same incident energy per kilogram of
air becomes larger for higher Qsup/m, and this effect is also manifested in the thermal efficiency. For
Qsup/m = 600 kJ.kg−1, the maximum thermal efficiency difference between simulations is 3.1 percent-
age points; whereas for Qsup/m = 1600 kJ.kg−1, that difference can be as high as 9.8 percentage points.
The purpose of showing data in function of Qsup/m is to normalize it. In theory, this allows to
(a) Outlet air temperature (b) Thermal efficiency
(c) Cycle efficiency (d) System efficiency
Figure 4.8: Overall results of the simulations (as a function of the incident energy per kilogram of air).
52
compare the results against other works with the same conditions [41]. Nevertheless, it has been seen
that simulations with the same incident energy per kilogram of air yield contrasting results, which has
produced differences up to 15.3% in ηthermal. According to Gomez-Garcia et al. [41], this is because the
mass flow rate and the incident irradiance have an independent effect on the outlet air temperature and
consequently on the efficiency. Therefore, Figure 4.9 plots the same simulations for different Ginc, but
this time explicitly as a function of m. With this representation, the mass flow and the irradiance can be
treated separately, which leads to a better understanding of the results. For instance, in Figure 4.9(b),
one can observe that ηthermal is actually lower for higher incident irradiances (that have the same mass
flow). Additionally, this graph suggests that the dependence of ηthermal on m is stronger than on Ginc,
which is a difficult conclusion to extract from Figure 4.8(b). Hence, the graphs in function to the mass
will be preferred through the remaining of this work.
With the attained outlet temperature, the efficiency of the power cycle can be estimated. For this
analysis, it is supposed that the heated air maintains its energy until reaching the heat exchanger, so
that TH = Tf,out. One also assumes that the TC = Tamb. With this considerations, one can see
in Figure 4.9(c) that the profile of the cycle efficiency curves are similar to the outlet fluid temperature
curves in Figure 4.9(a). For an increasing mass flow rate, ηcycle decreases and ηthermal raises. Thus, the
overall system efficiency has an optimal operating point for m, that maximizes the production of energy
(a) Outlet air temperature (b) Thermal efficiency
(c) Cycle efficiency (d) System efficiency
Figure 4.9: Overall results of the simulations (as a function of the mass flow rate).
53
for each Ginc. Figure 4.9(d) shows that for a higher Ginc the point of maximum system efficiency occurs
for larger m. Moreover, if the mass flow is kept constant, ηsystem becomes higher as the irradiance
gets larger. For the performed simulations, a maximum system efficiency of 39.2 % is achieved for
Ginc = 1000 kW.m−2 and m ≈ 0.71 kg.s−1.
4.4.3 Equilibrium time
The presented simulations have started with the all the nodes at ambient temperature. At t = 0 s,
the incident irradiance suddenly increases (from 0 kW.m−2 to the desired Ginc), and these simulations
evolved in time until a steady temperature profile in both media is attained. Thus, the time where thermal
equilibrium is reached can be found.
To do that, one needs to know how to compute time in this model. In theory, it is possible to attain
the thermal properties of the receiver at each time step. However, to reduce the simulation time, a
higher time interval between log values was imposed. The calculation of the printed time is then given
by equation 4.3:
tpr = ∆tct2 nt1 . (4.3)
The equilibrium time (teq) can then be obtained if a ”quasi-steady” state is reached. According to Wu
and Wang [57], one arrives at a ”quasi-steady” state when each measured node is able to maintain a
temperature difference (in respect to time) below 15 K for at least five minutes. Using this definition, the
following general procedure is proposed to calculate teq in each simulation:
1. Record the steady state temperatures of nodes 0 and n (of both media).
2. For each node and phase, report the minimum time where the temperature difference (between
steady state temperature and minimum time temperature) is below 15 K.
3. Finally, subtract the maximum attained time to the initial time (where the incident irradiance pertur-
bation has begun).
Figure 4.10(a) shows the equilibrium time calculated for every simulation. For the same absorber’s
properties, one can see that teq is inversely proportional to the mass flow rate. This conclusion isn’t
unexpected (remember that ∆t ∝ m−1). A larger mass flow rate distributes the energy more quickly
along the receiver, which leads to a faster increase of the temperature, specifically at the absorber’s
outlet. Note that the maximum attained time in the procedure has always occurred for the node n, and
this fact also sustains the latter statement.
If another configuration of absorber’s properties is considered, the values of the equilibrium time
change as well (see Table E.3). This change on the equilibrium time is particularly noticeable when the
porosity of the material is shifted, as shown in Figure 4.10(b). From ϕ = 0.7 to ϕ = 0.9, the equilibrium
time diminishes over a minute. The particle diameter and the thermal conductivity also have impact on
teq, but it is lower when compared to the effect that the porosity induces.
54
(a) teq(m) (b) teq(ϕ)
Figure 4.10: Equilibrium time (teq) as a function of the mass flow rate (m) and as a function of theporosity (ϕ)
4.5 Transient analysis
The previous sections have focused on the steady thermal behaviour of the model, where its param-
eters (defined in Table 4.2) remained undisturbed when performing each simulation individually. In this
section, the response (in time) of the receiver’s model is evaluated, by adding some perturbations on the
external conditions while executing the simulations. These perturbations attempt to recreate the start-up
and shutdown procedures, as well as the passage of a cloud above the heliostat field
In order to perform a successful transient simulation (with perturbations), the following method is
proposed [57]. The algorithm previously defined is applied, so that the steady state temperatures (of both
media along the absorber) are recorded. The attained temperatures are used as the initial conditions
of the simulation, as well as the mass flow and incident irradiance applied in the previous step. In a
central receiver system, the latter quantities can be altered while operating with the absorber. However,
the mass flow rate was assumed to be constant while the absorber is operated. Therefore, Ginc is the
only external parameter that is manipulated along the execution of the simulation.
Note that when dealing with the second stage, some modifications must be made to the algorithm.
In step 4, the steady state temperatures are applied to each node instead of Tamb. Additionally, the
incident irradiance perturbations are inserted in the algorithm between step 10 and 11. Once triggered,
the perturbation can either be a step function or a slope (in respect to time). For the former, Ginc
instantly changes to its setpoint, while for the latter the irradiance gradually varies for a period of time
until reaching its desired value.
4.5.1 Start-up and shutdown response
In the previous section, the receiver has been heated from ambient temperature until a steady state
temperature. These simulations can be associated to the daily start-up procedure of the solar tower
system. In the same way, one can also represent the equivalent of the daily shutdown procedure, that
is, when the absorber is cooled from its steady state temperature to Tamb
55
The conditions for the shutdown procedure will be similar to the ones performed for the start-up with
Ginc = 600 kW.m−2. The same incident irradiance is initially applied, which diminishes to 0 kW.m−2.
Note that this decrease is instantaneous, since it was implicitly applied a step change to the start-up
simulations. Additionally, the selected mass flows in each simulation are equal to the ones at the start-
up. This means that the steady state temperatures are the ones already attained in Figure 4.7(b).
The equilibrium time of both start-up and shutdown is illustrated in Figure 4.11. As one can observe,
the shutdown equilibrium time is also inversely proportional to the mass flow. However, it is always
higher than the start-up teq. In addition, the difference between the start-up and shutdown teq becomes
larger as m decreases.
For the shutdown simulation, convection is the main effect to cool down the absorber. With a low
mass flow rate, air takes more time to pass through the receiver, and thus takes more time to decrease
its temperature. On the other hand, the main mechanism to heat the absorber (in the start-up simulation)
is the absorption of incident irradiance, which happens along the receiver instantly. In the end, the
power absorbed in the start-up procedure is always greater than the power lost through convection (in
the shutdown procedure), and this difference is amplified when m decreases. Therefore, the shutdown
procedure requires again more time to cool the receiver, especially for low mass flow rates.
4.5.2 Cloudy weather response
In a solar tower system, the volumetric receiver is heated due to the direct incident irradiance that is
focused by the heliostats. If a cloud passes nearby the solar tower, it can cast a shadow that covers the
heliostat field (partially or even fully), which decreases the available incident irradiance on the receiver.
The thermal response of the absorber is then evaluated in these conditions, for different time intervals
in which the cloud shades the heliostat field (∆tlo).
Figure 4.11: Start-up and shutdown comparison of the receiver response.
56
Two values for the incident irradiance are assumed in each set of simulations: a lower value for when
the heliostat field is covered (Ginc,lo), and an upper value for when it’s not (Ginc,up). When the irradiance
changes between those values, it doesn’t occur instantaneously. Instead, the irradiance is considered
to vary linearly (from Ginc,up to Ginc,lo or vice versa) in a period of 5 s, to better define the passage of a
cloud. Figure 4.12 shows a set of transient simulations for different ∆tlo. Notice that for each simulation,
∆tlo is the time interval in which the incident irradiance is lower than Ginc,up.
When the irradiance diminishes, the convective and radiative losses become much greater than the
absorbed irradiance. Therefore, the receiver inlet temperature decreases abruptly immediately after
the irradiance drop. However, as Ts,0 decreases, the mentioned losses also diminish. Therefore, this
temperature lowers more slowly as time passes, until it reaches thermal equilibrium (as seen in Fig-
ure 4.12(d)). Note that an analogous analysis can be performed for the inlet temperature increase (after
the cloud passes).
Regarding the outlet fluid, it has been observed in Figure 4.12 that its temperature doesn’t react
immediately to irradiance variations. This occurs because the receiver’s solid structure doesn’t respond
to these changes uniformly, as shown in Figure 4.13(a). During the shutdown procedure, the colder
inlet air receives energy from the absorber, which increases the fluid temperature and decreases the re-
(a) ∆tlo = 10 s (b) ∆tlo = 20 s
(c) ∆tlo = 35 s (d) ∆tlo = 95 s
Figure 4.12: Transient thermal response of the receiver inlet (Ts,0) and fluid outlet (Tf,out) to a temporarydecrease of incident irradiance (Ginc), for different time intervals of the cloud passage (∆tlo).
57
ceiver’s temperature at the inlet region. At the beginning of this procedure, the energy received from the
absorber’s inlet region is such that air reaches the outlet region at thermal equilibrium with the receiver.
One can see in Figure 4.13(a) that the outlet temperature has barely decreased when the shutdown
procedure has started 5 s ago, while the inlet temperature has diminished abruptly. The absorber’s out-
let region can only be cooled when the receiver’s inlet region has significantly lowered its temperature,
which in Figure 4.13(a) is only possible when t > 5 s. This is the reason for the thermal latent response
between the receiver’s edges observed.
Due to this latency, the outlet air temperature isn’t as affected as the receiver inlet temperature for
low ∆tlo. In Figure 4.12(a), a maximum temperature difference of 63.9 K has been recorded for Tf,out
when ∆tlo = 10 s, which is lower than the 122.7 K attained for Ts,0. Nevertheless, the thermal behaviour
at the outlet becomes similar to the inlet thermal behaviour when a higher value of ∆tlo is considered,
since thermal equilibrium has been attained. Once again, note that an analogous study can be done for
the temperature increase.
With the transient results, one can also find the equilibrium times teq,lo and teq,up, which are associ-
ated to the lower and upper irradiances respectively. When the cloud is shading the heliostat field,teq,lo
is the time needed for the system to reach a ”quasi-steady” state; while teq,up is the time required for the
system to recover its normal operation after the passage of the cloud. Since each simulation starts from
a steady state condition, teq,lo is constant in each set of simulations (as seen in Table E.5). However,
when Ginc is increased, the temperature profile of both media might be different. Therefore, Figure 4.14
presents the different upper equilibrium times as a function of the cloud interval time. Notice that in this
figure, the green line corresponds to the simulations performed in Figure 4.12.
When ∆tlo is greater than the lower equilibrium time, a new ”quasi-steady” state is temporarily at-
Figure 4.13: Transient response for a shutdown procedure, with Ginc,up = 800 kW.m−2 and m =0.5 kg.s−1: (a) Temperature profile; (b) Receiver temperature difference.
58
tained in the receiver (when Ginc = Ginc,lo). As a consequence, teq,up becomes constant for higher
cloud interval times. On the other hand, one can see that the upper equilibrium time decreases as ∆tlo
diminishes. This is due to the latency previously identified in the outlet fluid. As a result, a smaller air
temperature drop results in a faster recovery of the absorber’s normal operating conditions. One of the
most interesting results of this work was to see that for low ∆tlo, the recovery time is much larger than
the perturbation itself. For example, in the simulation illustrated in Figure 4.12(a), a ∆tlo = 10 s results
in a upper equilibrium time of 34.4 s.
Figure 4.14 and Table E.5 also compare different sets of simulations. One can see that when the
irradiance difference is maintained (blue and red simulations in Figure 4.14), a higher mass flow rate
results in lower equilibrium times, which is consistent with the previous results. On the other hand, for a
constant m (blue and green simulations in Figure 4.14), a higher irradiance difference generates larger
equilibrium times. This may seem contradictory when compared to the results of Figure 4.10, since one
would expect similar teq for different Ginc differences. However, the simulations performed in Figure 4.10
have the same starting temperature of the receiver and different Ginc,up; while the simulations of this
section have the same Ginc,up but different starting temperatures of the absorber. The variation of the
starting temperatures is responsible for the different teq,up attained.
The blue and green simulations in Figure 4.14 can be compared as an example. Assuming that
∆tlo > teq,lo, the blue simulation has a Tf,out = Tamb when Ginc starts to increase; whereas in the
green simulation, Tf,out = 1000.1 K when Ginc begins to rise. Since the desired setpoint is the same,
the temperature difference in the blue simulation is much greater than in the green, and thus a higher
teq,up is needed.
Figure 4.14: Equilibrium time required for the system to recover its normal operation after the cloudpassage.
59
4.5.3 Thermal stress
As discussed previously, the transient regime of the receiver displays a non-uniform thermal be-
haviour. During the shutdown procedure, represented in Figure 4.13(a), the absorber’s outlet region
temperature is significantly higher than the inlet region temperature. In the majority of time stamps, it
is observed a noticeable increase of temperature along the receiver. Hence, a study on local thermal
stress (σth,i) throughout the absorber can be performed for this situation (using expression 4.2).
Figure 4.13(b) depicts the temperature difference between adjacent nodes along the absorber, for
each time stamp. A maximum ∆Ts,i of almost 15 K is attained for t = 5 s. Additionally, for time stamps
in the range of 5 s ≤ t ≤ 20 s, a ∆Ts,i > 5 K is recorded for most of the receiver’s length. The highlighted
results present significant temperature differences if one takes into account the short distance between
adjacent nodes, which in these simulations is 2∆x = 0.4 mm. Therefore, the observed ∆Ts,i can produce
local thermal stresses along the absorber.
The presented analysis on thermal stress was done for the shutdown procedure, but it can applied to
every situation where a sudden change of the incident irradiance is observed. Thus, a similar behaviour
of ∆Ts,i is expected in all of the simulations performed in this work.
Note that in a real CRS plant, a high thermal gradient along the absorber must be prevented. If the
temperature difference is high enough, the generated stress has the potential to permanently damage
the receiver, and that can ultimately result in the malfunction of the entire power plant. This is a problem
that happens not only to volumetric absorbers but also to other solar tower receivers. In order to avoid
this situation, different philosophies have been employed in these systems to mitigate radiative flux
variations and, consequently, temperature gradients along the absorber [88–90]. A standard procedure
of CRS power plants is to preheat the receiver in the morning, from ambient temperature to a operating
temperature in a gradual manner. An analogous policy is implemented for the shutdown process at the
end of the day. Moreover, the absorber is kept on a standby operation shortly before the solar tower plant
gets shaded by clouds. This standby mode allows the receiver to rapidly resume its normal operation
after the cloud passage, but without provoking an excessive temperature gradient. When all of these
measures are applied, the temperature along the receiver varies more evenly, and thus the durability of
the receiver is greatly increased [88, 89].
60
Chapter 5
Conclusions
The objective of this thesis was to develop a model to study the transient thermal behaviour of an
open-cell foam volumetric receiver. In this work, the model was greatly simplified due to the inclusion of
a unidimensional representation. This was possible because the cross section dimensions of an OCF
absorber are usually much larger than the receiver’s flow length. This has allowed to implement the two-
flux approximation as a simplification of the radiative transfer equation. As the literature hasn’t reported
an one-dimensional transient model of an OCF receiver, the development of this model is relevant in
transient analysis.
As proposed, the presented heat transfer model shows flexibility. It can be used for absorbers with
porosities ranging from 0.66 to 0.9. In addition, it is capable of predicting fluid temperatures between
100 K to 1600 K. Note that the upper temperature boundary is well below the melting point of the material
used in the simulations [10]. Although this model was designed for a silicon carbide absorber, it can be
applied to other materials. To that end, the optical correlation of β must be modified to match the
material’s characteristics.
The initial simulations were focused on the steady state behaviour of the model. In a first analysis,
the receiver’s intrinsic properties have been studied in order to achieve an optimal design. Tests results
have shown that porosity and particle diameter have the greatest impact on the absorber’s thermal
efficiency. After evaluating its properties, the thermal performance of the absorber was tested under
different external conditions. Since the goal of central receiver systems is to maximize power production,
the optimum system efficiency has been attained for each set of simulations with the same incident
irradiance. During this study, it was concluded that the Qsup/m usually found in literature doesn’t offer
a proper normalization of the results. Moreover, in a real solar power plant, both Ginc and m can be
controlled while the receiver is being operated. Hence, a representation depending on the mass flow
rate has been preferred, due to its practical application and to better understand the influence of these
parameters individually.
Regarding the absorber’s transient response, an important parameter is the equilibrium time. It has
been proven in this work that this parameter is strongly dependent on the mass flow rate. Regarding
the absorber’s intrinsic properties, it was verified that porosity has the most significant influence on
61
equilibrium time.
The transient analysis was also performed considering cloudy weather. Since the presented work
hasn’t addressed explicitly the heliostat field, the executed simulations have dealt with the heliostat
shading indirectly. Even though this assumption simplified the weather issue, some effects that occur in
the mirrors have been neglected in this model. For instance, a recent research shows that the impact of
passing clouds can be associated with the optical efficiency of the heliostat field [17]. Nevertheless, this
study has revealed that the normal operation of the solar tower gets immediately disrupted as soon as
the incident radiation lowers, which indicates that the system is highly sensitive to fluctuations of Ginc.
It was also concluded that for short-duration perturbations, the recovery time is much larger than the
perturbation itself, which is one the most relevant results of this work.
The absorber’s sensitivity to the incident irradiance was identified in all of the presented simulations.
It was observed that a sudden change of this parameter creates a temporary temperature difference
along the solid structure, which induces local thermal stresses that can damage the receiver. To prevent
this issue, operational strategies proposed in the literature have been identified in order to extend the
absorber’s lifetime.
5.1 Future Work
During the model validation, it was discussed the influence of the inlet convection term in the fluid
boundary condition. However, this deliberation has only considered the presence or the absence of the
convection term, instead of evaluating the convective energy that re-enters the absorber as a fraction.
Therefore, further work could be investigating the relation between inlet air velocity and wind speed, in
order to assess the ratio of convective power that is lost.
Additionally, one could perform longer simulations of the volumetric receiver model. These simula-
tions should include the operating philosophies proposed above, where a weather predicting tool can
be added to the model as well. Using meteorogical data of a specific site, the simulations could esti-
mate the power produced over a day or even over a year, in order to assess the feasibility of its real
implementation.
62
Bibliography
[1] EASAC. Concentrating solar power: Its potencial contribution to a sustainable energy future. Ger-
man Academy of Sciences Leopoldina, Halle, 2011. (policy report 16).
[2] IEA. World Energy Outlook 2012. OECD/IEA, Paris, 2012.
[3] O. Behar, A. Khellaf, and K. Mohammedi. A review of studies on central receiver solar thermal
power plants. Renewable and Sustainable Energy Reviews, 23:12–39, 2013.
[4] IRENA. Renewable capacity statistics 2017. International Renewable Energy Agency (IRENA),
Abu Dhabi, 2017.
[5] IEA. Technology roadmap: Concentrating solar power. https://www.iea.org/publications/
freepublications/publication/csp_roadmap.pdf, 2010. (Retrieved June 19, 2019).
[6] IRENA. Concentrating solar power. Renewable Energy Technologies: Cost Analysis Series, 1(2),
2012.
[7] World Energy Council. World energy resources 2016. https://www.worldenergy.org/
wp-content/uploads/2016/10/World-Energy-Resources-Full-report-2016.10.03.pdf, 2016.
(Retrieved June 19, 2019).
[8] CSPPLAZA. Global CSP installed capacity increased to
6069 MW In 2018, grew 18.23%. http://en.cspplaza.com/
global-csp-installed-capacity-increased-to-6069-mw-in-2018-grew-18-23, January
2019. (Retrieved June 19, 2019).
[9] M. Romero, R. Buck, and J. E. Pacheco. An update on solar central receiver systems, projects, and
technologies. Journal of Solar Energy Engineering, 124(2):98–108, 2002.
[10] A. L. Avila Marın. Volumetric receivers in solar thermal power plants with central receiver system
technology: A review. Solar Energy, 85(5):891–910, 2011.
[11] K. Lovegrove and W. Stein. Introduction to concentrating solar power (CSP) technology. In K. Love-
grove and W. Stein, editors, Concentrating solar power technology: Principles, developments and
applications, chapter 1, pages 3–15. Woodhead Publishing, Cambridge, 2012.
63
[12] K. Lovegrove and J. Pye. Fundamental principles of concentrating solar power (CSP) systems.
In K. Lovegrove and W. Stein, editors, Concentrating solar power technology: Principles, develop-
ments and applications, chapter 2, pages 16–67. Woodhead Publishing, Cambridge, 2012.
[13] M. Romero-Alvarez and E. Zarza. Concentrating solar thermal power. In F. Kreith and D. Y.
Goswami, editors, Handbook of energy efficiency and renewable energy, chapter 21, pages 21–1–
21–98. Taylor & Francis, Boca Raton, 2007.
[14] European Comission. European research on concentrated solar thermal energy. Office for Official
Publications of the European Communities, Luxembourg, 2004.
[15] M. Romero and A. Steinfeld. Concentrating solar thermal power and thermochemical fuels. Energy
& Environmental Science, 5(11):9234–9245, 2012.
[16] A. L. Avila-Marin, J. Fernandez-Reche, and A. Martinez-Tarifa. Modelling strategies for porous
structures as solar receivers in central receiver systems: A review. Renewable and Sustainable
Energy Reviews, 111:15–33, 2019.
[17] F. Crespi, A. Toscani, P. Zani, D. Sanchez, and G. Manzolini. Effect of passing clouds on the
dynamic performance of a csp tower receiver with molten salt heat storage. Applied Energy, 229:
224–235, 2018.
[18] C. B. Laney. Computational gasdynamics. Cambridge University Press, Cambridge, 1998.
[19] T. L. Bergman, A. S. Lavine, F. P. Incropera, and D. P. DeWitt. Fundamentals of heat and mass
transfer. John Wiley & Sons, Hoboken, 7th edition, 2011.
[20] R. Courant, K. Friedrichs, and H. Lewy. On the partial difference equations of mathematical physics.
IBM Journal of Research and Development, 11(2):215–234, 1967.
[21] C. K. Ho and B. D. Iverson. Review of high-temperature central receiver designs for concentrating
solar power. Renewable and Sustainable Energy Reviews, 29:835–846, 2014.
[22] G. J. Kolb, R. B. Diver, and N. Siegel. Central-station solar hydrogen power plant. Journal of Solar
Energy Engineering, 129(2):179–183, 2007.
[23] C. K. Ho. Advances in central receivers for concentrating solar applications. Solar Energy, 152:
38–56, 2017.
[24] N. P. Siegel, C. K. Ho, S. S. Khalsa, and G. J. Kolb. Development and evaluation of a prototype
solid particle receiver: On-sun testing and model validation. Journal of Solar Energy Engineering,
132(2):021008–1–021008–8, 2010.
[25] P. K. Falcone. A handbook for solar central receiver design. Technical report, Sandia National
Laboratories Livermore, 1986.
[26] W. Wu, B. Gobereit, C. Singer, L. Amsbeck, and R. Pitz-Paal. Direct absorption receivers for high
temperatures. In Proceedings of SolarPACES 2011, pages 20–23. SolarPACES, 2011.
64
[27] J. Leon, M. Sanchez, and J. E. Pacheco. Internal film receiver possibilities for the third generation
of central receiver technology. Journal de Physique IV France, 9(PR3):Pr3–525–Pr3–530, 1999.
[28] C. Singer, R. Buck, R. Pitz-Paal, and H. Muller-Steinhagen. Assessment of solar power tower
driven ultrasupercritical steam cycles applying tubular central receivers with varied heat transfer
media. Journal of Solar Energy Engineering, 132(4):041010–1–041010–12, 2010.
[29] F. J. Miller and R. W. Koenigsdorff. Thermal modeling of a small-particle solar central receiver.
Journal of Solar Energy Engineering, 122(1):23–29, 2000.
[30] W. B. Bienert, H. Rind, and D. A. Wolf. Conceptual design of an open cycle air Brayton solar
receiver: Phase 1 final report. Technical report, California Institute of Technology - Jet Propulsion
Laboratory, 1979.
[31] T. Fend. High porosity materials as volumetric receivers for solar energetics. Optica Applicata, 40
(2):271–284, 2010.
[32] M. Becker, T. Fend, B. Hoffschmidt, R. Pitz-Paal, O. Reutter, V. Stamatov, M. Steven, and D. Trimis.
Theoretical and numerical investigation of flow stability in porous materials applied as volumetric
solar receivers. Solar Energy, 80(10):1241–1248, 2006.
[33] A. Kribus, R. Zaibel, D. Carey, A. Segal, and J. Karni. A solar-driven combined cycle power plant.
Solar Energy, 62(2):121–129, 1998.
[34] A. Kribus, P. Doron, R. Rubin, J. Karni, R. Reuven, S. Duchan, and E. Taragan. A multistage solar
receiver: The route to high temperature. Solar Energy, 67(1–3):3–11, 1999.
[35] M. J. Marcos, M. Romero, and S. Palero. Analysis of air return alternatives for CRS-type open
volumetric reciever. Energy, 29(5–6):677–686, 2004.
[36] B. Hoffschmidt, V. Fernandez, A. G. Konstandopoulos, I. Mavroidis, M. Romero, P. Stobbe, and
F. Tellez. Development of ceramic volumetric receiver technology. In Proceedings of 5th Cologne
Solar Symposium, pages 51–61. DLR, 2001.
[37] R. Capuano, T. Fend, H. Stadler, B. Hoffschmidt, and R. Pitz-Paal. Optimized volumetric solar
receiver: Thermal performance prediction and experimental validation. Renewable Energy, 114
(B):556–566, 2017.
[38] R. Buck, T. Brauning, T. Denk, M. Pfander, P. Schwarzbozl, and F. Tellez. Solar-hybrid gas turbine-
based power tower systems (REFOS). Journal of Solar Energy Engineering, 124(1):2–9, 2002.
[39] European Commission. SOLGATE: Solar Hybrid Gas Turbine Electric Power System. Office for
Official Publications of the European Communities, Luxembourg, 2005.
[40] C. C. Agrafiotis, I. Mavroidis, A. G. Konstandopoulos, B. Hoffschmidt, P. Stobbe, M. Romero, and
V. Fernandez-Quero. Evaluation of porous silicon carbide monolithic honeycombs as volumetric
65
receivers/collectors of concentrated solar radiation. Solar Energy Materials and Solar Cells, 91(6):
474–488, 2007.
[41] F. Gomez-Garcia, J. Gonzalez-Aguilar, G. Olalde, and M. Romero. Thermal and hydrodynamic be-
havior of ceramic volumetric absorbers for central receiver solar power plants: A review. Renewable
and Sustainable Energy Reviews, 57:648–658, 2016.
[42] R. Pitz-Paal, B. Hoffschmidt, M. Bohmer, and M. Becker. Experimental and numerical evaluation
of the performance and flow stability of different types of open volumetric absorbers under non-
homogeneous irradiation. Solar Energy, 60(3–4):135–150, 1997.
[43] B. Hoffschmidt, R. Pitz-Paal, M. Bohmer, T. Fend, and P. Rietbrock. 200 kWth open volumetric
air receiver (HiTRec) of DLR reached 1000 oC average outlet temperature at PSA. Journal de
Physique IV France, 9(PR3):Pr3–551–Pr3–556, 1999.
[44] B. Hoffschmidt, F. M. Tellez, A. Valverde, J. Fernandez, and V. Fernandez. Performance evaluation
of the 200-kWth HiTRec-II open volumetric air receiver. Journal of Solar Energy Engineering, 125
(1):87–94, 2003.
[45] F. M. Tellez Sufrategui. Thermal performance evaluation of the 200 kWth SolAir volumetric solar re-
ceiver. Technical report, Centro de Investigaciones Energeticas Medioambientales y Tecnologicas
(CIEMAT), 2003.
[46] S. Zunft, M. Hanel, M. Kruger, V. Dreißigacker, F. Gohring, and E. Wahl. Julich solar power tower
- Experimental evaluation of the storage subsystem and performance calculation. Journal of Solar
Energy Engineering, 133(3):031019–1–031019–5, 2011.
[47] J. Petrasch, P. Wyss, and A. Steinfeld. Tomography-based monte carlo determination of radia-
tive properties of reticulate porous ceramics. Journal of Quantitative Spectroscopy and Radiative
Transfer, 105(2):180–197, 2007.
[48] A. Kribus, P. Doron, R. Rubin, R. Reuven, E. Taragan, S. Duchan, and J. Karni. Performance of the
directly-irradiated annular pressurized receiver (DIAPR) operating at 20 bar and 1,200oC. Journal
of Solar Energy Engineering, 123(1):10–17, 2001.
[49] M. Lacroix, P. Nguyen, D. Schweich, C. P. Huu, S. Savin-Poncet, and D. Edouard. Pressure drop
measurements and modeling on SiC foams. Chemical Engineering Science, 62(12):3259–3267,
2007.
[50] T. Fend, B. Hoffschmidt, R. Pitz-Paal, O. Reutter, and P. Rietbrock. Porous materials as open volu-
metric solar receivers: Experimental determination of thermophysical and heat transfer properties.
Energy, 29(5-6):823–833, 2004.
[51] Z. Wu, C. Caliot, F. Bai, G. Flamant, Z. Wang, J. Zhang, and C. Tian. Experimental and numerical
studies of the pressure drop in ceramic foams for volumetric solar receiver applications. Applied
Energy, 87(2):504–513, 2010.
66
[52] M. F. Modest. Radiative heat transfer. Academic Press, Oxford, 3rd edition, 2013.
[53] S. Mey, C. Caliot, G. Flamant, A. Kribus, and Y. Gray. Optimization of high temperature SiC volu-
metric solar absorber. Energy Procedia, 49:478–487, 2014.
[54] A. Kribus, M. Grijnevich, Y. Gray, and C. Caliot. Parametric study of volumetric absorber perfor-
mance. Energy Procedia, 49:408–417, 2014.
[55] A. Kribus, Y. Gray, M. Grijnevich, G. Mittelman, S. Mey-Cloutier, and C. Caliot. The promise and
challenge of solar volumetric absorbers. Solar Energy, 110:463–481, 2014.
[56] Z. Wu, C. Caliot, G. Flamant, and Z. Wang. Coupled radiation and flow modeling in ceramic foam
volumetric solar air receivers. Solar Energy, 85(9):2374–2385, 2011.
[57] Z. Wu and Z. Wang. Fully coupled transient modeling of ceramic foam volumetric solar air receiver.
Solar Energy, 89:122–133, 2013.
[58] D. A. Nield and A. Bejan. Convection in porous media. Springer, Cham, 5th edition, 2017.
[59] S. C. Chapra and R. P. Canale. Numerical methods for engineers. McGraw-Hill, New York, 6th
edition, 2010.
[60] E. Isaacson and H. B. Keller. Analysis of numerical methods. Dover Publications, New York, 1994.
[61] K. C. Park. Synthesis tools for structural dynamics and partitioned analysis of coupled systems. In
NATO Advanced Research Workshop, pages 50–111, 2004.
[62] J. A. Duffie and W. A. Beckman. Solar engineering of thermal processes. John Wiley & Sons,
Hoboken, 4th edition, 2013.
[63] J. M. Coulson, J. F. Richardson, J. R. Backhurst, and J. H. Harker. Coulson & Richardson’s chemical
engineering, volume 1. Butterworth-Heinemann, Oxford, 6th edition, 1999.
[64] N. J. Higham. Accuracy and stability of numerical algorithms. Society for Industrial and Applied
Mathematics, Philadelphia, 2nd edition, 2002.
[65] J. Taine, F. Bellet, V. Leroy, and E. Iacona. Generalized radiative transfer equation for porous
medium upscaling: Application to the radiative Fourier law. International Journal of Heat and Mass
Transfer, 53(19–20):4071–4081, 2010.
[66] F. Gomez-Garcia, J. Gonzalez-Aguilar, S. Tamayo-Pacheco, G. Olalde, and M. Romero. Numerical
analysis of radiation propagation in a multi-layer volumetric solar absorber composed of a stack of
square grids. Solar Energy, 121:94–102, 2015.
[67] J. Yang, J. Wang, S. Bu, M. Zeng, Q. Wang, and A. Nakayama. Experimental analysis of forced con-
vective heat transfer in novel structured packed beds of particles. Chemical Engineering Science,
71:126–137, 2012.
67
[68] C. Hsu. Dynamic modeling of convective heat transfer in porous media. In K. Vafai, editor, Hand-
book of porous media, chapter 2, pages 39–80. Taylor & Francis, Boca Raton, 2nd edition, 2005.
[69] J. F. Richardson, J. H. Harker, and J. R. Backhurst. Coulson & Richardson’s chemical engineering,
volume 2. Butterworth-Heinemann, Oxford, 5th edition, 2002.
[70] N. Wakao and S. Kaguei. Heat and mass transfer in packed beds, volume 1. Gordon and Breach,
New York, 1982.
[71] F. Kuwahara, M. Shirota, and A. Nakayama. A numerical study of interfacial convective heat transfer
coefficient in two-energy equation model for convection in porous media. International Journal of
Heat and Mass Transfer, 44(6):1153–1159, 2001.
[72] A. Nakayama. A note on the confusion associated with the interfacial heat transfer coefficient for
forced convection in porous media. International Journal of Heat and Mass Transfer, 79:1–2, 2014.
[73] Z. Wu, C. Caliot, G. Flamant, and Z. Wang. Numerical simulation of convective heat transfer
between air flow and ceramic foams to optimise volumetric solar air receiver performances. Inter-
national Journal of Heat and Mass Transfer, 54(7–8):1527–1537, 2011.
[74] T. J. Hendricks and J. R. Howell. Absorption/scattering coefficients and scattering phase functions
in reticulated porous ceramics. Journal of Heat Transfer, 118(1):79–87, 1996.
[75] R. Viskanta. Combustion and heat transfer in inert porous media. In K. Vafai, editor, Handbook of
porous media, chapter 15, pages 607–644. Taylor & Francis, Boca Raton, 2nd edition, 2005.
[76] Y. Zhao and G. H. Tang. Monte Carlo study on extinction coefficient of silicon carbide porous media
used for solar receiver. International Journal of Heat and Mass Transfer, 92:1061–1065, 2016.
[77] Thermal conductivity of air vs. temperature. http://bouteloup.pierre.free.fr/lica/phythe/
don/air/air_k_plot.pdf, . (Retrieved June 20, 2019).
[78] Specific heat of air vs. temperature. https://ninova.itu.edu.tr/en/courses/
faculty-of-aeronautics-and-astronautics/965/uck-421/ekkaynaklar?g96162, . (Retrieved
June 20, 2019).
[79] C. Borgnakke and R. E. Sonntag. Fundamentals of thermodynamics. John Wiley & Sons, Hoboken,
7th edition, 2009.
[80] P. J. Pritchard and J. C. Leylegian. Fox and McDonald’s introduction to fluid mechanics. John Wiley
& Sons, Hoboken, 8th edition, 2011.
[81] B. D. Tapley and T. R. Poston. Eshbach’s handbook of engineering fundamentals. John Wiley &
Sons, New York, 4th edition, 1990.
[82] S. Irmay. On the theoretical derivation of Darcy and Forchheimer formulas. Eos, Transactions
American Geophysical Union, 39(4):702–707, 1958.
68
[83] S. Ergun. Fluid flow through packed columns. Chemical Engineering Progress, 48(2):89–94, 1952.
[84] G. I. Garrido, F. C. Patcas, S. Lang, and B. Kraushaar-Czarnetzki. Mass transfer and pressure
drop in ceramic foams: A description for different pore sizes and porosities. Chemical Engineering
Science, 63(21):5202–5217, 2008.
[85] E. Rebhan. Efficiency of nonideal carnot engines with friction and heat losses. American Journal
of Physics, 70(11):1143–1149, 2002.
[86] I. I. Nokinov. The efficiency of atomic power stations (a review). J. Nuclear Energy, 7(1–2):125–128,
1958.
[87] F. P. Beer, E. R. Johnston, J. T. DeWolf, and D. F. Mazurek. Mechanics of materials. McGraw-Hill,
New York, 7th edition, 2015.
[88] O. A. J. de Meyer. Optimisation in plant operations for a 100 MW Central Receiver CSP plant with
focus on the plant operating strategies. PhD thesis, Stellenbosch University, 2018.
[89] R. Terdalkar, H. Qian, and G. Ye. Unique challenges in the design and operation philosophy of solar
thermal power plants. Energy Procedia, 49:2521–2531, 2014.
[90] D. Faille and P. Haessig. Heliostat aiming points optimization for concentrated solar power plant. In
60th Annual ISA Power Industry Division Symposium (POWID 2017), 2017.
69
Appendix A
Finite-divided-difference formulas
When solving numerical problems, a partial derivative can be computed using finite-divided-difference
formulas, which are derived from the Taylor series expansion. The following equations present the for-
ward, backward and centered finite-divided-difference formulas for the first and second derivative. Note
that for each derivative and approximation type, the first and second order formulas are shown.
A.1 Forward finite-divided-difference formulas
First derivative:
f ′(xi) =f(xi+1)− f(xi)
δx. (A.1a)
f ′(xi) =−f(xi+2) + 4f(xi+1)− 3f(xi)
2δx. (A.1b)
Second derivative:
f ′′(xi) =f(xi+2)− 2f(xi+1) + f(xi)
(δx)2. (A.2a)
f ′′(xi) =−f(xi+3) + 4f(xi+2)− 5f(xi+1) + 2f(xi)
(δx)2. (A.2b)
A.2 Backward finite-divided-difference formulas
First derivative:
f ′(xi) =f(xi)− f(xi−1)
δx. (A.3a)
f ′(xi) =3f(xi)− 4f(xi−1) + f(xi−2)
2δx. (A.3b)
71
Second derivative:
f ′′(xi) =f(xi)− 2f(xi−1) + f(xi−2)
(δx)2. (A.4a)
f ′′(xi) =2f(xi)− 5f(xi−1) + 4f(xi−2)− f(xi−3)
(δx)2. (A.4b)
A.3 Centered finite-divided-difference formulas
First derivative:
f ′(xi) =f(xi+1)− f(xi−1)
2δx. (A.5a)
f ′(xi) =−f(xi+2) + 8f(xi+1)− 8f(xi−1) + f(xi−2)
12δx. (A.5b)
Second derivative:
f ′′(xi) =f(xi+1)− 2f(xi) + f(xi−1)
(δx)2. (A.6a)
f ′′(xi) =−f(xi+2) + 16f(xi+1)− 30f(xi) + 16f(xi−1)− f(xi−2)
12(δx)2. (A.6b)
72
Appendix B
Auxiliary calculations
B.1 Transient heat transfer: Interior nodes
After applying the finite-divided-difference formulas to the partial derivatives, the heat transfer equa-
tions of interior nodes become:
(1− ϕ)(ρcP )s,iT t+1s,i − T ts,i
∆t= (1− ϕ)ks,i
T t+1s,i+1 − 2T t+1
s,i + T t+1s,i−1
∆x2− dqR
dx
∣∣∣∣i
+ hi(Ttf,i − T ts,i) , (B.1a)
ϕ(ρcP )f,iT t+1f,i − T tf,i
∆t+ (ρcP )f,iuf,i
T t+1f,i − T
t+1f,i−1
∆x= ϕkf,i
T t+1f,i+1 − 2T t+1
f,i + T t+1f,i−1
∆x2+ hi(T
ts,i − T tf,i) .
(B.1b)
By manipulating expression (B.1), one gets:
T t+1s,i −T
ts,i =
∆t
(ρcP )s,i
ks,i∆x2
[(T t+1s,i+1−2T t+1
s,i +T t+1s,i−1
)− ∆x2
(1− ϕ)ks,i
dqRdx
∣∣∣∣i
+hi∆x
2
(1− ϕ)ks,i
(T tf,i−T ts,i
)], (B.2a)
T t+1f,i −T
tf,i+uf,i
∆t
ϕ
T t+1f,i − T
t+1f,i−1
∆x=
∆t
(ρcP )f,i
kf,i∆x2
[(T t+1f,i+1−2T t+1
f,i +T t+1f,i−1
)+hi∆x
2
ϕkf,i
(T ts,i−T tf,i
)]. (B.2b)
It is useful to define the Fourier number(λs(f),i = ∆t
(ρcP )s(f),i
ks(f),i∆x2
)and Biot number
(φs(f),i = hi∆x
2
ks(f),i
)to
equation (B.2), so that:
T t+1s,i − T
ts,i = λs,i
[(T t+1s,i+1 − 2T t+1
s,i + T t+1s,i−1
)− ∆x2
(1− ϕ)ks,i
dqRdx
∣∣∣∣i
+φs,i
1− ϕ(T tf,i − T ts,i
)], (B.3a)
T t+1f,i − T
tf,i + uf,i
∆t
ϕ∆x
(T t+1f,i − T
t+1f,i−1
)= λf,i
[(T t+1f,i+1 − 2T t+1
f,i + T t+1f,i−1
)+φf,iϕ
(T ts,i − T tf,i
)], (B.3b)
73
and rearranging equation (B.3), one gets:
−λs,iT t+1s,i−1 +
(1 + 2λs,i
)T t+1s,i −λs,iT
t+1s,i+1 =
(1−λs,i
φs,i1− ϕ
)T ts,i−λs,i
∆x2
(1− ϕ)ks,i
dqRdx
∣∣∣∣i
+λs,iφs,i
1− ϕT tf,i ,
(B.4a)
−(λf,i+uf,i
∆t
ϕ∆x
)T t+1f,i−1 +
(1+2λf,i+uf,i
∆t
ϕ∆x
)T t+1f,i −λf,iT
t+1f,i+1 =
(1−λf,i
φf,iϕ
)T tf,i+λf,i
φf,iϕT ts,i .
(B.4b)
B.2 Transient heat transfer: Boundary conditions
B.2.1 Inlet
Solid
(1− ϕ)αGinc − (1− ϕ)qrad,in − (1− ϕ)qconv,in = −(1− ϕ)ks,0∂T
∂x
∣∣∣∣0
. (B.5)
There aren’t any nodes of the receiver before the entrance of the receiver, so the forward finite-divided-
difference formula is used to replace the partial derivative. Substituting also equation (3.6) in expres-
sion (B.5) gives:
αGinc − εσ[(T ts,0
)4 − (T tsky)4]− hin(T t+1s,0 − T
t+1f,in
)= ks,0
T t+1s,0 − T
t+1s,1
∆x, (B.6)
and rearranging equation (B.6), one gets:
(hin +
ks,0∆x
)T t+1s,0 −
ks,0∆x
T t+1s,1 = αGinc + hinT
t+1f,in + εσ
[(T tsky
)4 − (T ts,0)4] . (B.7)
Fluid
− (cP )f,0 m(T t+1f,0 − T
t+1f,in
)−Aϕkf,0
∂T
∂x
∣∣∣∣0−
= −Aϕkf,0∂T
∂x
∣∣∣∣0+
. (B.8)
In expression (B.8), there are two partial derivatives associated to the conduction terms. On the left side,
the backward finite-divided-difference formula is used; and on the right side, the forward finite-divided-
difference formula is employed. By also replacing equation (3.8) in (B.8), one attains:
(ρcP )f,0uf,0(T t+1f,0 − T
t+1f,in
)+ ϕkf,0
T t+1f,0 − T
t+1f,in
∆x= ϕkf,0
T t+1f,1 − T
t+1f,0
∆x, (B.9)
and rearranging equation (B.9) gives:
[(ρcP )f,0uf,0 + 2
ϕkf,0∆x
]T t+1f,0 −
ϕkf,0∆x
T t+1f,1 =
[(ρcP )f,0uf,0 +
ϕkf,0∆x
]T t+1f,in . (B.10)
74
B.2.2 Outlet
Solid
− (1− ϕ)ks,n∂T
∂x
∣∣∣∣n
= (1− ϕ)qconv,out + (1− ϕ)qrad,out . (B.11)
Some assumptions can be made at the receiver’s outlet. First, one can consider that the wall in front
of the absorber is always at thermal equilibrium with the outlet air (T t+1w = T t+1
f,out). Additionally, it is
assumed that the outlet fluid temperature is the temperature of the fluid at node ’n’ in the previous time
step (T t+1f,out = T tf,n). Analogously to the inlet case, there aren’t any nodes of the receiver after the
receiver’s outlet, so the backward finite-divided-difference formula is used to replace the partial deriva-
tive. Considering all this assumptions, and using the definition of equation (3.10), expression (B.11)
becomes:
ks,nT t+1s,n−1 − T t+1
s,n
∆x= hout
(T t+1s,n − T tf,n
)+ εσ
[(T ts,n
)4 − (T tf,n)4] , (B.12)
and rearranging equation (B.12), one gets:
− ks,n∆x
T t+1s,n−1 +
(ks,n∆x
+ hout
)T t+1s,n = houtT
tf,n + εσ
[(T tf,n
)4 − (T ts,n)4] . (B.13)
Fluid
− (cP )f,n m(T t+1f,n − T
t+1f,n−1
)−Aϕkf,n
∂T
∂x
∣∣∣∣n−
= −(cP )f,nm(T t+1f,out − T
t+1f,n
)−Aϕkf,n
∂T
∂x
∣∣∣∣n+
. (B.14)
Similar to the inlet case, expression (B.14) has two partial derivatives. The left one uses the backward
finite-divided-difference formula and the right one employs the forward finite-divided-difference formula.
Considering once again that T t+1f,out = T tf,n, expression (B.14) gives:
(ρcP )f,nuf,n(T t+1f,n − T
t+1f,n−1
)+ ϕkf,n
T t+1f,n − T
t+1f,n−1
∆x= (ρcP )f,nuf,n
(T tf,n − T t+1
f,n
)+ ϕkf,n
T tf,n − Tt+1f,n
∆x,
(B.15)
and rearranging equation (B.15), one gets:
−[(ρcP )f,nuf,n +
ϕkf,n∆x
]T t+1f,n−1 + 2
[(ρcP )f,nuf,n +
ϕkf,n∆x
]T t+1f,n =
[(ρcP )f,nuf,n +
ϕkf,n∆x
]T tf,n . (B.16)
75
B.3 Radiative heat transfer: Boundary conditions
The boundary conditions of the two-flux approximation applied to the diffuse radiation are given by
expression (B.17):
x = 0 : Gd,0 + 2qd,0 = 4J1 , (B.17a)
x = Lr : Gd,n − 2qd,n = 4J2 . (B.17b)
Before solving the boundary conditions, it is useful to rewrite equation (3.39a) to obtain a expression
for qd(x):
qd = − 1
4β
dGddx
. (B.18)
Using equations (B.18), (3.40b) and (3.40c), expression (B.17) becomes:
Gd,0 −1
2β
dGddx
∣∣∣0
= 4ϕ
[ζGinc + εσ
(T ts,0
)4], (B.19a)
Gd,n +1
2β
dGddx
∣∣∣n
= 4ϕεσ(T ts,n
)4. (B.19b)
The forward and backward finite-divided-difference formulas are now applied to equations (B.19a) and (B.19b)
respectively, in order to resolve the first order derivative:
Gd,0 −1
2β
Gd,1 −Gd,0∆x
= 4ϕ
[ζGinc + εσ
(T ts,0
)4], (B.20a)
Gd,n +1
2β
Gd,n −Gd,n−1
∆x= 4ϕεσ
(T ts,n
)4, (B.20b)
and rearranging expression (B.20), one gets:
(1 + 2β∆x)Gd,0 −Gd,1 = 8β∆xϕ
[ζGinc + εσ
(T ts,0
)4], (B.21a)
−Gd,n−1 + (1 + 2β∆x)Gd,n = 8β∆xϕεσ(T ts,n
)4. (B.21b)
76
Appendix C
Thomas algorithm
The Thomas algorithm is a computational method that simplifies the Gaussian elimination for tridiag-
onal systems. These kind of systems can be expressed by:
Mw = b , (C.1)
where:
M =
f0 g0 0 · · · 0
e1 f1 g1. . .
...
0. . . . . . . . . 0
.... . . en−1 fn−1 gn−1
0 · · · 0 en fn
w =
w0
w1
...
wn−1
wn
b =
b0
b1...
bn−1
bn
. (C.2)
This algorithm has three steps. First, one performs a LU decomposition, where M is divided in a
lower and upper matrices:
M = LU =
ψ0 0 · · · · · · 0
e1 ψ1. . . . . .
...
0. . . . . . . . .
......
. . . en−1 ψn−1 0
0 · · · 0 en ψn
·
1 γ0 0 · · · 0
0 1 γ1. . .
......
. . . . . . . . . 0...
. . . . . . 1 γn−1
0 · · · · · · 0 1
, (C.3)
where
ψ0 = f0 , (C.4a)
ψi = fi − eiγi−1, i = 1, 2, . . . , n , (C.4b)
γi =giψi, i = 0, 1, . . . , n− 1 . (C.4c)
77
Then, a forward substitution is applied with the lower matrix:
Lr = b , (C.5)
where:
r0 =b0ψ0
, (C.6a)
ri =bi − eiri−1
ψi, i = 1, 2, . . . , n . (C.6b)
Finally, the solution is attained after performing a backward substitution with the upper matrix:
Uw = r , (C.7)
where:
wn = rn , (C.8a)
wi = ri − γiwi+1, i = n− 1, n− 2, . . . , 0 . (C.8b)
This algorithm is only stable in certain situations. One necessary conditions for numerical stability is
the matrix M to be diagonal dominant:
|f0| > |g0| > 0 , (C.9a)
|fi| ≥ |ei|+ |gi|, eigi 6= 0, i = 1, 2, . . . , n− 1 . (C.9b)
|fn| > |en| > 0 . (C.9c)
In these conditions, the matrix M is considered to be non-singular, and therefore LU decomposition
is possible. When this method cannot be applied, Gaussian elimination with partial pivoting is recom-
mended [64].
78
Appendix D
Numerical implementation of the
algorithm
In Chapter 3, an algorithm to solve the heat transfer model was proposed. The source code (in C) of
its numerical implementation is presented below (where the standard conditions shown in Table 4.2 are
applied).
1 # inc lude <s t d i o . h>
2 # inc lude <s t d l i b . h>
3 # inc lude <s t r i n g . h>
4 # inc lude <t ime . h>
5 # inc lude <math . h>
6
7 / / / / / / / / / / / / / / / / Ob jec t i ve : Temperature p r o f i l e / / / / / / / / / / / / / / /
8 double Ts [101 ] , Tf [ 1 0 1 ] ;
9
10 / / / / / / / / / / / / / / / / / / / / / / Global Var iab les / / / / / / / / / / / / / / / / / / / / / / /
11
12 / / User data
13 long i n t Ginc=600000;
14 double Lr =0.02 , Area=1 , dotm =0.6 , Tamb=300 , Tsky =273.15;
15
16 / / I n t e r n a l data
17 i n t n = 100;
18 double varph i =0.8 , alpha =0.9 , eps i l on =0.9 , dp=0.0008 , zeta =0 .1 ;
19 double sigma=5.67∗pow(10 ,−8) , Tk = 273.15 , c1 = 1.458∗pow(10 ,−6) , c2 = 110.4 , pf0 = 101325;
20 double Cmax=0.1 , rhofmin =0.21 , k1k2min = 1.94∗pow(10 ,−6) ;
21
22 / / Model ’ s constants
23 double dx , alphap , a lphasf , dh , kappa , sigmasc , beta , dt , k1 , k2 , k1k2 , umax ;
24
25 / / Time , t ime counters and space counters
26 long i n t t =0 , n t =15000, taux =0 , ntaux =2000;
79
27 double t r e a l ;
28 i n t i ;
29
30 / / Rad ia t i ve p r o f i l e
31 double Gc[101 ] , Gd[101 ] , Gt [ 101 ] , dq [ 1 0 1 ] ;
32
33 / / Temperature c o r r e l a t i o n s
34 double ks [101 ] , cps [101 ] , rhos [101 ] , as [ 1 0 1 ] ;
35 double k f [ 101 ] , cp f [ 101 ] , r ho f [ 101 ] , muf [ 101 ] , nuf [ 101 ] , p f [ 101 ] , u f [ 1 0 1 ] ;
36 i n t Rf = 287;
37
38 / / Temperature dependent p r o p e r t i e s
39 double Re[101 ] , Pr [ 101 ] , Nu[101 ] , hs f [ 101 ] , h [ 1 0 1 ] ;
40 double lambdas [101 ] , lambdaf [ 101 ] , ph is [ 101 ] , p h i f [ 1 0 1 ] ;
41
42 / / Other temperatures and c o r r e l a t i o n s
43 double T f i n ;
44
45 / / Record o ld values
46 double Tsold [101 ] , T fo ld [ 1 0 1 ] ;
47 double T f i n o l d ;
48
49
50 / / / / / / / / / / / / / / / / / / / / / / Funct ions / / / / / / / / / / / / / / / / / / / / / / /
51
52 / / D i f f use r a d i a t i o n , t o t a l r a d i a t i o n and energy deposi ted
53 vo id Gdqcalc ( )
54
55 / / Arrays
56 double f [ 101 ] , g [101 ] , e [101 ] , b [ 1 0 1 ] ;
57 double ps i [ 101 ] , gama[101 ] , r [ 101 ] , w[ 1 0 1 ] ;
58
59 / / Clear v a r i a b l e s
60 f o r ( i =0; i<=n ; ++ i )
61
62 w[ i ] = 0 ;
63 ps i [ i ] = 0 ;
64 gama [ i ] = 0 ;
65 r [ i ] = 0 ;
66
67
68 / / Set i n i t i a l values
69 f [ 0 ] = 1 + 2∗beta∗dx ;
70 g [ 0 ] = −1;
71 e [ 0 ] = 0 ;
72 b [ 0 ] = 8∗beta∗dx∗ varph i ∗ ( zeta ∗ ( double ) Ginc + eps i l on ∗sigma∗pow( Tsold [ 0 ] , 4 ) ) ;
73 f [ n ] = 1 + 2∗beta∗dx ;
74 g [ n ] = 0 ;
75 e [ n ] = −1;
80
76 b [ n ] = 8∗beta∗dx∗ varph i ∗eps i l on ∗sigma∗pow( Tsold [ n ] , 4 ) ;
77
78 f o r ( i =1; i<=n−1; ++ i )
79
80 f [ i ] = 2 + 4∗beta∗kappa∗dx∗dx ;
81 g [ i ] = −1;
82 e [ i ] = −1;
83 b [ i ] = 16∗beta∗kappa∗dx∗dx∗ varph i ∗sigma∗pow( Tsold [ i ] , 4 ) + 4∗beta∗sigmasc∗dx∗dx∗Gc[ i ] ;
84
85
86 / / LU Decomposit ion
87 ps i [ 0 ] = f [ 0 ] ;
88 gama [ 0 ] = g [ 0 ] / ps i [ 0 ] ;
89
90 f o r ( i =1; i<=n−1; ++ i )
91
92 ps i [ i ] = f [ i ] − e [ i ]∗gama [ i −1];
93 gama [ i ] = g [ i ] / ps i [ i ] ;
94
95 ps i [ n ] = f [ n ] − e [ n ]∗gama [ n−1];
96
97 / / Forward s u b s t i t u t i o n ( Lr = b )
98 r [ 0 ] = b [ 0 ] / ps i [ 0 ] ;
99
100 f o r ( i =1; i<=n ; ++ i )
101
102 r [ i ] = ( b [ i ] − e [ i ]∗ r [ i −1]) / ps i [ i ] ;
103
104
105 / / Backward s u b s t i t u t i o n (Ua = r )
106 w[ n ] = r [ n ] ;
107
108 f o r ( i =n−1; i >=0; −− i )
109
110 w[ i ] = r [ i ] − gama [ i ]∗w[ i + 1 ] ;
111
112
113 / / Ca lcu la te Gd, Gt e dq
114 f o r ( i =0; i<=n;++ i )
115
116 Gd[ i ] = w[ i ] ;
117 Gt [ i ] = Gc [ i ] + Gd[ i ] ;
118 dq [ i ] = kappa∗(4∗ varph i ∗sigma∗pow( Tsold [ i ] , 4 ) − Gt [ i ] ) ;
119
120
121
122 / / F l u i d temperature c o r r e l a t i o n s
123 vo id Tcor rca lc ( )
124
81
125 / / I n l e t
126 k f [ 0 ] = −3.9333∗pow(10 ,−4) + 1.0184∗pow(10 ,−4)∗Tfo ld [ 0 ] − 4.8574∗pow(10 ,−8)∗pow( T fo ld [ 0 ] , 2 )
+ 1.5207∗pow(10 ,−11)∗pow( T fo ld [ 0 ] , 3 ) ;
127 cpf [ 0 ] = 1.0575∗pow(10 ,3 ) − 4.4890∗pow(10 ,−1)∗Tfo ld [ 0 ] + 1.1407∗pow(10 ,−3)∗pow( T fo ld [ 0 ] , 2 ) −
7.9999∗pow(10 ,−7)∗pow( T fo ld [ 0 ] , 3 ) + 1.9327∗pow(10 ,−10)∗pow( T fo ld [ 0 ] , 4 ) ;
128 muf [ 0 ] = ( c1∗pow( T fo ld [ 0 ] , 1 . 5 ) ) / ( T fo ld [ 0 ] + c2 ) ;
129 r ho f [ 0 ] = p f [ 0 ] / ( ( double ) Rf∗Tfo ld [ 0 ] ) ;
130 nuf [ 0 ] = muf [ 0 ] / r ho f [ 0 ] ;
131 uf [ 0 ] = dotm / ( Area∗ r ho f [ 0 ] ) ;
132
133 / / I n t e r i o r nodes
134 f o r ( i =1; i<=n ; ++ i )
135
136 k f [ i ] = −3.9333∗pow(10 ,−4) + 1.0184∗pow(10 ,−4)∗Tfo ld [ i ] − 4.8574∗pow(10 ,−8)∗pow( T fo ld [ i
] , 2 ) + 1.5207∗pow(10 ,−11)∗pow( T fo ld [ i ] , 3 ) ;
137 cpf [ i ] = 1.0575∗pow(10 ,3 ) − 4.4890∗pow(10 ,−1)∗Tfo ld [ i ] + 1.1407∗pow(10 ,−3)∗pow( T fo ld [ i ] , 2 )
− 7.9999∗pow(10 ,−7)∗pow( T fo ld [ i ] , 3 ) + 1.9327∗pow(10 ,−10)∗pow( T fo ld [ i ] , 4 ) ;
138 muf [ i ] = ( c1∗pow( T fo ld [ i ] , 1 . 5 ) ) / ( T fo ld [ i ]+ c2 ) ;
139 pf [ i ] = s q r t (pow( p f [ i −1] ,2)−2∗Rf∗Tfo ld [ i ]∗ dx ∗ ( ( ( muf [ i ]∗dotm ) / ( k1∗Area ) ) + (1 / k2 ) ∗pow( dotm /
Area , 2 ) ) ) ;
140 r ho f [ i ] = p f [ i ] / ( ( double ) Rf∗Tfo ld [ i ] ) ;
141 nuf [ i ] = muf [ i ] / r ho f [ i ] ;
142 uf [ i ] = dotm / ( Area∗ r ho f [ i ] ) ;
143
144
145
146
147 / / Temperature dependent p r o p e r t i e s
148 vo id Tpropcalc ( )
149
150 / / Reynolds and Prand t l number
151 f o r ( i =0; i<=n ; ++ i )
152
153 Re[ i ] = ( u f [ i ]∗dh ) / ( va rph i ∗nuf [ i ] ) ;
154 Pr [ i ] = ( cp f [ i ]∗muf [ i ] ) / k f [ i ] ;
155
156
157 / / Nussel t number and heat t r a n s f e r c o e f f i c i e n t ( i n l e t and o u t l e t )
158 Nu[0]=2.0696∗pow( varphi , 0 . 3 8 ) ∗pow(Re[ 0 ] , 0 . 4 3 8 ) ;
159 Nu[ n]=2.0696∗pow( varphi , 0 . 3 8 ) ∗pow(Re [ n ] , 0 . 4 3 8 ) ;
160 hsf [ 0 ] = ( k f [ 0 ]∗Nu [ 0 ] ) / dp ;
161 hsf [ n ] = ( k f [ n ]∗Nu[ n ] ) / dp ;
162 f o r ( i =1; i<=n−1; ++ i )
163
164 / / Nussel t number and heat t r a n s f e r c o e f f i c i e n t ( i n t e r i o r nodes )
165 Nu[ i ]=2+((12∗(1− varph i ) ) / va rph i ) + s q r t (1−varph i ) ∗pow( Pr [ i ] , ( 1 / 3 ) ) ∗pow ( ( Re [ i ]∗dp∗ varph i ) / dh
, 0 . 6 ) ;
166 hsf [ i ] = ( k f [ i ]∗Nu[ i ] ) / dp ;
167 h [ i ]= hs f [ i ]∗ a lphas f ;
82
168
169 / / Four ie r and B io t number
170 lambdas [ i ] = ( d t∗ks [ i ] ) / ( rhos [ i ]∗ cps [ i ]∗ dx∗dx ) ;
171 phis [ i ] = ( h [ i ]∗ dx∗dx ) / ks [ i ] ;
172 lambdaf [ i ] = ( d t∗ k f [ i ] ) / ( r ho f [ i ]∗ cpf [ i ]∗ dx∗dx ) ;
173 p h i f [ i ] = ( h [ i ]∗ dx∗dx ) / k f [ i ] ;
174
175
176
177
178 / / Temperatures outs ide the rece i ve r
179 vo id Textca lc ( )
180
181 T f i n = Tamb;
182
183
184
185 / / Ca lcu la te the s o l i d phase temperature
186 vo id Tscalc ( )
187
188 / / Arrays
189 double f [ 101 ] , g [101 ] , e [101 ] , b [ 1 0 1 ] ;
190 double ps i [ 101 ] , gama[101 ] , r [ 101 ] , w[ 1 0 1 ] ;
191
192 / / Clear v a r i a b l e s
193 f o r ( i =0; i<=n ; ++ i )
194
195 w[ i ] = 0 ;
196 ps i [ i ] = 0 ;
197 gama [ i ] = 0 ;
198 r [ i ] = 0 ;
199
200
201 / / Set i n i t i a l values
202 f [ 0 ] = hsf [ 0 ] + ( ks [ 0 ] / dx ) ;
203 g [ 0 ] = −(ks [ 0 ] / dx ) ;
204 e [ 0 ] = 0 ;
205 b [ 0 ] = alpha ∗ ( double ) Ginc + hsf [ 0 ]∗ T f i n + eps i l on ∗sigma ∗ (pow( Tsky , 4 ) − pow( Tsold [ 0 ] , 4 ) ) ;
206 f [ n ] = ( ks [ n ] / dx ) + hsf [ n ] ;
207 g [ n ] = 0 ;
208 e [ n ] = −(ks [ n ] / dx ) ;
209 b [ n ] = hsf [ n ]∗ Tfo ld [ n ] + eps i l on ∗sigma ∗ (pow( T fo ld [ n ] , 4 ) − pow( Tsold [ n ] , 4 ) ) ;
210
211 f o r ( i =1; i<=n−1; ++ i )
212
213 f [ i ] = 1 + 2∗ lambdas [ i ] ;
214 g [ i ] = −lambdas [ i ] ;
215 e [ i ] = −lambdas [ i ] ;
216 b [ i ] = (1−(( lambdas [ i ]∗ phis [ i ] ) /(1− varph i ) ) ) ∗Tsold [ i ] − ( ( lambdas [ i ]∗ dx∗dx ) /((1− varph i ) ∗ks
83
[ i ] ) ) ∗dq [ i ] + ( ( lambdas [ i ]∗ phis [ i ] ) /(1− varph i ) ) ∗Tfo ld [ i ] ;
217
218
219 / / LU Decomposit ion
220 ps i [ 0 ] = f [ 0 ] ;
221 gama [ 0 ] = g [ 0 ] / ps i [ 0 ] ;
222
223 f o r ( i =1; i<=n−1; ++ i )
224
225 ps i [ i ] = f [ i ] − e [ i ]∗gama [ i −1];
226 gama [ i ] = g [ i ] / ps i [ i ] ;
227
228 ps i [ n ] = f [ n ] − e [ n ]∗gama [ n−1];
229
230 / / Forward s u b s t i t u t i o n ( Lr = b )
231 r [ 0 ] = b [ 0 ] / ps i [ 0 ] ;
232
233 f o r ( i =1; i<=n ; ++ i )
234
235 r [ i ] = ( b [ i ] − e [ i ]∗ r [ i −1]) / ps i [ i ] ;
236
237
238 / / Backward s u b s t i t u t i o n (Ua = r )
239 w[ n ] = r [ n ] ;
240
241 f o r ( i =n−1; i >=0; −− i )
242
243 w[ i ] = r [ i ] − gama [ i ]∗w[ i + 1 ] ;
244
245
246 / / Ca lcu la te Ts
247 f o r ( i =0; i<=n;++ i )
248
249 Ts [ i ] = w[ i ] ;
250
251
252
253
254 / / Ca lcu la te the f l u i d phase temperature
255 vo id T fca l c ( )
256
257 / / Arrays
258 double f [ 101 ] , g [101 ] , e [101 ] , b [ 1 0 1 ] ;
259 double ps i [ 101 ] , gama[101 ] , r [ 101 ] , w[ 1 0 1 ] ;
260
261 / / Clear v a r i a b l e s
262 f o r ( i =0; i<=n ; ++ i )
263
264 w[ i ] = 0 ;
84
265 ps i [ i ] = 0 ;
266 gama [ i ] = 0 ;
267 r [ i ] = 0 ;
268
269
270 / / Set i n i t i a l values
271 f [ 0 ] = rho f [ 0 ]∗ cpf [ 0 ]∗ uf [ 0 ] + ( (2∗ varph i ∗ k f [ 0 ] ) / dx ) ;
272 g [ 0 ] = −(( va rph i ∗ k f [ 0 ] ) / dx ) ;
273 e [ 0 ] = 0 ;
274 b [ 0 ] = rho f [ 0 ]∗ cpf [ 0 ]∗ uf [ 0 ]∗ T f i n + ( ( va rph i ∗ k f [ 0 ] ) / dx ) ∗T f i n ;
275 f [ n ] = 2∗ r ho f [ n ]∗ cpf [ n ]∗ uf [ n ] + ( (2∗ varph i ∗ k f [ n ] ) / dx ) ;
276 g [ n ] = 0 ;
277 e [ n ] = −( r ho f [ n ]∗ cpf [ n ]∗ uf [ n ] + ( ( va rph i ∗ k f [ n ] ) / dx ) ) ;
278 b [ n ] = rho f [ n ]∗ cpf [ n ]∗ uf [ n ]∗ Tfo ld [ n ] + ( ( va rph i ∗ k f [ n ] ) / dx ) ∗Tfo ld [ n ] ;
279
280 f o r ( i =1; i<=n−1; ++ i )
281
282 f [ i ] = 1 + 2∗ lambdaf [ i ] + ( ( u f [ i ]∗ dt ) / ( va rph i ∗dx ) ) ;
283 g [ i ] = −lambdaf [ i ] ;
284 e [ i ] = −( lambdaf [ i ] + ( ( u f [ i ]∗ dt ) / ( va rph i∗dx ) ) ) ;
285 b [ i ] = (1−(( lambdaf [ i ]∗ p h i f [ i ] ) / va rph i ) ) ∗Tfo ld [ i ] + ( ( lambdaf [ i ]∗ p h i f [ i ] ) / va rph i ) ∗Tsold [ i
] ;
286
287
288 / / LU Decomposit ion
289 ps i [ 0 ] = f [ 0 ] ;
290 gama [ 0 ] = g [ 0 ] / ps i [ 0 ] ;
291
292 f o r ( i =1; i<=n−1; ++ i )
293
294 ps i [ i ] = f [ i ] − e [ i ]∗gama [ i −1];
295 gama [ i ] = g [ i ] / ps i [ i ] ;
296
297 ps i [ n ] = f [ n ] − e [ n ]∗gama [ n−1];
298
299 / / Forward s u b s t i t u t i o n ( Lr = b )
300 r [ 0 ] = b [ 0 ] / ps i [ 0 ] ;
301
302 f o r ( i =1; i<=n ; ++ i )
303
304 r [ i ] = ( b [ i ] − e [ i ]∗ r [ i −1]) / ps i [ i ] ;
305
306
307 / / Backward s u b s t i t u t i o n (Ua = r )
308 w[ n ] = r [ n ] ;
309
310 f o r ( i =n−1; i >=0; −− i )
311
312 w[ i ] = r [ i ] − gama [ i ]∗w[ i + 1 ] ;
85
313
314
315 / / Ca lcu la te Tf
316 f o r ( i =0; i<=n;++ i )
317
318 Tf [ i ] = w[ i ] ;
319
320
321
322 / / / / / / / / / / / / / / / / / / Main f u n c t i o n / / / / / / / / / / / / / / / / / / / / / / / / / /
323
324 i n t main ( i n t argc , char ∗∗argv )
325
326
327 i n t j ;
328 / / F i l e to p r i n t t r a n s i e n t values
329 FILE ∗ t t r a n s i e n t ;
330
331 t t r a n s i e n t = fopen ( ” D a t a t r a ns i e n t . t x t ” , ” wt ” ) ;
332
333 f p r i n t f ( t t r a n s i e n t , ” t r e a l \ t \ t Ts0 \ t \ t Ts100 \ t \ t Ts200 \ t \ t Tf0 \ t \ t Tf100 \ t \ t
Tf200 \n ” ) ;
334
335 p r i n t f ( ” Begin a lgo r i t hm \n ” ) ;
336
337 / / Clear ar rays
338 f o r ( i =0; i<=n ; ++ i )
339
340 Ts [ i ] = 0 ;
341 Tf [ i ] = 0 ;
342 Gc[ i ] = 0 ;
343 ks [ i ] = 0 ;
344 cps [ i ] = 0 ;
345 rhos [ i ] = 0 ;
346 as [ i ] = 0 ;
347 k f [ i ] = 0 ;
348 cpf [ i ] = 0 ;
349 r ho f [ i ] = 0 ;
350 muf [ i ] = 0 ;
351 nuf [ i ] = 0 ;
352 pf [ i ] = 0 ;
353 uf [ i ] = 0 ;
354 Re[ i ] = 0 ;
355 Pr [ i ] = 0 ;
356 Nu[ i ] = 0 ;
357 hsf [ i ] = 0 ;
358 h [ i ] = 0 ;
359 lambdas [ i ] = 0 ;
360 lambdaf [ i ] = 0 ;
86
361 phis [ i ] = 0 ;
362 p h i f [ i ] = 0 ;
363 Gd[ i ] = 0 ;
364 Gt [ i ] = 0 ;
365 dq [ i ] = 0 ;
366 Tsold [ i ] = 0 ;
367 Tfo ld [ i ] = 0 ;
368
369
370 / / Def ine mesh
371 dx = Lr / ( double ) n ;
372 umax = dotm / ( Area∗ rhofmin ) ;
373 dt = ( dx∗Cmax) / umax ;
374 p r i n t f ( ”%f \n ” , d t ) ;
375
376 / / Def ine pore diameter and pe rm eab i l i t y
377 alphap = 6/ dp ;
378 a lphas f = (1−varph i ) ∗alphap ;
379 dh = (4∗ varph i ) / a lphas f ;
380 k1 =(pow( varphi , 3 ) ∗dp∗dp ) / (150∗pow(1−varphi , 2 ) ) ;
381 k2 =(pow( varphi , 3 ) ∗dp ) /(1.75∗(1− varph i ) ) ;
382 k1k2=k1 / k2 ;
383 / / Evaluate p o s s i b i l i t y o f f l ow i n s t a b i l i t i e s
384 i f ( k1k2 >= k1k2min )
385 p r i n t f ( ”K1 / K2 OK\n ” ) ;
386 else
387 p r i n t f ( ”K1 / K2 NOT OK OK\n ” ) ;
388
389 / / Op t i ca l c o r r e l a t i o n
390 beta = (12.64∗pow(1−varphi , 0 . 7 ) ) / ( pow( dh , 0 . 7 9 ) ) ;
391 kappa = alpha∗beta ;
392 sigmasc = (1−alpha ) ∗beta ;
393
394 / / I n l e t pressure
395 pf [ 0 ] = pf0 ;
396
397 / / P rope r t i es o f the absorber
398 f o r ( i =0; i<=n ; ++ i )
399
400 ks [ i ] = 40;
401 cps [ i ] = 1244;
402 rhos [ i ] = 3210;
403
404
405 / / Def ine i n i t i a l cond i t i ons ( temperature p r o f i l e , co l l ima ted r a d i a t i o n and cu r ren t t ime )
406 t r e a l = taux∗nt + d t∗ t ;
407 T f i n = Tamb;
408 T f i n o l d = T f i n ;
409 f o r ( i =0; i<=n ; ++ i )
87
410
411 Gc[ i ] = varph i∗(1−zeta ) ∗Ginc∗exp(− i ∗beta∗dx ) ;
412 Ts [ i ] = Tamb;
413 Tf [ i ] = Tamb;
414 Tsold [ i ] = Ts [ i ] ;
415 Tfo ld [ i ] = Tf [ i ] ;
416
417
418 / / Evo lu t i on i n t ime
419 f o r ( taux =0; taux<=ntaux ; ++taux )
420
421 f o r ( t =1; t<=nt ; ++ t )
422
423 t r e a l = d t ∗ ( taux∗nt + t ) ;
424 / / Ca lcu la te d i f f u s e r a d i a t i o n , t o t a l r a d i a t i o n and energy deposi ted i n the s o l i d
425 Gdqcalc ( ) ;
426
427 / / Ca lcu la te temperature c o r r e l a t i o n s
428 Tcor rca lc ( ) ;
429
430 / / Ca lcu la te temperature dependent p r o p e r t i e s
431 Tpropcalc ( ) ;
432
433 / / Ca lcu la te temperatures outs ide the rece i ve r
434 Textca lc ( ) ;
435
436 / / Ca lcu la te the s o l i d phase temperature
437 Tscalc ( ) ;
438
439 / / Ca lcu la te the f l u i d phase temperature
440 Tfca l c ( ) ;
441
442 / / Store temperatures i n backup v a r i a b l e s
443 f o r ( i =0; i<=n ; ++ i )
444
445 Tsold [ i ] = Ts [ i ] ;
446 Tfo ld [ i ] = Tf [ i ] ;
447
448 T f i n o l d = T f i n ;
449
450
451 / / P r i n t t r a n s i e n t values
452 f p r i n t f ( t t r a n s i e n t , ” %.5 f \ t %.2 f \ t %.2 f \ t %.2 f \ t %.2 f \ t %.2 f \ t %.2 f \n ” , t r e a l , Ts
[ 0 ] , Ts [ 5 0 ] , Ts [100 ] , Tf [ 0 ] , Tf [ 5 0 ] , Tf [ 1 0 0 ] ) ;
453
454
455 f c l o s e ( t t r a n s i e n t ) ;
456
457 / / F i l e to p r i n t f i n a l values ( poss ib l y i n steady−s ta te c o n d i t i o n )
88
458 FILE ∗ t f i n a l ;
459 t f i n a l = fopen ( ” Data steady . t x t ” , ” wt ” ) ;
460
461 / / P r i n t f i n a l values
462 f p r i n t f ( t f i n a l , ” t r e a l \ t \ t i \ t x \ t \ t \ t Ginc \ t \ t Gc \ t \ t Gd \ t G \ t \ t dq \ t \ t p \ t
\ t u \ t \ t Ts \ t \ t Tf \n ” ) ;
463 f o r ( i =0; i<=n ; ++ i )
464
465 f p r i n t f ( t f i n a l , ” %.5 f \ t %d \ t %.5 f \ t %l d \ t %.0 f \ t %.0 f \ t %.0 f \ t %.0 f \ t %.0 f \ t %.2 f \
t %.2 f \ t %.2 f \n ” , t r e a l , i , i ∗dx , Ginc , Gc [ i ] , Gd[ i ] , Gt [ i ] , dq [ i ] , p f [ i ] , u f [ i ] , Ts [ i ] ,
Tf [ i ] ) ;
466
467
468 f c l o s e ( t f i n a l ) ;
469
470 p r i n t f ( ”End a lgo r i t hm \n ” ) ;
471 r e t u r n 0 ;
472
89
Appendix E
Detailed results of the simulation
The following appendix presents detailed information about the simulation results in Chapter 4.
91
TableE
.1:D
etailedresults
ofsimulations
performed
inthe
validation.
InitialconditionsS
imulation
results
Node
0N
oden
Maxim
um
mQinc
mTs
Tf
Ts
Tf
xs
Ts
xf
Tf
uf,0
Qf
ηth
erm
al
[kg.s −
1][k
J.k
g−1]
[K]
[K]
[K]
[K]
[mm
][K
][m
m]
[K]
[m.s −
1][W
][−
]
Using
equation(3.7)
0.0
3478
625.7
3745.3
303.9
710.0
710.0
0.0
0745.3
23.9
2710.4
1.7
914754
0.6
78
0.0
3061
711.1
0782.9
305.1
759.4
759.5
0.0
0782.9
23.4
6760.0
1.5
814621
0.6
72
0.0
2476
879.1
3855.1
308.2
851.7
851.8
0.0
0855.1
22.0
8852.8
1.2
914337
0.6
59
0.0
2463
883.8
1857.1
308.3
854.1
854.2
0.0
0857.1
21.6
2855.3
1.2
814329
0.6
58
0.0
1878
1158.6
8969.1
315.1
990.7
990.8
18.8
6993.2
19.3
2993.2
1.0
013819
0.6
35
0.0
1501
1450.0
11078.8
324.3
1118.0
1118.2
16.5
61123.0
17.0
21123.0
0.8
213247
0.6
09
Using
equation(4.1)
0.0
3478
625.7
3781.9
366.9
757.6
757.6
0.0
0781.9
23.4
6758.1
2.1
616546
0.7
60
0.0
3061
711.1
0825.6
380.2
813.9
814.0
0.0
0825.6
22.0
8814.7
1.9
716448
0.7
56
0.0
2476
879.1
3910.2
409.5
920.0
920.1
20.7
0921.5
21.1
6921.5
1.7
216228
0.7
46
0.0
2463
883.8
1912.5
410.4
922.8
922.9
19.7
8924.3
20.2
4924.3
1.7
116222
0.7
45
0.0
1878
1158.6
81045.5
467.1
1082.4
1082.6
18.4
01086.0
18.8
61086.0
1.4
815800
0.7
26
0.0
1501
1450.0
11177.2
536.4
1233.3
1233.5
15.1
81240.6
15.6
41240.6
1.3
615283
0.7
02
92
Tabl
eE
.2:
Ene
rgy
bala
nce
ofsi
mul
atio
nspe
rform
edin
the
valid
atio
n.
Initi
alco
nditi
ons
Sim
ulat
ion
para
met
ers
Ene
rgy
bala
nce
mQsup
mhsf,in
hsf,out
Gc,n
ηth
erm
al
Loss
1L
oss
2L
oss
3L
oss
4L
oss
5L
oss
6
[kg.s−1]
[kJ.k
g−1]
[W.m
−2.K
−1]
[W.m
−2.K
−1]
[W.m
−2]
[%]
[%]
[%]
[%]
[%]
[%]
[%]
Usi
ngeq
uatio
n(3
.7)
0.0
3478
625.7
31156.7
1782.9
0.0
67.8
1.2
8.6
2.2
0.0
0.0
20.2
0.0
3061
711.1
01096.2
1740.0
0.0
67.2
1.5
8.9
2.2
0.0
0.0
20.3
0.0
2476
879.1
31004.4
1672.6
0.0
65.9
2.1
9.3
2.2
0.0
0.0
20.5
0.0
2463
883.8
11002.2
1671.0
0.0
65.8
2.7
10.3
2.2
0.0
0.0
19.0
0.0
1878
1158.6
8900.6
1591.0
0.0
63.5
3.4
10.1
2.2
0.0
0.0
20.8
0.0
1501
1450.0
1829.2
1529.5
0.0
60.9
5.3
10.8
2.2
0.0
0.0
20.8
Usi
ngeq
uatio
n(4
.1)
0.0
3478
625.7
31279.9
1838.1
0.0
76.0
1.4
0.0
2.2
0.0
0.0
20.3
0.0
3061
711.1
01233.1
1797.2
0.0
75.6
1.8
0.0
2.2
0.0
0.0
20.4
0.0
2476
879.1
31168.5
1733.9
0.0
74.6
2.7
0.0
2.2
0.0
0.0
20.6
0.0
2463
883.8
11167.0
1732.4
0.0
74.5
2.7
0.0
2.2
0.0
0.0
20.6
0.0
1878
1158.6
81108.2
1660.5
0.0
72.6
4.7
0.0
2.2
0.0
0.0
20.5
0.0
1501
1450.0
11077.2
1609.4
0.0
70.2
7.5
0.0
2.2
0.0
0.0
20.1
93
TableE
.3:D
etailedresults
ofsteadystate
simulations
performed
inC
hapter4.Initialconditions
Sim
ulationresults
Node
0N
oden
Maxim
um
Ginc
mQsup
mks
ϕdp
ζTs
Tf
Ts
Tf
xs
Ts
xf
Tf
teq
uf,0
Qf
ηthermal
ηcycle
ηsystem
[kW.m
−2
][kg.s −
1]
[kJ.k
g−
1]
[W.m
−1.K
−1
][−
][m
m]
[−]
[K]
[K]
[K]
[K]
[mm
][K
][m
m]
[K]
[s][m.s −
1]
[W]
[−]
[−]
[−]
500
0.8
3333
600
40
0.8
00.8
0.1
764.6
305.3
765.8
765.9
11.0
766.6
13.0
766.5
46.6
50.7
2403913
0.8
08
0.3
74
0.3
02
500
0.6
2500
800
40
0.8
00.8
0.1
876.3
310.8
891.2
891.4
10.2
892.9
11.4
892.8
63.3
00.5
5389561
0.7
79
0.4
20
0.3
27
500
0.5
0000
1000
40
0.8
00.8
0.1
976.9
318.8
1000.0
1000.1
9.2
1002.9
10.2
1002.9
78.2
50.4
5373144
0.7
46
0.4
52
0.3
38
500
0.4
1667
1200
40
0.8
00.8
0.1
1065.1
329.1
1092.5
1092.7
8.6
1097.2
9.4
1097.1
90.8
70.3
9355395
0.7
11
0.4
76
0.3
38
500
0.3
5714
1400
40
0.8
00.8
0.1
1141.1
341.5
1170.4
1170.7
8.0
1177.1
8.6
1177.1
101.0
80.3
5337133
0.6
74
0.4
94
0.3
33
500
0.3
1250
1600
40
0.8
00.8
0.1
1206.0
355.4
1235.7
1235.9
7.4
1244.5
8.0
1244.4
109.0
70.3
1319063
0.6
38
0.5
07
0.3
24
600
1.0
0000
600
40
0.8
00.8
0.1
780.2
304.1
771.5
771.6
0.0
780.2
13.6
772.1
38.8
10.8
6490879
0.8
18
0.3
76
0.3
08
600
0.7
5000
800
40
0.8
00.8
0.1
892.7
308.3
901.8
902.0
10.4
903.4
12.0
903.3
53.0
90.6
5476371
0.7
94
0.4
23
0.3
36
600
0.6
0000
1000
40
0.8
00.8
0.1
996.4
314.5
1016.8
1017.0
9.6
1019.8
10.8
1019.7
66.1
50.5
3459376
0.7
66
0.4
57
0.3
50
600
0.5
0000
1200
40
0.8
00.8
0.1
1089.5
322.7
1116.6
1116.8
8.8
1121.3
9.8
1121.2
77.6
20.4
6440476
0.7
34
0.4
82
0.3
54
600
0.4
2857
1400
40
0.8
00.8
0.1
1171.2
332.9
1201.9
1202.2
8.2
1208.9
9.0
1208.7
87.1
70.4
0420466
0.7
01
0.5
00
0.3
51
600
0.3
7500
1600
40
0.8
00.8
0.1
1242.2
344.7
1274.3
1274.6
7.8
1283.7
8.2
1283.6
95.0
90.3
7400121
0.6
67
0.5
15
0.3
43
800
1.3
3333
600
40
0.8
00.8
0.1
811.4
302.7
778.5
778.7
0.0
811.4
14.6
779.0
29.0
11.1
4664815
0.8
31
0.3
79
0.3
15
800
1.0
0000
800
40
0.8
00.8
0.1
923.2
305.5
915.3
915.5
0.0
923.2
12.8
916.6
39.7
50.8
7650356
0.8
13
0.4
28
0.3
48
800
0.8
0000
1000
40
0.8
00.8
0.1
1029.2
309.5
1038.8
1039.1
10.0
1041.6
11.6
1041.5
50.1
60.7
0632803
0.7
91
0.4
63
0.3
66
800
0.6
6667
1200
40
0.8
00.8
0.1
1127.3
315.0
1148.8
1149.2
9.2
1153.5
10.4
1153.4
59.7
20.5
9612518
0.7
66
0.4
89
0.3
74
800
0.5
7143
1400
40
0.8
00.8
0.1
1216.4
322.0
1245.4
1245.8
8.8
1252.5
9.6
1252.3
68.1
30.5
2590128
0.7
38
0.5
09
0.3
76
800
0.5
0000
1600
40
0.8
00.8
0.1
1295.9
330.5
1329.2
1329.6
8.0
1339.1
8.8
1338.9
75.2
20.4
7566409
0.7
08
0.5
25
0.3
72
1000
1.6
6667
600
40
0.8
00.8
0.1
843.2
302.0
782.7
782.8
0.0
843.2
15.6
783.1
23.1
71.4
3838582
0.8
39
0.3
81
0.3
19
1000
1.2
5000
800
40
0.8
00.8
0.1
953.3
304.0
923.4
923.7
0.0
953.3
13.6
924.6
31.7
01.0
8824410
0.8
24
0.4
30
0.3
55
1000
1.0
0000
1000
40
0.8
00.8
0.1
1059.0
306.9
1052.4
1052.8
0.0
1059.0
12.2
1054.9
40.1
30.8
7806725
0.8
07
0.4
66
0.3
76
1000
0.8
3333
1200
40
0.8
00.8
0.1
1158.9
310.8
1169.3
1169.8
9.6
1173.9
11.0
1173.6
48.1
60.7
3785772
0.7
86
0.4
94
0.3
88
1000
0.7
1429
1400
40
0.8
00.8
0.1
1251.6
315.8
1273.9
1274.4
9.0
1280.8
10.2
1280.6
55.4
80.6
4761939
0.7
62
0.5
15
0.3
92
1000
0.6
2500
1600
40
0.8
00.8
0.1
1336.2
322.1
1366.2
1366.7
8.6
1376.1
9.4
1375.9
61.8
90.5
7735932
0.7
36
0.5
31
0.3
91
600
0.6
0000
1000
10.8
00.8
0.1
1006.4
313.8
1029.1
1030.5
13.0
1036.8
13.8
1036.7
45.2
60.5
3468649
0.7
81
0.4
60
0.3
60
600
0.6
0000
1000
20
0.8
00.8
0.1
986.2
314.2
1019.5
1019.8
10.2
1023.8
11.2
1023.7
62.3
70.5
3461309
0.7
69
0.4
58
0.3
52
600
0.6
0000
1000
60
0.8
00.8
0.1
1001.1
314.6
1015.7
1015.8
9.4
1017.9
10.4
1017.8
67.7
30.5
3458544
0.7
64
0.4
57
0.3
49
600
0.6
0000
1000
80
0.8
00.8
0.1
1003.7
314.7
1015.0
1015.1
9.2
1016.8
10.4
1016.8
68.4
60.5
3458083
0.7
63
0.4
56
0.3
48
600
0.6
0000
1000
40
0.7
00.8
0.1
959.6
323.2
960.6
960.7
5.2
962.9
5.8
962.9
97.4
40.5
5420844
0.7
01
0.4
41
0.3
09
600
0.6
0000
1000
40
0.7
50.8
0.1
980.1
319.1
987.3
987.4
7.0
990.0
7.8
990.0
82.2
20.5
4439035
0.7
32
0.4
49
0.3
28
600
0.6
0000
1000
40
0.8
50.8
0.1
1000.7
309.6
1049.0
1049.4
13.4
1050.9
15.2
1050.6
49.6
70.5
3481677
0.8
03
0.4
65
0.3
74
600
0.6
0000
1000
40
0.9
00.8
0.1
976.5
305.2
1066.6
1064.7
19.0
1066.7
20.0
1064.7
33.5
00.5
2492304
0.8
21
0.4
69
0.3
85
600
0.6
0000
1000
40
0.8
00.2
0.1
828.7
355.5
847.0
847.0
4.2
849.1
4.2
849.1
57.3
30.6
0344339
0.5
74
0.4
05
0.2
32
600
0.6
0000
1000
40
0.8
00.5
0.1
942.2
328.2
969.8
969.8
7.2
972.8
7.6
972.8
63.4
20.5
6427057
0.7
12
0.4
44
0.3
16
600
0.6
0000
1000
40
0.8
01.1
0.1
1036.8
308.8
1039.6
1040.0
11.2
1042.0
13.6
1041.6
68.0
40.5
2475182
0.7
92
0.4
63
0.3
67
600
0.6
0000
1000
40
0.8
01.4
0.1
1073.5
306.0
1051.9
1052.3
0.0
1073.5
16.8
1052.7
69.7
20.5
2483717
0.8
06
0.4
66
0.3
76
600
0.6
0000
1000
40
0.8
00.8
0993.5
314.4
1018.0
1018.2
9.8
1020.9
10.8
1020.8
66.2
60.5
3460160
0.7
67
0.4
57
0.3
51
600
0.6
0000
1000
40
0.8
00.8
0.2
999.3
314.5
1015.7
1015.9
9.4
1018.7
10.4
1018.6
66.1
50.5
3458592
0.7
64
0.4
57
0.3
49
600
0.6
0000
1000
40
0.8
00.8
0.4
1005.2
314.7
1013.4
1013.6
9.0
1016.5
10.2
1016.4
66.1
50.5
3457018
0.7
62
0.4
56
0.3
47
600
0.6
0000
1000
40
0.8
00.8
0.6
1011.0
314.8
1011.1
1011.3
8.2
1014.3
9.6
1014.2
66.1
50.5
4455445
0.7
59
0.4
55
0.3
46
600
0.6
0000
1000
40
0.8
00.8
0.8
1016.8
315.0
1008.8
1009.0
0.0
1016.8
8.8
1012.1
66.1
50.5
4453865
0.7
56
0.4
55
0.3
44
600
0.6
0000
1000
40
0.8
00.8
11022.6
315.1
1006.5
1006.7
0.0
1022.6
7.8
1010.1
66.1
50.5
4452286
0.7
54
0.4
54
0.3
42
94
Tabl
eE
.4:
Ene
rgy
bala
nce
ofst
eady
stat
esi
mul
atio
nspe
rform
edin
Cha
pter
4.
Initi
alco
nditi
ons
Sim
ulat
ion
para
met
ers
Ene
rgy
bala
nce
Ginc
mQsup
mks
ϕdp
ζhsf,in
hsf,out
Gc,n
ηthermal
Loss
1L
oss
2L
oss
3L
oss
4L
oss
5L
oss
6
[kW.m
−2
][k
g.s
−1
][k
J.k
g−
1]
[W.m
−1.K
−1
][−
][m
m]
[−]
[W.m
−2.K
−1
][W.m
−2.K
−1
][W.m
−2
][%
][%
][%
][%
][%
][%
][%
]
500
0.8
3333
600
40
0.8
00.8
0.1
512.4
816.4
9.3
680.8
3.4
9.5
2.0
0.0
0.0
4.3
500
0.6
2500
800
40
0.8
00.8
0.1
456.2
772.4
9.3
677.9
6.0
10.5
2.0
0.0
0.0
3.6
500
0.5
0000
1000
40
0.8
00.8
0.1
419.4
739.5
9.3
674.6
9.2
11.4
2.0
0.0
0.0
2.8
500
0.4
1667
1200
40
0.8
00.8
0.1
393.9
712.7
9.3
671.1
13.1
12.1
2.0
0.0
0.0
1.8
500
0.3
5714
1400
40
0.8
00.8
0.1
375.6
689.7
9.3
667.4
17.2
12.6
2.0
0.0
0.0
0.7
500
0.3
1250
1600
40
0.8
00.8
0.1
361.9
669.4
9.3
663.8
21.5
13.1
2.0
0.0
0.0
−0.5
600
1.0
0000
600
40
0.8
00.8
0.1
553.8
887.3
11.2
481.8
3.1
8.9
2.0
0.0
0.0
4.2
600
0.7
5000
800
40
0.8
00.8
0.1
491.9
841.3
11.2
479.4
5.4
9.7
2.0
0.0
0.0
3.5
600
0.6
0000
1000
40
0.8
00.8
0.1
450.9
807.4
11.2
476.6
8.3
10.5
2.0
0.0
0.0
2.6
600
0.5
0000
1200
40
0.8
00.8
0.1
422.2
780.3
11.2
473.4
11.9
11.1
2.0
0.0
0.0
1.5
600
0.4
2857
1400
40
0.8
00.8
0.1
401.2
757.5
11.2
470.1
16.0
11.7
2.0
0.0
0.0
0.3
600
0.3
7500
1600
40
0.8
00.8
0.1
385.6
737.4
11.2
466.7
20.2
12.1
2.0
0.0
0.0
−1.0
800
1.3
3333
600
40
0.8
00.8
0.1
626.7
1010.7
14.9
883.1
2.7
8.0
2.0
0.0
0.0
4.1
800
1.0
0000
800
40
0.8
00.8
0.1
555.2
960.9
14.9
881.3
4.6
8.7
2.0
0.0
0.0
3.4
800
0.8
0000
1000
40
0.8
00.8
0.1
507.1
925.3
14.9
879.1
7.1
9.2
2.0
0.0
0.0
2.5
800
0.6
6667
1200
40
0.8
00.8
0.1
472.6
898.0
14.9
876.6
10.3
9.8
2.0
0.0
0.0
1.4
800
0.5
7143
1400
40
0.8
00.8
0.1
447.1
875.8
14.9
873.8
13.9
10.2
2.0
0.0
0.0
0.0
800
0.5
0000
1600
40
0.8
00.8
0.1
427.6
856.8
14.9
870.8
18.0
10.6
2.0
0.0
0.0
−1.4
1000
1.6
6667
600
40
0.8
00.8
0.1
690.2
1117.3
18.7
383.9
2.6
7.5
2.0
0.0
0.0
4.1
1000
1.2
5000
800
40
0.8
00.8
0.1
610.6
1064.0
18.7
382.4
4.2
8.0
2.0
0.0
0.0
3.4
1000
1.0
0000
1000
40
0.8
00.8
0.1
556.6
1026.8
18.7
380.7
6.4
8.4
2.0
0.0
0.0
2.5
1000
0.8
3333
1200
40
0.8
00.8
0.1
517.4
999.2
18.7
378.6
9.2
8.9
2.0
0.0
0.0
1.3
1000
0.7
1429
1400
40
0.8
00.8
0.1
487.9
977.8
18.7
376.2
12.5
9.3
2.0
0.0
0.0
0.0
1000
0.6
2500
1600
40
0.8
00.8
0.1
465.1
960.3
18.7
373.6
16.2
9.6
2.0
0.0
0.0
−1.5
600
0.6
0000
1000
10.8
00.8
0.1
450.4
812.5
11.2
478.1
8.7
10.6
2.0
0.0
0.1
0.5
600
0.6
0000
1000
20
0.8
00.8
0.1
450.7
808.5
11.2
476.9
8.0
10.3
2.0
0.0
0.0
2.8
600
0.6
0000
1000
60
0.8
00.8
0.1
451.0
806.9
11.2
476.4
8.5
10.5
2.0
0.0
0.0
2.5
600
0.6
0000
1000
80
0.8
00.8
0.1
451.1
806.7
11.2
476.3
8.6
10.6
2.0
0.0
0.0
2.5
600
0.6
0000
1000
40
0.7
00.8
0.1
364.2
625.4
0.0
070.1
7.2
12.0
3.0
0.0
0.0
7.7
600
0.6
0000
1000
40
0.7
50.8
0.1
402.2
704.5
0.0
873.2
7.8
11.4
2.5
0.0
0.0
5.1
600
0.6
0000
1000
40
0.8
50.8
0.1
519.0
951.4
653.0
580.3
8.5
9.1
1.5
0.1
0.0
0.5
600
0.6
0000
1000
40
0.9
00.8
0.1
628.6
1169.5
15826.9
582.1
7.7
7.1
1.0
2.6
0.0
−0.5
600
0.6
0000
1000
40
0.8
00.2
0.1
1049.7
1614.9
0.0
057.4
4.0
18.5
2.0
0.0
0.0
18.1
600
0.6
0000
1000
40
0.8
00.5
0.1
600.9
1028.0
0.1
071.2
6.7
12.9
2.0
0.0
0.0
7.3
600
0.6
0000
1000
40
0.8
01.1
0.1
373.3
682.4
117.6
079.2
9.8
9.2
2.0
0.0
0.0
−0.2
600
0.6
0000
1000
40
0.8
01.4
0.1
324.4
599.3
488.4
980.6
11.2
8.4
2.0
0.1
0.0
−2.3
600
0.6
0000
1000
40
0.8
00.8
0450.9
807.8
12.4
876.7
8.2
10.4
2.0
0.0
0.0
2.6
600
0.6
0000
1000
40
0.8
00.8
0.2
451.0
807.0
9.9
976.4
8.4
10.5
2.0
0.0
0.0
2.6
600
0.6
0000
1000
40
0.8
00.8
0.4
451.1
806.1
7.4
976.2
8.6
10.6
2.0
0.0
0.0
2.6
600
0.6
0000
1000
40
0.8
00.8
0.6
451.2
805.2
4.9
975.9
8.8
10.7
2.0
0.0
0.0
2.5
600
0.6
0000
1000
40
0.8
00.8
0.8
451.3
804.3
2.5
075.6
9.0
10.8
2.0
0.0
0.0
2.5
600
0.6
0000
1000
40
0.8
00.8
1451.4
803.5
0.0
075.4
9.3
10.9
2.0
0.0
0.0
2.5
95
Table E.5: Summary of transient simulation results performed in Chapter 4.
Initial conditions Simulation resultsm Ginc,up Ginc,lo ∆tlo teq,lo teq,up
[kg.s−1] [kW.m−2] [kW.m−2] [s] [s] [s]
0.5 800 0 10 104.33 50.780.5 800 0 20 104.33 65.270.5 800 0 35 104.33 72.200.5 800 0 65 104.33 76.360.5 800 0 95 104.33 77.360.5 800 0 125 104.33 77.62
0.8 800 0 10 61.58 38.750.8 800 0 20 61.58 47.640.8 800 0 35 61.58 51.190.8 800 0 65 61.58 52.530.8 800 0 95 61.58 52.680.8 800 0 125 61.58 52.76
0.5 800 500 10 63.76 34.400.5 800 500 20 63.76 47.880.5 800 500 35 63.76 53.900.5 800 500 65 63.76 56.700.5 800 500 95 63.76 57.200.5 800 500 125 63.76 57.20
96