Documenti di Didattica
Documenti di Professioni
Documenti di Cultura
Vorrei ringraziare il Prof. Stefano Campanari per aver riposto la sua fiducia in me e con
cui confrontarsi é stato illuminante.
Un ringraziamento speciale va a Luca, che mi ha guidato step by step nella realiz-
zazione del progetto, da vicino e anche da lontano con svariati mezzi di comunicazione,
per avermi dato nuove visioni, per la disponibilitá e pazienza sfoderate.
Nel team per la realizzazione di questa tesi non posso non citare Andrea, che ha
contribuito con la sua conoscenza ’solare’ condita con perle di saggezza. Voglio ricordare
anche gli esperti Stefano e Giulio, senza l’aiuto dei quali avrei litigato maggiormente
col caro Aspen, l’ing. Matti che ho contattato troppe volte per i miei problemi con la
tecnologia.
Inoltre, in questi ringraziamenti, vorrei includere i miei genitori, sostegno a tutto
questo, e le sisters, insostituibili complici, che hanno potuto testare la loro capacitá di
sopportazione, mettiamola cosı́; Laura, che si é sorbita tutte le puntate di questa tesi, ma
quante risate vitali insieme, perché ci piace sdrammatizzare su tutto; gli amici colleghi
che hanno condiviso con me tanti momenti, tra cui le avventure milanesi con Silvi, i
viaggi (spesso simili a odissee) con i pendolari Pazzi, i pranzi in bovisa e gli aperitivi in
allegria con la Palestra.
Tante persone e situazioni hanno contribuito a rendere unico questo viaggio al Poli,
che mi ha plasmato e da cui esco fortificata. Insomma queste parole per dirvi semplice-
mente grazie! E spero di aver lasciato anche io una traccia in ognuno di voi.
Extended abstract
1 Introduction
The global energy system is undergoing a profound transformation, with a view to achieving the decar-
bonisation across all sectors. This challenge could be addressed by the use of hydrogen produced from
renewable sources; therefore the development of near-zero emission technologies for hydrogen production
is of paramount importance. In particular the coupling of high temperature electrolysis technology, per-
forming electrochemical water splitting, with a CSP system has stirred up interest in the research, for its
capability to deliver both thermal and electric energy required for carrying out the electrolysis reaction.
Mohammadi and Mehrpooya [11] proposed a system that integrates SOEC with a parabolic trough
section. Solar thermal energy plays its part to pre-heat the reacting steam, along with the hot streams
exiting the SOEC stack and an auxiliary electric heater. For electricity generation, part of the thermal
energy is delivered to an Organic Rankine Cycle. A sensitivity analysis on a number of key parame-
ters related to operating conditions of ORC and SOEC is performed and, by way of a multi-objective
optimization, it was found that the highest achievable exergy efficiency is 26.81%.
The feasibility of a solar tower system coupled with SOEC was deeply analyzed by Lin and Haussener
[8], providing evidence that, among different proposed scenarios, this solution is able to work at high
solar to hydrogen efficiency (10.6%), but at the expense of a greater LCOH. Also Sanz-Bermejo et al.
[14] delved into the coupling of SOEC with a 10 M We direct steam generation solar tower plant; several
configurations were evaluated, in order to find out the optimal integration strategy, in terms of overall
hybrid plant performance.
This work focuses on small scale hydrogen production, specifically conceived for mobility application,
since the plant size is that of a typical hydrogen refuelling station. The idea is to pursue the concept of
distributed generation also when dealing with hydrogen production. In particular the aim is to design
a solar driven hydrogen production plant, by coupling the Solid Oxide Electrolysis technology to a
parabolic dish system. In this way, the high modularity of both technologies could be wisely exploited
and combined. To gain insight into the operation of the SOEC within the plant, a 1D cell model is studied,
enabling a detailed description of the cell from the electrochemical and thermodynamic standpoint. A
techno-economic optimization is subsequently performed to assess the optimal design conditions of the
plant.
2 Methodology
2.1 SOEC modeling
The model of the SOEC is built in Aspen Custom Modeler. It is one dimensional along the cell axis and
reproduces a co-flow planar configuration of the cell. The cell, schematized as depicted in figure 1, is
discretized axially in a number of unit elements; in each of them, the electrochemical, mass and energy
equations are solved.
The assumptions under which the model is developed are:
i
Figure 1: Representation of the SOEC cell model [6]
Fuel (f u) electrode and sweep (sw) electrode designations are used here instead of cathode and anode,
to avoid confusion. For writing the model equations, the reaction is considered in this direction
1
H2 + O2 −→ H2 O (1)
2
Hence to simulate the model in the electrolyzer mode, and not in the fuel cell one, a negative current
density must be specified.
Electrochemical model
It computes the cell voltage as
In the expression, the definition of ohmic, activation, and concentration overpotentials is included, which
are responsible for the deviation of the cell voltage from VN ernst , called Nernst potential.
VN ernst is expressed by the Nernst’s law:
1
∆G RTpen xH2 · xO2 2 RTpen psw
VN ernst =− + ln + ln (3)
2F 2F xH2 O 4F p0
where the resistances related to the pen structure are calculated as the ratio of the thickness to the
conductivity of the considered layer; Rc accounts for possible contact resistances.
The total activation overpotential includes contributions from both the fuel and sweep electrodes;
the model presented by [17] is implemented. The modified Butler-Volmer equation, which describes the
functional relationship between the activation loss and current density, is
(βa,f u + 1)F ∆Vact,f uel βc,f u F ∆Vact,f uel
j = j0,f u exp − exp − (5)
RTpen RTpen
βa,sw F ∆Vact,sweep βc,sw F ∆Vact,sweep
j = j0,sw exp − exp − (6)
RTpen RTpen
j0,f u and j0,sw are the exchange current densities, expressed as
ii
βa,f
2
u 1− βa,f
2
u βa,sw
2
pH2 pH2 O p O2
p∗
H p0 p∗O
∗ 2 ∗ 2
j0,f u = jH 2 21 j0,sw = jO 2 21 (7)
pH2 p O2
1+ p∗ 1+ p∗
H 2 O
2
∗ ∗
The following Arrhenius expression are employed for describing and jH 2
jO 2
∗ Eact,f u ∗ Eact,sw
jH = γ f u exp − j O2 = γ sw exp − (8)
2
RTpen RTpen
Expressions for the equilibrium pressures p∗H2 and p∗O2 can be found in [17] [7].
The total concentration overpotential is defined as the difference between the theoretical Nernst
potential eq. (3), which is based on the bulk flow chemical composition, and the real Nernst potential,
calculated with the TPB molar fractions of the species. By rearranging, it is possible to recognize two
terms, related to the fuel and sweep sides
RTpen xH2 · xH2 O,T P B
∆V conc,f uel = ln (9)
2F xH2 ,T P B · xH2 O
RTpen xO2
∆V conc,sweep = ln (10)
4F xO2 ,T P B
The computation of the species molar fraction at cell reaction sites is performed as done by [1].
The model is validated with the calibration presented by [9]. Experimental tests were performed by
[5] on a planar button SOFC produced at Risø National Laboratory, constituted by a NiO/YSZ fuel
electrode, a YSZ electrolyte and a LSM/YSZ sweep electrode. Input data to the model are summarized
in table 1.
Cell geometry
δf u [µm] 310
δsw [µm] 10
δel [µm] 10
L [m] 0.01852
W [m] 0.00926
Hf u ch [m] 0.001
Hsw ch [m] 0.001
Material electrical properties
σf u [S/m] 107 / Tpen exp(1150/Tpen )
σsw [S/m] 4.6 · 106 / Tpen exp(1100/Tpen )
σel [S/m] 3.6 · 107 / Tpen exp(-8 · 104 /RTpen )
Kinetic parameters
βa 0.5
βc 0.5
Exchange current density fitted parameters
Eact,f u [kJ/mol] 87.4
Eact,sw [kJ/mol] 88.75
γf u [A/m2 ] 3.50428 · 108
2
γsw [A/m ] 1.69850 · 108
Electrodes’ microstructural parameters
φ 0.35
τ 3
rpore [µm] 0.15
iii
1.6
1.4
1.2
Voltage [V]
0.8
Exp 750
0.6 Exp 850
Exp 950
0.4 Mod 750
Mod 850
0.2 Mod 950
0
0 0.5 1 1.5 2 2.5 3 3.5 4
Current density [A/cm 2 ]
Figure 2: Model validation. Dashed lines are the simulated polarization curves; experimental data are marked
with dots
Three polarization curves are fitted, related to three set of experiments, performed at different tem-
peratures and same pressure (1 bar). Looking at figure 2, it is observed a good agreement between
numerical results and measured data; consequently the model input parameters succeed in accurately
matching the simulated cell performances with the real one.
The streams entering the electrode compartments of the SOEC stack are preheated to the design tem-
perature (750 ) in a single stream. The feed water passes first through a heat exchange section (EE),
then is directed to the solar plant (simulated by a heater block), where superheated steam is produced.
By way of a splitter, the amount of steam constituting the sweep gas flow is set; the remainder of steam,
with the addition of the hydrogen reflux, creates the fuel flow to the fuel channel. In the SOEC, the
electrolysis reaction takes place and the output flows are a mixture of H2 and H2 O at the fuel electrode,
and steam enriched with O2 at the sweep electrode side. Both streams represent the hot streams in the
heat exchange section. At the exit of the economizer, they are further cooled in two different condensers.
The hydrogen reflux is partitioned before the condenser, it is re-compressed through a H2 recirculation
blower and re-heated in a separated heat exchanger (HR). Two adiabatic flashes allow for the condensate
separation, which is then recirculated. On the other hand, the hydrogen and oxygen flows are sent each
one to an intercooled compression unit and attain the storage pressure of 200 bar. A feed water pump,
electrically driven, is needed to recover the plant pressure drops.
The absence of a sweep gas allows simplifying the plant arrangement, with great potential for cost
reduction. The oxygen separator is not necessary anymore and the two condensers are removed as a result
of the lower temperature levels achieved at the outlet of the economizing section. The reflux of hydrogen
is split after the hydrogen separator, ensuring a single phase stream approaching the H2 blower.
The SOEC cell employed in the plant is 594 cm2 active area. For each operating condition, the total
number of cells required to produce the design hydrogen capacity is obtained by iterative process. As
regards the inlet flows to the SOEC, the stream at the fuel electrode is a mixture of steam and hydrogen,
with molar fractions of 0.9 and 0.1 respectively; its inlet temperature depends on the temperature of the
iv
Figure 3: Schematic of plant layout in case of sweep gas
reflux. The sweep consists of steam at 750 , whose mass flow rate is set to 0.5 kg/h per cell in the sweep
gas case; the no sweep gas case instead is simulated by executing the SOEC submodel with a very little
steam flow rate. Further plant input specifications, concerning the considered block units, are gathered
in table 2.
It is decided to use the ideal equation of state except in the compression train, where the species deviation
from ideality is supposed to be more evident. The Peng-Robinson method is chosen accordingly.
For atmospheric operation, the pump discharge pressure is set in order not to run any component
in sub-atmospheric condition (the point of minimum pressure is corresponding to the flash separators).
For pressurized operation, the pressure at inlet of the SOEC is set to 5 bar, aiming for cell efficiency
optimization.
v
Solar field dimensioning
It is taken into account the integration with a parabolic dish system, which is designed to deliver both
electricity Pel,tot and heat Qth,tot required by the hydrogen production plant. Specifically, a multi-dish
configuration is decided, where a number of solar dishes are devoted to the electric supply and they are
labeled as ’electric dishes’, while at least one solar dish (’thermal dish’), upstream the SOEC unit, has
the task to bring the steam to 750 . Two different approaches are adopted:
1. Solar only: the dimensioning is aimed at providing the annual energy requirement of the plant,
only by exploiting the solar energy source. Calculations are performed on a yearly energy basis,
starting from the Annual Direct Normal Irradiation of the chosen site; it is supposed that Pel,tot
and Qth,tot are mean powers.
2. Hybrid : the dimensioning is aimed at assuring the plant power needs in a specific design condition
(DN I 900 W/m2 , Tamb 35 , SF 0); to fulfill the overall plant duty in the other conditions, it is
supposed an integration with a natural gas source. Calculations are performed on a power basis.
Depending on the values of solar dish related efficiencies (solar to thermal, solar to electric) a base case
and a best case are performed, as clarified in table 3.
Table 3: Base and best case efficiencies for solar only (yearly values) and hybrid approach (nominal values)
where at the numerator there are the actualized capital expenditure, O&M expenditure and tax contri-
bution (computed on the net income), while the term at the denominator is the present value of the total
energy output by H2 generation.
vi
Table 4: Equipment cost data
the combination of the Nernst potential decrease and the high heat capacity lets enlarge operation in
endothermic mode, with an advantageous decrement of the electric power absorbed by the SOEC itself.
The derivation of cell internal profiles reveals how the evolution of polarization losses and their mag-
nitude directly reflects on the current density distribution (figure 5(b)) and the thermal characteristic of
the SOEC. Figure 5(a) shows the temperature profiles relative to the anodic and cathodic streams and to
the pen structure, at two different values of current density. At low current density, the behaviour of the
SOEC is endothermic because the endothermicity of the electrolysis reaction prevails over the heating
associated with polarization losses. As a result, a decrease in the plotted temperatures is observed. A
higher current density contribute to boosting irreversibilities within the cell, therefore both streams, as
well as pen, are heated up by the cell heat generation.
780 0.45
Uf 0.4
775 Uf 0.6
0.4
Uf 0.8
770
Current density [A/cm2 ]
765
T sweep 0.3 A/cm 2
760 T pen 0.3 A/cm 2 0.3
T fuel 0.9 A/cm 2
755 T sweep 0.9 A/cm 2 0.25
750 T pen 0.9 A/cm 2
iso T cell 0.55 A/cm 2 0.2
745
0.15
740
735 0.1
0 0.002 0.004 0.006 0.008 0.01 0.012 0.014 0.016 0.018 0 0.002 0.004 0.006 0.008 0.01 0.012 0.014 0.016 0.018
Cell length [m] Cell length [m]
(a) (b)
Figure 5: (a) Temperature profiles of pen, fuel and sweep streams at 0.3 A/cm2 and 0.9 A/cm2 - Tin streams 750
, Uf 0.6, msweep 0.01 kg/h; (b) Current density profiles and variation with steam utilization
Looking at figure 5(b), at the very beginning of the cell, as soon as the reacting steam is consumed,
vii
overall the losses get smaller and the current density appreciably increases in order to keep a constant
cell voltage. Then it has a decreasing trend coherently with the reduction of steam concentration at the
fuel side from inlet to outlet. At higher steam utilization, expressed as
jAactive
nH2,prod 2F
Uf = = (12)
nH2 O,in xH2 O · nH2 O,in
a larger amount of hydrogen is generated at the inlet of the cell. This because in the second half of the
cell, activation at sweep electrode and concentration loss at the fuel electrode are prevailing and, along
with the Nernst voltage, act such that current density decreases more at high Uf .
As a result of a sensitivity analysis on the sweep gas flow rate, a map of the cell net thermal flux is
obtained (figure 6(a)), which clearly identifies the different thermal conditions of the SOEC. The isoline
corresponding to QSOEC = 0 represents the thermoneutral, featuring a cell voltage equal to 1.285 V . This
condition occurs progressively at higher current densities as steam, employed as sweep gas, is increasing;
at high sweep flow rate, a more rapid rise of the heat generated by polarization losses leads to a slight
reduction in the current density at the thermoneutral. On the right of the plot, the cell functioning is
above the thermoneutral voltage, QSOEC becomes negative, thus the cell needs to be cooled in order
to maintain an isothermal condition. Conversely, on the left, the cell shows an endothermic behaviour,
therefore it requires a thermal input to carry out the electrolysis at constant temperature; a maximum
in the net thermal flux is realized for each sweep gas flow rate considered.
0.08
0.02
0.06
0
0.1
0.02 1.4
0.11
0.01 0.13
0.005
Sweep flow rate [kg/h]
1.3
5
-0.2
-0.2
-0.3
-0.1
-0.4
0 0.02
-0.5
-0.6
-0.0
0.06
0.1
Qsoec [W/cm 2 ]
1e-3 1.2
Voltage [V]
-0.4
1.1
1e-4
0.08
-0.6 1
0.02
0.06
0.9 1 bar
1e-5
0
-0.8 3 bar
-0.8
-0.05
0.06
0.02
-0.6
-0.5
-0.1
-0.4
-0.3
-0.2
0.8 5 bar
-1
10 bar
-1
5e-7 0.7
0 0.3 0.6 0.9 1.2 1.5 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
2 2
Current density [A/cm ] Current density [A/cm ]
(a) (b)
Figure 6: (a) Map of QSOEC at 750 ; (b) Polarization curves at different values of pressure
As last observation, pressurized cell operation is simulated: results (figure 6 (b)) show that polarization
curves at high pressure cross the one at atmospheric pressure, meaning that at higher current density
(beyond the intersection point), the decrease in the overpotentials prevails over the Nernst voltage increase
and thus succeeds in reducing the cell voltage.
viii
Figure 7: Number of cells obtained for each current density. Values refer to the 60% Uf simulation case
SOEC absorbs more and more electricity to increase its hydrogen production rate. As for the steam
utilization effect, it decreases with Uf elevation, because the negative variation of the electric consumption
of the equipment (except for SOEC), mainly the intercooled compressor of H2 , overtakes the increasing
power absorbed by SOEC. Viceversa, the latter is minimized at the highest current density. In fact
at high j, due to the exothermic behaviour of the SOEC, the outlet streams from the sweep and fuel
channels are hotter and more heat can be transferred to the feed water in the heat exchange section, so
that the stream going to the solar section is characterized by a higher vapor fraction. The lowest power
level occurs at the highest Uf because of the lower amount of working fluid that the plant is processing.
Solar only The solar to hydrogen efficiency, whose map is provided in figure 8(a), is expressed as
mH2 ,target LHVH2
ηsun−to−H2 ,year = (13)
Pel,tot Qth,tot
+
ηtot,sun−to−el,year ηtot,sun−to−th,year
where ηtot,sun−to−el,year and ηtot,sun−to−th,year are the yearly averaged conversion efficiencies of the elec-
tric dish and thermal dish respectively, based on the effective Annual Direct Normal Irradiation. This
efficiency is representative of the yearly energetic performance of the plant.
Coherently with the previous discussion, best efficiency points appear at the highest steam utilization.
A maximum of 12.5% is found at a medium value of current density (0.4 A/cm2 ), because the contribution
of electric power at the denominator weighs more than the thermal power: at high current density, the
increment in the electric power requirement slightly prevails over the thermal power reduction. In a best
case scenario, an upgrade by 30% is observed, compared to the base case values.
ix
80 0.179
0.216 80
0.216
0.216
0.178
0.214
Solar to hydrogen efficiency [-]
Figure 8: (a) Yearly solar to hydrogen efficiency map for solar only best case; (b) Nominal solar to hydrogen
efficiency map for hybrid best case - no sweep gas, atmospheric
At the denominator, two contributions appears, the first being the annual solar energy, the second the
fuel energy associated to the natural gas consumption. For the best efficiency point, the resulting hybrid
efficiency is 18.4% in the base case and 26.4% in the best case.
In the sweep gas case, the need to heat the steam employed as sweep gas leads to demand a greater
duty from the thermal solar dish section. In fact in the cooling process of the hot streams exiting the
SOEC, the latent heat is inevitably lost and needs to be delivered in part by the thermal dish once more.
Consequently, lower solar to hydrogen efficiencies are obtained: 11.4% that raises to 12.2% in the best
case (solar only approach).
Pressurized operation
Adopting the same scheme of the atmospheric operation section, design performance maps are derived,
from which it is possible to appreciate the improvement in the solar to hydrogen efficiency. The reason is
that not only the electrochemical performance of the SOEC itself is enhanced, as already pointed out in
the Model results, but also the plant can take advantage of the pressurized functioning, since the electricity
consumption linked to the compression blocks reduces as well, and the evaporation and condensation of
water are thermodynamically favored. Now, for the well performing no sweep gas case, the trade-off
between the electric power increase and the thermal power decrease at high j leads to locate the best
solar to hydrogen efficiency point at 0.6 A/cm2 . The following table 5 gathers the resulting values of
solar efficiency for the best efficiency point. Also the needed number of electric dishes is specified (fixing
a diameter of 12 m) while only one thermal dish is sufficient to fulfill the heat duty required from the
plant.
From an efficiency standpoint, the system turns out to be competitive if compared to other solar hydrogen
technologies, like PEM coupled with photovoltaics [15], or SOEC coupled to a solar tower system [8].
Table 5: Solar efficiencies and number of electric dishes for solar only and hybrid approach in base case and best
case scenarios - no sweep gas, pressurized. Values refer to the best efficiency point
x
3.3 Economic results
Table 6 summarizes the results of the economic assessment. In BS, the lowest values of LCOH occur for
A B C D
Sweep gas - - • •
Pressurized - • - •
j [A/cm2 ] 0.4 0.6 1 1
Uf [%] 80 80 80 80
eCO2 [kgCO2 /kgH2 ] 20.2 20.1 23.6 22.1
BS cost
TP [Me] 1.113 1.010 1.177 1.096
CO&M,f ixed [Me/y] 0.071 0.062 0.067 0.064
CO&M,var [Me/y] 0.105 0.104 0.122 0.114
LCOH [e/kg] 7.0 6.5 7.45 6.98
OS cost
TP [Me] 0.820 0.752 0.920 0.853
LCOH [e/kg] 5.6 5.3 6.3 5.9
cases B and D, confirming that operation under pressure is a promising technical solution for making
the technology more economically viable. B takes advantage of the reduced plant complexity which
characterizes the no sweep layout, whereas in D, the employment of a larger number of components is
offset by the higher operating current density. The major contribution to the total investment cost is
from the solar dish section (above 70%), being the mirrors and the thermal receiver the dominant cost
items. Moving to OS, the LCOH decreases, as expected, achieving a minimum value of 5.3 e/kg in B:
the capital cost is in fact affecting by about 40% the LCOH, immediately followed by the variable O&M
costs, while fixed O&M costs represent around 16% of the final cost of hydrogen. The substantial impact
of variable O&M acts such that the LCOH decrease is dampened. Nearly 64% of the thermal energy
input is provided by the natural gas source, to ensure the target volume production, and this is reflected
in the specified carbon dioxide emissions. The cost of hydrogen in this study is in line with the outcomes
present in literature [8] [10].
4 Conclusion
In this paper, the coupling between SOEC and a parabolic dish system for solar hydrogen production was
investigated. Focusing in principle on the SOEC modeling, a 1D cell model was defined and subsequently
applied to generate useful results for any relevant electrochemical and thermodynamic variable, enabling
the discussion of the effects of different operating conditions on the internal SOEC behaviour.
Afterwards, the overall plant model was developed on Aspen Plus, in which the SOEC unit was
incorporated as a submodel. In pursuit of solar to hydrogen efficiency optimization, different operating
nominal conditions of the SOEC within the plant were examined, in terms of pressure, current density,
steam utilization and sweep gas flow rate (steam), with the objective to produce a hydrogen throughput
of 150 kg/d. It was found that the overall system conversion efficiency could be enhanced by (1) sweep
gas minimization, (2) high steam utilization, (3) pressurization, (4) medium current density values (in
the range [0.4 0.6] A/cm2 for the no sweep gas case), corresponding to a voltage for each cell of about
1.3 V .
Hence it is possible to state that the most suitable set of operating conditions is always a trade-off between
many aspects, connected not only to the SOEC stack but to all the balance of plant involving it. The
best solar to hydrogen efficiency is found to be 18.9%, on a yearly averaged basis, corresponding to the
no sweep pressurized scenario, accordingly. Within the hybrid approach, a solar to hydrogen efficiency
xi
of 22.9% is attained in design conditions, leading to a hybrid efficiency of 26.5%, on a yearly basis.
Results of the economic analysis showed that pressurization and high current density operation to-
gether move in the direction of a more affordable solar hydrogen generation. However the resulting value
of LCOH (5.3 e/kg) confirms that the proposed system is still not competitive if compared to steam
reforming hydrogen production, mainly due to the minor technological maturity of the parabolic dish
system; nevertheless it is in line with other solar hydrogen solutions. Specific carbon dioxide emissions
are around 20 kgCO2 /kgH2 for the presented cases, which are greater than typical values from steam
methane reforming (∼ 10 kgCO2 /kgH2 ). If a TES was incorporated, the hybrid system could exploit
the solar source to a greater extent and, consequently, CO2 emissions could experience a substantial
reduction.
Off-design analysis, that will be the subject of future works, could deepen the feasibility of a totally
renewable solar hydrogen production, by examining the coupling of the solar energy distribution with
typical hydrogen demand profiles.
xii
References
[1] S. Campanari and P. Iora. Definition and sensitivity analysis of a finite volume SOFC model for a
tubular cell geometry. Journal of Power Sources, 132(1-2):113–126, 2004.
[2] S. Campanari and E. Macchi. Technical and Tariff Scenarios Effect on Microturbine Trigenerative
Applications. Journal of Engineering for Gas Turbines and Power, 126(July 2004):747–757, 2004.
[3] A. Giostri and E. Macchi. An advanced solution to boost sun-to-electricity efficiency of parabolic
dish. Solar Energy, 139:337–354, 2016.
[4] B. D. James and D. DeSantis. Manufacturing Cost and Installed Price - Analysis of Stationary Fuel
Cell Systems. Technical Report September, 2015.
[5] S. H. Jensen, P. H. Larsen, and M. Mogensen. Hydrogen and synthetic fuel production from renewable
energy sources. International Journal of Hydrogen Energy, 32(15 SPEC. ISS.):3253–3257, 2007.
[6] P. Kazempoor and R. J. Braun. Model validation and performance analysis of regenerative solid
oxide cells: Electrolytic operation. International Journal of Hydrogen Energy, 39(6):2669–2684, 2014.
[7] P. Kazempoor and R. J. Braun. Model validation and performance analysis of regenerative solid
oxide cells for energy storage applications: Reversible operation. International Journal of Hydrogen
Energy, 39(11):5955–5971, 2014.
[8] M. Lin and S. Haussener. Techno-economic modeling and optimization of solar-driven high-
temperature electrolysis systems. Solar Energy, 155:1389–1402, 2017.
[9] V. Menon, V. M. Janardhanan, and O. Deutschmann. A mathematical model to analyze solid oxide
electrolyzer cells (SOECs) for hydrogen production. Chemical Engineering Science, 110:83–93, 2014.
[10] A. Mohammadi and M. Mehrpooya. A comprehensive review on coupling different types of elec-
trolyzer to renewable energy sources. Energy, 158:632–655, 2018.
[11] A. Mohammadi and M. Mehrpooya. Thermodynamic and economic analyses of hydrogen production
system using high temperature solid oxide electrolyzer integrated with parabolic trough collector.
Journal of Cleaner Production, 212:713–726, 2019.
xiii
[16] A. UK. DECARBit: Enabling advanced pre-combustion capture techniques and plants - European
best practice guidelines for assessment of CO2 capture technologies. 2011.
xiv
Contents
Abstract iii
Sommario v
Introduction 1
3 SOEC model 33
3.1 Electrochemical model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 34
3.2 Mass and energy balances . . . . . . . . . . . . . . . . . . . . . . . . . . . 39
3.3 Model validation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 42
4 Model results 47
4.1 Sweep gas type . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 47
4.2 Cell internal profiles . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 53
4.3 Sweep flow rate sensitivity analysis . . . . . . . . . . . . . . . . . . . . . . 60
4.4 Pressurized cell . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 65
i
5 Plant model 69
5.1 Description . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 69
5.2 Assumptions and design . . . . . . . . . . . . . . . . . . . . . . . . . . . . 71
5.3 SOEC operation within the plant . . . . . . . . . . . . . . . . . . . . . . . 73
5.4 Solar field dimensioning . . . . . . . . . . . . . . . . . . . . . . . . . . . . 75
5.4.1 Solar only approach . . . . . . . . . . . . . . . . . . . . . . . . . . 78
5.4.2 Hybrid approach . . . . . . . . . . . . . . . . . . . . . . . . . . . . 80
6 Plant results 87
6.1 Atmospheric operation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 87
6.1.1 No sweep gas case . . . . . . . . . . . . . . . . . . . . . . . . . . . 87
6.1.2 Sweep gas case . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 101
6.2 Pressurized operation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 109
6.2.1 No sweep gas case . . . . . . . . . . . . . . . . . . . . . . . . . . . 109
6.2.2 Sweep gas case . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 114
6.3 Solar field considerations . . . . . . . . . . . . . . . . . . . . . . . . . . . . 118
6.4 Efficiency comparison with other solar hydrogen technologies . . . . . . . 123
Conclusion 137
Bibliography 141
ii
Abstract
The global energy system is going towards the decarbonization across all sectors, in order
to accomplish the requirements in the Paris Agreement. In this scenario, hydrogen plays
a key role as it constitutes a clean and flexible energy carrier. Alongside its conventional
uses in the refinery and chemical industry, more recent applications include the produc-
tion of synthetic methane, the hydrogen storage for the integration of renewable energy
into the grid and the use as fuel for Fuel Cell vehicles and buses. The roadmap for
the hydrogen uptake passes through the development of near-zero emission technologies
for hydrogen production. This is the rationale for the investigation of solar hydrogen
technologies. In particular, the coupling between High Temperature Electrolysis (HTE)
and a Concentrating Solar Power (CSP) system is analyzed in this study. The proposed
plant is conceived to supply hydrogen for a small refueling station in Lancaster (CA).
It is based on the Solid Oxide Electrolyser Cell (SOEC) technology, which performs the
electrochemical water splitting, in order to produce a target of 150 kg/d of hydrogen.
The plant is integrated with a parabolic solar dish system, that is designed to provide
both electricity and thermal energy, necessary for the electrolysis reaction to take place.
Specifically a multi-dish configuration is selected, in which the share of electric power
is produced by an air Micro Gas Turbine placed in the dish focus, the share of thermal
power by means of a receiver that acts as a boiler, bringing the anodic and cathodic
streams to the cell operating temperature (750 ). To gain insight into the operation of
the SOEC within the plant, a 1D cell model is studied, enabling a detailed description
of the cell from the electrochemical and thermodynamic viewpoint. The plant process is
simulated with different conditions of the SOEC, in terms of current density, steam uti-
lization, pressure and sweep gas flow rate. A series of design performance maps are built,
accordingly. Furthermore, a preliminary economic analysis is carried out and the Lev-
elized Cost Of Hydrogen is subsequently obtained. The results of the techno-economic
study are used to find out the optimal design operating point.
Keywords Solar hydrogen, Parabolic dish, CSP integration, SOEC, 1D cell model
iii
iv
Sommario
v
vi
List of Tables
4.1 Input variables for simulation of streamwise molar fraction at fuel and
sweep channels . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 53
4.2 Input conditions for the sweep sensitivity analysis . . . . . . . . . . . . . . 60
6.1 Number of dishes for solar only approach in base case and best case sce-
narios - no sweep gas . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 97
6.2 Number of dishes for hybrid approach in base case and best case scenarios
- no sweep gas . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 101
6.3 Number of cells - sweep gas, atmospheric . . . . . . . . . . . . . . . . . . 101
6.4 Number of dishes for solar only approach in base case and best case sce-
narios - sweep gas . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 106
6.5 Number of dishes for hybrid approach in base case and best case scenarios
- sweep gas . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 106
6.6 Number of electric dishes for solar only and hybrid approaches in base
case and best case scenarios - no sweep gas, pressurized . . . . . . . . . . 114
vii
6.7 Solar efficiencies for solar only and hybrid approaches in base case and
best case scenarios - no sweep gas, pressurized . . . . . . . . . . . . . . . . 114
6.8 Comparison of solar efficiencies between pressurized and atmospheric op-
erations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 115
6.9 Number of dishes for solar only and hybrid approaches in base case and
best case scenarios - sweep gas, pressurized . . . . . . . . . . . . . . . . . 118
6.10 Results of shading calculation for the for the hybrid base case, in absence
of sweep . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 120
6.11 Results of shading calculation for the sweep atmospheric case . . . . . . . 122
viii
List of Figures
4.1 Comparison of polarization curves with air, steam and oxygen as sweep
gases . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 48
4.2 Separated overpotentials versus current density for each sweep gas case:
air (a), steam (b) and oxygen (c) . . . . . . . . . . . . . . . . . . . . . . . 49
4.3 Cell temperature variation with current density for the three sweep gas
cases . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 50
ix
4.4 Exchange current density at sweep electrode versus O2 molar fraction at
750 and 780 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 51
4.5 Nernst potential versus current density for the three sweep gas cases . . . 52
4.6 (a) SOEC efficiency and (b) electric efficiency . . . . . . . . . . . . . . . . 53
4.7 Molar fraction distribution at fuel and sweep channels . . . . . . . . . . . 54
4.8 Temperature profiles of pen, fuel and sweep streams at 0.3 A/cm2 and 0.9
A/cm2 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 55
4.9 Temperature profiles of interconnects at 0.3 A/cm2 and 0.9 A/cm2 . . . . 56
4.10 Variation of temperature profiles with fuel composition at (a) 0.3 A/cm2
and (b) 0.9 A/cm2 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 57
4.11 Current density profiles and variation with steam utilization . . . . . . . . 58
4.12 Separated losses contribution and Nernst potential in the streamwise di-
rection . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 59
4.13 Map of QSOEC at 750 . . . . . . . . . . . . . . . . . . . . . . . . . . . . 61
4.14 Maps of (a) Qreact and (b) Q at 750 . . . . . . . . . . . . . . . . . .
loss 62
4.15 Maps of QSOEC at (a) 700 and (b) 800 . . . . . . . . . . . . . . . . 64
4.16 Polarization curves at different values of pressure . . . . . . . . . . . . . . 65
4.17 Comparison of activation, concentration and ohmic overpotentials at 1
bar and 3 bar . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 66
4.18 VN ernst variation with pressure . . . . . . . . . . . . . . . . . . . . . . . . 67
4.19 Average cell temperature variation with current density, at different values
of pressure . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 68
6.1 Number of cells obtained for each current density - no sweep gas, atmo-
spheric . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 88
x
6.2 Mass flow rate and hydrogen content versus steam utilization for H2 re-
circulation stream and inlet fuel stream to SOEC . . . . . . . . . . . . . . 89
6.3 Cell voltage map - no sweep gas, atmospheric . . . . . . . . . . . . . . . . 90
6.4 Inlet fuel temperature to SOEC - no sweep gas, atmospheric . . . . . . . . 91
6.5 Outlet fuel temperature from SOEC - no sweep gas, atmospheric . . . . . 91
6.6 First law efficiency map - no sweep gas, atmospheric . . . . . . . . . . . . 92
6.7 Electric power contributions - no sweep gas, atmospheric . . . . . . . . . . 93
6.8 Power plant requirements - no sweep gas, atmospheric . . . . . . . . . . . 94
6.9 Feed vapor fraction variation with current density - no sweep gas, atmo-
spheric . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 95
6.10 SOEC efficiency map - no sweep gas, atmospheric . . . . . . . . . . . . . . 96
6.11 Thermal dish aperture area as function of Uf for solar only case- no sweep
gas, atmospheric . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 97
6.12 Solar to hydrogen efficiency map for solar only base case - no sweep gas,
atmospheric . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 98
6.13 Solar to hydrogen efficiency map for solar only best case - no sweep gas,
atmospheric . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 98
6.14 Thermal dish aperture area as function of Uf for hybrid case - no sweep
gas, atmospheric . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 99
6.15 Solar to hydrogen efficiency map for hybrid best case - no sweep gas,
atmospheric . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 100
6.16 Solar to hydrogen efficiency map for hybrid best case - no sweep gas,
atmospheric . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 100
6.17 Cell voltage map - sweep gas, atmospheric . . . . . . . . . . . . . . . . . . 102
6.18 Reaction thermal flux at thermoneutral as function of Uf - sweep gas,
atmospheric . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 103
6.19 Total electric power distribution - sweep gas, atmospheric . . . . . . . . . 103
6.20 Electric power variation with Uf - sweep gas, atmospheric . . . . . . . . . 104
6.21 Total thermal power distribution - sweep gas, atmospheric . . . . . . . . . 105
6.22 I law efficiency map - sweep gas, atmospheric . . . . . . . . . . . . . . . . 105
6.23 Solar to hydrogen efficiency maps for the solar only approach - sweep gas,
atmospheric . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 107
6.24 Solar to hydrogen efficiency maps for the hybrid approach - sweep gas,
atmospheric . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 108
6.25 SOEC efficiency map - no sweep gas, pressurized . . . . . . . . . . . . . . 110
xi
6.26 I law efficiency map - no sweep gas, pressurized . . . . . . . . . . . . . . . 110
6.27 Outlet fuel temperature from SOEC - no sweep gas, pressurized . . . . . . 111
6.28 Solar to hydrogen efficiency maps for the solar only approach - no sweep
gas, pressurized . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 112
6.29 Solar to hydrogen efficiency maps for the hybrid approach - no sweep gas,
pressurized . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 113
6.30 I law efficiency map - sweep gas, pressurized . . . . . . . . . . . . . . . . . 115
6.31 Solar to hydrogen efficiency maps for the solar only approach - sweep gas,
pressurized . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 116
6.32 Solar to hydrogen efficiency maps for the hybrid approach - sweep gas,
pressurized . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 117
6.33 Mean Shade Factor, Ground Cover Ratio and total land area as function
of Aspect Ratio . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 119
6.34 Aerial view of the rectangular solar field layout, for the selected no sweep
base case . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 120
6.35 Aerial view of the solar field layout in the best scenario and its 3D repre-
sentation at a specific time instant . . . . . . . . . . . . . . . . . . . . . . 121
6.36 Aerial view of the diamond solar field layout, for the selected no sweep
base case . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 122
6.37 Example of random layout, for the selected no sweep base case . . . . . . 123
xii
Nomenclature
Following are lists of (1) acronyms, (2) Greek letters, (3) Roman uppercase letters and
(4) Roman lower case letters adopted in this work. When appropriate, they are given
along with the dimensional units employed in this study.
xiii
∆Vact Activation overpotential K Convective heat transfer coefficient,
∆Vconc Concentration overpotential W/m2 /K
xiv
Introduction
The global energy system is going towards the decarbonisation across all sectors, in order
to accomplish the requirements in the Paris Agreement [34]. In this energy transition,
hydrogen plays a key role as it constitutes a clean and flexible energy carrier: it could
indeed be involved in power-to-gas applications, for direct injection into an existing
natural gas infrastructure, and could be used as fuel for Fuel Cell vehicles and buses in
power-to-fuel applications. Moreover it could serve as energy storage system, enabling a
larger share of renewable energy to be integrated into the electric grid. Hence, hydrogen
can link different energy sectors and energy T&D networks and increase the operational
flexibility of future low-carbon energy systems [32]. This concept is outlined in figure 1.
Figure 1: Representation of the energy system today and in the future [32]
The roadmap for the hydrogen uptake passes through the development of near-zero
emission technologies for hydrogen generation. In this respect, research on the coupling
between a water splitting technology and a solar energy system has been taking root,
1
Introduction
Focus is made on High Temperature Electrolysis, notably on the Solid Oxide Electrol-
ysis technology, in that it shows great potential for future cost reduction and efficiency
improvements. In fact, thanks to the operation at temperatures above 650 , SOEC
can take advantage of reducing the required electric energy for the process itself. In lit-
erature, a good deal of research have been devoted to the mathematical modeling of the
SOEC cell. Thanks to the reversible feature of the cell, which can be employed either in
fuel cell or electrolysis mode, previous works on SOFC modeling could be conveniently
exploited.
2
Udagawa et al. [75] developed a one-dimensional dynamic model of a cathode-
supported planar SOEC, consisting of an electrochemical model, a mass balance, and
four energy balances (for the different cell layers). This was then employed to study the
steady state behaviour of a SOEC stack at different current densities and temperatures.
Menon et al. [51] proposed a quasi-two-dimensional model in which the species
composition is resolved in 1D in the gas channels (along the direction of axial flow) as
well as in the porous media (transverse to the direction of axial flow). For each axial
position in the flow channel, the porous media is resolved across its thickness. The
simulation applies detailed models for electrochemical conversion at the three-phase
boundary, an elementary heterogeneous reaction mechanism for the thermo-catalytic H2
electrode chemistry, dusty-gas model to account for multicomponent diffusion through
porous media, and a plug flow model for flow through the channels. Also the performance
of SOEC in co-electrolysis mode, for syngas production, is studied in [50].
Ferrero et al. [23] developed a dynamic model for the simulation of planar rSOCs.
In the study, the performance of commercial-size cells made with different air elec-
trode materials (LSM/YSZ and LSCF) has been investigated and an integrated thermo-
electrochemical model has been first calibrated and then validated on the experimental
data. Other works devoted to rSOC modeling are available (1D - Kazempoor and Braun
[38] [39], Wendel et al. [78], 0D - Ni et al. [56]).
Among the solar hydrogen technologies, attention is drawn to the coupling between
SOEC and a concentrated solar system. The reason is that CSP offers the opportunity
to deliver both electricity and high temperature heat at the same site, so it can fulfill
the energy requirements of the SOEC, in terms of electric and thermal duty, as well. As
a result, the combination of hydrogen production by high temperature electrolysis and
concentrated solar thermal energy is very attractive since this could create synergy with
much gain for both technologies.
A variety of solar hydrogen solutions developed, with different operating conditions
and applications. Derbal-Mokrane et al. [21] proposed a system that integrates high
temperature electrolysis with a parabolic trough section, devoted to thermal energy pro-
duction; the electric input is instead provided separately by a PV power station. Water
is superheated to 350 by indirect heat exchange with the thermal oil circuit of the
parabolic trough field, and attains 900 , by the exploitation of the hot streams exiting
from SOEC and an auxiliary electric heater. Results, derived for different installation
sites in Algeria, showed that the power requirement for a hydrogen throughput of 400
kg/h is 5 M W thermal and 14 M W electric; about 14 ha area is necessary, 97% of which
3
Introduction
4
needed temperature step to achieve the electrolyser’s operating temperature (700 ).
The target of 400 kg/d of hydrogen is accomplished with this system and, depending
on the number of TES modules, it can reach values above 550 kg/d. Besides, it was
considered a scenario in which the SOEC can operate reversibly for grid balancing.
The feasibility of a solar tower system coupled with SOEC was deeply analyzed by
Lin and Haussener [45], providing evidence that, among different proposed scenarios,
this solution is able to work at high solar to hydrogen efficiency (10.6%), but at the
expense of a greater LCOH.
This work focuses on small scale hydrogen production, specifically conceived for mo-
bility application. The idea is to pursue the concept of distributed generation also when
dealing with hydrogen production. Following this outlook, hydrogen refueling stations
could be directly supplied on site, leading to the minimization of the distribution man-
agement and economic impact. More precisely, the present study aims at designing a
solar driven hydrogen production plant, by coupling the Solid Oxide Electrolysis tech-
nology to a parabolic dish system. In this way, the high modularity of both technologies
could be wisely exploited and combined. From a design perspective, a multi-dish con-
figuration is selected, in which the share of electric power is produced by an air Micro
Gas Turbine placed in the dish focus, the share of thermal power by means of a thermal
receiver, bringing the anodic and cathodic streams to the cell operating temperature.
To gain insight into the operation of the SOEC within the plant, a 1D cell model is
studied, enabling a detailed description of the cell from the electrochemical and ther-
modynamic standpoint. A techno-economic optimization is subsequently performed to
assess the optimal design conditions of the plant. Finally, the competitiveness of the
plant is evaluated in comparison with other solar fuels’ technologies.
The structure of the thesis is the following: Chapter 1 includes a general overview on
hydrogen applications and sources; water splitting technologies are then presented, with
major emphasis on electrolysis. Moreover a brief description on the types of concentrated
solar power systems is reported.
In Chapter 2 the operating principles of Solid Oxide Electrolysis cells are explained.
Afterwards, depending on the adopted CSP system type, solar hydrogen solutions are
presented in order to assess their compatibility with the high temperature electrolysis
stack.
Chapter 3 describes the one dimensional SOEC model. The equations are progres-
sively shown and commented, subdividing in electrochemical, mass and energy equations.
The validation of the model is then performed.
5
Introduction
In Chapter 4, the effects on the SOEC performance of central parameters, like tem-
perature, pressure, flow rate and compositions of the inlet streams, are studied through
numerical simulations at the cell level.
In Chapter 5, the plant layout is described in detail, accompanied by the input con-
ditions and assumptions. The approaches for the solar dish dimensioning are explained,
together with the corresponding computation procedure.
The plant process is simulated with different conditions of the SOEC, in terms of
current density, pressure, steam utilization and sweep gas flow rate. Chapter 6 presents
the results of these plant simulations.
Eventually, in Chapter 7, a preliminary economic analysis is carried out on selected
plant operating points and the Levelized Cost Of Hydrogen is subsequently derived.
6
Chapter 1
7
Overview on hydrogen and CSP technologies
Power-to-feedstock The largest share of hydrogen demand is from the chemical sec-
tor, for the production of ammonia primarily, and methanol. Furthermore the refining
industry requires large amounts of hydrogen, during hydrocracking and desulphurisation
processes, for the conversion of raw materials into refinery products.
Power-to-fuel In the transport sector, hydrogen can be used as fuel in the so called
Fuel Cell Electric Vehicles (FCEVs). The focus is on the use of pure hydrogen, even
though other pathways are feasible, for example via the use of synthetic methane in
Compressed Natural Gas (CNG) vehicles or through conversion to methanol. FCEVs
are basically electric vehicles using hydrogen stored in a pressurised tank and a fuel cell
for on-board power generation. To date, FCEVs are fuelled with gaseous hydrogen at
pressure of 350 bar (for cars) or 700 bar (for trucks and buses). The on-road fuel economy
is around 1 kg of hydrogen per 100 km travelled, and demonstration cars have ranges of
around 500 km to 650 km. Since the driving performance of FCEVs is comparable to
conventional cars and refuelling time is about the same, FCEVs can provide the mobility
service of conventional cars at much lower carbon emissions [32].
FCEVs are complementary to battery electric vehicles (BEVs): they indeed expand the
market for electric mobility to high duty cycle segments (long-range or high utilisation
rate vehicles, such as trucks, trains, buses, taxis, ferry boats, forklifts) where batteries
are currently limited. Embracing this vision, several global automakers, such as Toyota,
Hyundai, Honda and the Chinese car manufacturer SAIC, have begun commercialisation
in certain regions of the world, including Japan, California, Europe and China [34].
8
Chapter 1
Stationary application Distributed generation of electricity and heat using micro co-
generation systems enables an increment of the overall energy efficiency in the buildings
sector. Fuel cell micro co-generation systems powered by hydrogen or natural gas are
an alternative to conventional ICE systems. Currently, the electrical efficiency of fuel
cell micro co-generation systems is around 42%, being around 10 percentage points
higher than for the corresponding ICE [32]; this benefit is though counterbalanced by
the significantly higher investment cost of the former one.
Currently, over 95% of hydrogen production is fossil-fuel based (see figure 1.1). Steam-
methane reforming (SR) is the leading way of producing hydrogen. The process, being
based on a mature technology, is cost-effective (1-3 e/kg of H2 , including the cost of
CO2 sequestration) and highly efficient (70-80%) . Oil and coal gasification are also used,
especially in China and Australia, albeit to a lesser extent than SR. Only around 4%
of global hydrogen supply is produced via electrolysis [34]. In the midst of the struggle
with the dependency on fossil sources, conventional hydrogen production methods are
considered transitional technologies; water is widely agreed to be the most interesting
source of sustainable hydrogen, mainly because the process lends itself to be integrated
with renewable energy sources [77]. An examination of water splitting technologies is
offered in the next section.
9
Overview on hydrogen and CSP technologies
Water splitting consists of the dissociation of the water molecule in hydrogen and oxygen,
as shown below.
1
H2 O −→ H2 + O2 ∆H ◦ = 285.84 kJ/mol ∆G◦ = 237.21 kJ/mol (1.1)
2
The reaction is endothermic (∆H ◦ > 0) and non spontaneous (∆G◦ > 0). Electrical en-
ergy and/or thermal energy can be used to drive the dissociation. Different technologies
allow for the splitting of water:
Electrolysis
Photolysis
Thermolysis
Thermochemical process
Among them, electrolysis is the most promising and has already found commercial
application. Electrolysis, thermolysis and thermochemical processes could be poten-
tially coupled with a renewable technology, that provides the energy requirement for the
splitting process. In particular, the integration with a solar technology is considered,
leading to the so called solar hydrogen production. It follows a brief description of the
aforementioned technologies.
The following table (1.1) sums up the electrochemical reactions for each type of elec-
trolyzer, and specifies the adopted electrolyte and the range of temperatures at which
they operate. As a result a split into Low Temperature Electrolysis (LTE) and High
Temperature Electrolysis (HTE) can be done.
10
Chapter 1
HTE
MCEC Molten carbonate (liquid) Cathode semi-reaction 600 - 700
H2 O + CO2 + 2e− H 2 + CO32−
Ion transport
CO32−
Anode semi-reaction
CO32− 1/2 O 2 + CO2 + 2e−
SOEC Ceramic oxide (solid) Cathode semi-reaction 650 - 950
H2 O + 2e− H 2 + O2−
Ion transport
O2−
Anode semi-reaction
O2− 1/2 O 2 + 2e−
In the field of LTE, state-of-the-art PEM electrolysers can operate more flexibly and
reactively than current ALK technology, since they offer a wider operating range and a
shorter response time. The downside, however, is that PEM lifetime is reduced by half
(40000 h PEM versus 80000 h ALK) and efficiency is still lower than ALK (57% PEM
versus 65% ALK) [34]. Although alkaline electrolysers are a mature and affordable
11
Overview on hydrogen and CSP technologies
technology, SOEC shows a greater potential for future cost reduction and efficiency
improvements. Unlike the previous technologies, SOEC is able to operate at very high
temperature, which results in faster reaction kinetics and lower internal irreversibilities.
The expensive noble metal electrocatalysts (Pt particles, typically) and membranes that
characterize PEM are not required anymore. As regards the durability of SOEC, it is
improving: indeed, much work has been done in testing cells and optimizing electrode
and electrolyte materials for steam electrolysis ([27], [63]). For current densities up to
0.75 A/cm2 , operation with little or no degradation at all was attained by removing
impurities from the inlet gases [28]. Durable high current density operation is a vital
condition for the process to be affordable. In the family of HTE, MCEC and SOEC share
the strong advantage of being capable of performing also CO2 electrolysis for synthetic
fuel production. MCEC is not discussed here since it requires CO2 as reacting species in
addition to H2 O, but it is mentioned for completeness. For solar hydrogen production,
to properly match the solar technology with the energy requirement of electrolysers, a
PV device and electric heaters could be conveniently coupled with LTE, while CSP is the
renewable source that could be integrated in a HTE plant. A more detailed description
of the SOEC technology will be performed in the next chapter.
12
Chapter 1
must have a high quantum yield (>80%) across its absorption band; (3) the system
must be stable (against corrosion) in an aqueous environment.
Photolysis has only been tested at a laboratory scale and needs significant breakthroughs
to become a feasible method of dissociation.
The first step requires temperatures up to 2000 depending on the cycle. Fewer steps
results in lower losses associated with products’ separation, heat transfer and transfer of
13
Overview on hydrogen and CSP technologies
materials between each step. However, there are cycles with three or more steps with a
maximum temperature below 1000 , that thus succeed in drastically curtailing thermal
stress on materials. On the other hand, the higher the number of steps, the lower the
maximum theoretical process efficiency, because of the irreversibility of each stage [60]
and because more materials and heat transfer must be managed. In addition to this,
multi-step cycles often involve corrosive chemicals.
Hundreds of possible cycles have been identified for hydrogen production with a
variety of materials and maximum operating temperatures [11]. For example, some in-
vestigations have focused on cycles using ceria based materials, since they can be reduced
at a lower temperature (<1500 ), they might retain their microstructure through the
redox cycles, and both the oxidized and reduced phases remain solid. Despite detailed
studies on this technology, it has to face a number of critical issues: (1) expensive mate-
rials (or equivalently, short material lifetimes) associated with high temperatures, rapid
temperature transients, corrosive chemical intermediates, (2) difficult separations of the
chemical intermediates, (3) energy losses across multiple steps from heat exchange, and
(4) undesired side reactions [28].
The solar-to-heat conversion efficiency is limited by re-radiation losses and the heat-
to-chemicals conversion efficiency is limited by thermodynamics, with further energy
losses from heat recuperation and from separation and quench steps. Practical efficiencies
for the net solar-to-chemicals conversion have been estimated in the range of 16-25%.
To improve efficiency, more effort has to be directed to the design of efficient and robust
heat recuperating solar reactors. This solution is expected to be a long-term technology
pathway, with potentially low or no greenhouse gas emissions.
A CSP system is based on the concept of concentrating the direct solar radiation on
a receiver; this receiver is able to transfer solar energy to a fluid, which is used in a
subsequently step as working fluid in conventional generators for electricity production.
A series of mirrors performs the concentration of the sunlight and, depending on their
shape, two main groups of CSP systems can be identified:
14
Chapter 1
onto a receiver pipe. This configuration allows a single axis tracking [17].
A brief description of them is here presented, together with a simple schematic of each
technology in figure 1.2.
Figure 1.2: Schematics of the four concentrating solar power technologies [65]
Solar dish A solar dish system (figure 1.3) consists of a parabolic mirror which reflects
the solar radiation on a receiver, integrated with a thermal engine located at the focal
point of the dish. These technology shows the highest CR (1000 up to 3000) and the
highest maximum temperature at the receiver (1500 ), leading to high conversion
efficiency. The mainly used heat engine is the Stirling engine: with a working fluid
temperature of over 700 and at pressure as high as 20 M P a, a solar to electric efficiency
of 29.4% can be achieved. The use of a Micro Gas Turbine has also been studied in more
recent years, with the aim of increasing the ratio of the electric power generated to the
solar energy collected and improving the operability in relation to solar energy short time
fluctuations [43]. The MGT exploits a recuperative Joule Brayton cycle, with typically
air as working fluid. The MGT has interesting characteristics, in particular high power
15
Overview on hydrogen and CSP technologies
to weight ratio, simplicity of design and reliability; another feature is the possibility
to exploit a higher receiver temperature (900 ) that matches with the Turbine Inlet
Temperature (TIT), leading to a higher cycle efficiency. In addition, parabolic dish
systems are modular and flexible but more expensive with respect to the other kinds of
CSP systems, therefore they are feasible for small scale and decentralized production.
Cost optimization indicates 10-12 m as maximum diameter of the reflector, therefore
limiting the single solar dish to a net power output of about 25-30 kW with a solar
radiation of about 1000 W/m2 [17].
Solar tower The solar tower technology (figure 1.4) utilizes a wealth of large, computer
controlled, sun-tracking mirrors (heliostats), to focus sunlight onto a central receiver,
mounted at the top of a tower. A heat transfer fluid, heated in the receiver, absorbs
indeed the highly concentrated radiation reflected by the heliostats and converts it into
thermal energy. This heat is usually coupled through a heat exchanger to a conventional
steam cycle to produce electricity. Solar towers typically stand about 75-150 m height;
they are best suited for large scale applications in the 10 to 200 M We range [65], in order
to take advantage of economies of scale. Being able to operate at a high temperature
level, this technology has high net solar to electric efficiency and is commercially proven.
In addition, it provides large heat storage capacities. The process design of the plant
depends on the adopted heat transfer fluid, which can be water, molten salt or air [71].
Parabolic trough The solar radiation is concentrated via parabolically shaped mirror
facets on a focal line, where the receivers (steel tubes) are placed inside an evacuated
16
Chapter 1
Figure 1.4: Solar tower plant from Gemasolar - Sevilla, Spain [70]
glass envelope to minimize convection heat losses. A heat transfer fluid, that could be
synthetic oil or molten salt, flowing in the receiver tube, brings thermal energy to a
steam generator, to produce superheated steam. This step is bypassed in the case of
direct steam generation (the heat transfer fluid is water). A conventional steam Rankine
cycle is generally coupled, exploiting the high temperature steam source to produce
electricity. The solar field consists of a number of collectors in parallel rows, aligned
with the tracking axis orientation. The aperture width is up to 12 m while the total
trough length can reach a value around 100-120 m. The characteristic size of a parabolic
trough plant varies between 15 and 100 M We and it can achieve an annual net solar
to electric efficiency around 14%. An example of parabolic trough system is shown in
figure 1.5.
Linear Fresnel Linear Fresnel reflectors (figure 1.6) incorporate long arrays of flat
mirrors that concentrate light onto a fixed linear receiver, above and along the arrays.
A heat transfer fluid circulates through the receiver, collecting and transporting thermal
energy to the power block (Rankine cycle typically) and storage tanks. This kind of
mirrors arrangement tries to approximate a parabolic trough collector shape, with in-
teresting advantages from the economic and practical viewpoints. The objective of this
technology is in fact to reproduce the performance of the parabolic trough collector tech-
nology but at lower costs. Linear Fresnel collector optimizes indeed material and cost
and simplifies the construction and installation thanks to (1) the use of cheaper flat glass
reflector, (2) direct steam production, (3) lower cost non-vacuum thermal absorbers, (4)
17
Overview on hydrogen and CSP technologies
light reflector structures (that do not have to support the receiver) close to the ground,
(5) minor land occupation, due to the reduced shading between rows, and (6) lower
O&M costs owing to easier access to primary reflectors. However the annual efficiency
of linear Fresnel is around 9% and thus lower than the one of parabolic troughs, because
the combination of a fixed and less expensive receiver and not curved mirrors results
into greater optical losses. The most common use of Linear Fresnel is for the direct
production of saturated steam at 270 , because of the use of non-vacuum absorbers,
featuring low efficiency at high temperature. At present, AREVA and NOVATEC are
looking for new absorbers able to work at temperatures above 450 [71].
18
Chapter 1
Figure 1.5: Parabolic trough from Andasol Solar Power project - Granada, Spain [70]
Figure 1.6: Linear Fresnel from Puerto Errado CSP project - Murcia, Spain [70]
19
Overview on hydrogen and CSP technologies
20
Chapter 2
This chapter focuses on SOEC electrolysis and solar hydrogen from HTE.
As can be seen in figure 2.1, the SOEC cell is composed of a cathode (fuel electrode) for
the hydrogen evolution reaction and an anode (sweep electrode) for the oxygen evolution
reaction, which are separated by a solid electrolyte, devoted to the O2− transport. These
three layers forms the so called pen structure (positive electrode, electrolyte, negative
electrode). The channels allow the H2 O/H2 mixture on the fuel side and the H2 O/O2
21
SOEC and integration with CSP
stream on the sweep side flowing through the cell; the flows bring the reacting species
to the active sites and incorporate the product species. Channels are enclosed by inter-
connects, which serve as electrical connectors between two adjacent cells and as physical
separators of the anode and cathode [7].
As already mentioned in table 1.1, these are the semi-reactions involved, together
giving the overall electrochemical reaction (equation 1.1).
1
Sweep electrode O2− −→ O2 + 2e− (2.2)
2
The reaction takes place at the Triple Phase Boundary (TPB), that is in a region where
three phases (gas, solid electrolyte, solid electrode) come together simultaneously [7].
Hence the simultaneous occurrence of gaseous, electronic and ionic species is essential
for the reaction to proceed. It is evident that TPB is of paramount importance since it
determines the overall cell performance. TPB strictly depends on material properties,
composition, particle size and sintering conditions, therefore research on the cell micro-
scale level is crucial.
The material composition of typical state-of-the-art SOECs is the following [7]:
A variety of alternative materials with improved properties for SOFCs are still under
development [73], many of which might be appropriate for use in SOECs [29].
Concerning the interconnects, they must show chemical stability in both reducing
and oxidizing atmospheres, high electronic conductivity while low ionic one and gas
tightness; moreover they have to thermally match the thermal expansion coefficient of
the electrodes [7]. For these reasons metallic interconnects are preferred over ceramic
ones [47]. Cr-based alloyed steels could be utilized; a number of protective coatings
22
Chapter 2
and deposition techniques are available to avoid chromium poisoning and oxide scales’
formation with high contact resistance.
∆H = W + Q = ∆G + T ∆S (2.3)
Two definitions can be introduced: the reversible voltage and the thermoneutral
voltage. The reversible cell voltage (Erev ) is the theoretical minimum potential required
for the electrolysis to occur. This voltage can be expressed as a function of ∆G by means
of
∆G
Erev = (2.4)
2F
Erev applies to the case in which pure reactants and products are separated, at standard
pressure. In practice, the incoming steam is mixed with some hydrogen in order to main-
tain reducing conditions at the cathode. It is also not desirable to run the electrolyzer
to 100% steam utilization, because localized steam starvation will occur, severely de-
grading the performance. On the oxygen-evolution side of the cell, a sweep gas is often
employed, so the oxygen partial pressure is decreased. In addition, the electrolysis sys-
tem can operate at higher pressure than the atmospheric one [58]. In order to account
for the range of gas compositions and pressures that occur in a real system, the Nernst
potential can be obtained from the Nernst equation
2
1
1
RT xH2 · xO2 p 2
VN ernst = −Erev + ln · (2.5)
2F xH 2 O p0
If the thermal energy T ∆S is provided by way of electricity, the actual minimum voltage
for the dissociation is the thermoneutral voltage Vtn , which can be obtained from the
following expression:
23
SOEC and integration with CSP
∆H
Vtn = (2.6)
2F
The energy consumption of the electrolysis process depends, nevertheless, on tem-
perature and pressure. Changes in these magnitudes have an influence on the above-
mentioned characteristic voltages.
temperature and it is very close in magnitude to hydrogen Lower Heating Value (LHV).
Note that in the liquid state (between 298 and 373 K), ∆H slightly reduces; above 373 K
the latent heat of water vaporization has to be accounted for in the total energy demand
of the process.
As a result, with increasing temperature, the proportion of total energy requirement
that can be supplied in the form of heat increases; the corresponding electric energy
consumption is considerably lower. However, a heat source with a temperature high
enough is essential in order to cover the thermal demand of the process.
24
Chapter 2
In addition to this, performing HTE has benefits from the standpoint of kinetics
and electrolyte conductivity, both of which improve: higher reaction rates and a lower
internal resistance are achieved at high temperature.
Pressure dependence Figure 2.3 shows the influence of pressure on the energy con-
sumption when operating at standard temperature (298.15 K). At this temperature and
within the pressure range (1-100 atm), the reaction takes place in the liquid state. The
electric energy demand ∆G increases with a logarithmic tendency as pressure increases.
On the contrary, the thermal consumption, T ∆S, decreases as the pressure rises. The
net energy consumption of the electrolysis reaction, ∆H, remains practically constant
despite the pressure increment [77]. This means that tweaking the amount of electric
or thermal energy needed by changing the operating pressure of the system could be an
interesting solution depending on the type of system integration explored.
Heat sources and sink are linked to two phenomena: (1) the electrochemical reaction,
(2) the polarization loss mechanisms.
25
SOEC and integration with CSP
(1) The steam reduction reaction is endothermic, therefore thermal energy is con-
sumed by the electrolysis reaction. The reaction heat flux Qreact [W/m2 ] is derived as
follows
j j
Qreact = T ∆S = (∆HR − ∆G) = j(Vtn − VN ernst ) (2.7)
2F 2F
∆S is the entropy change of the electrolysis reaction, which is dependent of the reactant
and product partial pressures and the operating temperature.
(2) The polarization heat flux Qloss [W/m2 ] is the heat generated by overpotentials,
in particular by Joule effect and by the rise of non-ohmic irreversibilities. It is given by
X
Qloss = j(V − VN ernst ) = j ∆Vlosses (2.8)
The net thermal flux is the difference between the abovementioned quantities
This means that, depending on the operating voltage, the net heat generation in the
SOEC may be negative, zero, or positive. This phenomenon is portrayed in figure 2.4.
The figure shows the respective internal heat sink/source fluxes associated with the
electrochemical reaction and the ohmic heating.
Figure 2.4: Variation with current density of Qreact and Qloss and resulting QSOEC at
800
26
Chapter 2
V <V N ernst
no water electrolysis can occur
V N ernst < V < Vtn QSOEC > 0
electrolysis is possible by adding heat, in order to maintain the isothermal condition
V =V tn QSOEC = 0
the cell is adiabatic, meaning that heat provided by irreversible losses is enough
to sustain the endothermic electrolysis reaction; the entire enthalpy change in the
electrolysis is exactly matched by the electrical energy input to the cell.
V >V tn QSOEC < 0
electrolysis takes place under heat dissipation to keep isothermal operation. The
power supplied to the cell is greater than the minimum required by the process
(Vtn ·j); the cell consumes an extra power QSOEC due to the cell losses, and becomes
apparent as heat thus increasing the process temperature unless the cooling system
evacuates this energy [77].
The thermoneutral voltage in the end separates the endothermic behaviour of the SOEC
(below Vtn ) from the exothermic one (above Vtn ). It is expected that sub-thermoneutral
operation goes in the direction of improving SOEC efficiency, because it allows to partly
replace the electrical energy consumption of the stack with a heat input. Not the totality
of this heat contribution must be supplied from external sources, since a share of it is
provided by internal irreversibilities. On the other hand, when lowering the operating
current density, a larger stack area is needed to support a give hydrogen production
rate, involving higher investment cost [75]. A trade-off is given and a techno-economic
analysis should be performed, to optimize the SOEC functioning.
27
SOEC and integration with CSP
the electric power consumption, the other accounting for the thermal power required for
the electrolysis reaction.
In the domain of electrolysers, it is customary to define a SOEC electric efficiency,
only by considering the electric power contribution at the denominator.
mH2,prod LHVH2
ηel,SOEC = (2.11)
Pel
As a result, the thermoneutral condition will be characterized by a unitary electric
efficiency, whereas ηel,SOEC will exhibit values greater than 1, for operation below the
thermoneutral voltage.
jAactive
nH2,prod 2F
Uf = = (2.12)
nH2 O,in xH2 O · nH2 O,in
It accounts for the steam conversion, and correspondingly for the hydrogen production,
at the fuel electrode, with respect to the amount of steam molar flow rate at the cathode
inlet. Uf is a central parameter for the SOEC operation, since it has an influence on
the cell voltage and efficiency. Furthermore, when inserting the SOEC within a plant,
its choice is fundamental for optimizing the overall system efficiency.
The energy needed to carry out the electrolysis process can be provided by renewable
sources. In particular, due to the need for high temperature, SOEC could be conve-
niently coupled with a Concentrated Solar Power system, which is able to deliver both
electricity and heat at the same site. The synergy generated by the combination of these
technologies makes this solution very attractive, with great potential of development in
the near future.
Deep research studies on the coupling between High Temperature Electrolysis and a
Concentrated Solar Power system have been carried out by SOPHIA, a European project
started on 2014 with the purpose of bringing forward the SOEC technology, for future
large scale production of solar hydrogen. Also co-electrolysis for syngas production has
been investigated.
In the following paragraphs a brief description of the integration of the four solar
energy concentrating technologies is presented, in order to assess their compatibility with
28
Chapter 2
the high temperature electrolysis stack. The information are included in the SOPHIA
public summary on most suitable schemes for solar energy integration [71].
Solar dish coupling with HTE The electricity required will be provided by the
Stirling engine. The heat rejection from the engine (a heat sink is usually needed to
cool down the engine) will be used to evaporate the water for the electrolyser and to
heat the sweep gas, by means of a heat exchanger. Exhaust lines of the HTE stack are
coupled to a heat recovery system which heats up HTE inlet stream to the cell operating
temperature [71].
Solar tower coupling with HTE In the case of direct steam generation solar towers,
the power block can be designed for cogeneration of electricity and heat. Thus steam
purged from the turbine is used as heat source for the electrolyser in a heat exchanger
and solar heat is used to evaporate the water required as reactant for the electrochemical
reaction. A part of the solar steam is also used to heat the sweep gas [71]. The same
strategy could be adopted in an indirect configuration with molten salts processing in
the tower. Moreover a molten salt thermal storage makes possible to maintain the
electrolyser in stable thermal and electrical operating conditions. If an air solar tower
is considered, while a part of solar heated air powers a steam generator, a share of it is
directly used as sweep gas in the electrolyser unit.
Parabolic trough coupling with HTE As regards the electric power demand, this
will be provided by the thermodynamic cycle of the parabolic trough system (typically
a Rankine cycle). A part of the solar heat will be used to evaporate the water for the
electrolyser and to heat the sweep gas. In addition to this, a heat recovery system
downstream of the electrolyser will exploit the enthalpy content of the outlet streams
to pre-heat the HTE inlet streams. A thermal storage could be integrated for transient
operation. With steam as heat transfer fluid, the operating temperature is 400 at
30 bar; with synthetic oil instead, the maximum operating temperature is 390 and is
limited by the degradation of the fluid itself [71].
Linear Fresnel coupling with HTE Similarly to the parabolic trough system, by
coupling Linear Fresnel with the high temperature steam electrolyser, the electricity
required will be provided by the associated power block; thermal energy is collected by
the heat transfer fluid (water) in the absorber tubes and is partly recovered from the
29
SOEC and integration with CSP
exhaust lines of the HTE stack. With steam as heat transfer fluid, the outlet temperature
from the solar receiver is 270 or near 480 with new absorbers in development [71].
It can be deduced that parabolic trough and Linear Fresnel technologies are the least
suitable for the coupling with HTE because the maximum allowable temperatures from
the solar field are limited, or rather not high enough for typical operating temperatures of
SOEC stacks. They could nonetheless be usefully taken into account for low temperature
electrolysis. The temperature level is not an issue for solar tower and solar dish systems,
that thus possess the appropriate characteristics for the combination with HTE. The
solar tower would be the selected technology whether the aim were the integration in
a MW-scale hydrogen production plant. For small scale and decentralized hydrogen
generation, the choice on the solar technology to be integrated would differently fall on
the parabolic dish, as will be performed in the present study.
It is worth reporting the major attainments undertaken by SOPHIA; detailed infor-
mation is available in the Final Report Summary [4]. In the multitude of researches and
projects, attention was drawn to the optimization of the electrolysis system at the cell
and stack levels. Testing and modeling activities were performed in support of these
development lines. In particular SOEC cells were developed with tailored electrode mi-
crostructure and compositions [44], aiming to improve the electrode efficiency and to
reduce cell degradation (less than 1% per 1000 h). Based on the experimental campaign
and durability results on cells, short-stacks were designed and prototyped consequently.
In parallel, micro-scale to macro-scale models are developed aiming at understanding
the underlying mechanisms for pressurized electrolysis [16] and co-electrolysis. In a fol-
lowing step, efforts were made in developing an optimized CSP+HTE system. A 3 kW
prototype was coupled to a concentrated solar energy system and tested under realistic
conditions, for on-site proof of concept. A comprehensive testing activity was performed
in the solar simulator at DLR in Cologne. To provide the electrolyser with superheated
steam (mass flow rate in the range 1 to 5 kg/h) at 700 , a tubular solar receiver was
developed and tested in DLR’s high flux simulator (figure 2.5). Operating pressure was
limited to 4 bar in a first phase, then the system was tested up to 15 bar. The results of
the experimental campaign were used to develop process flow charts for a corresponding
MW-scale plant.
30
Chapter 2
(a)
(b)
Figure 2.5: (a) Part of the building of the high flux solar furnace in Cologne [2]; (b)
model of the utilized solar receiver [54]
31
SOEC and integration with CSP
32
Chapter 3
SOEC model
The model of the SOEC is built in Aspen Custom Modeler. It is one dimensional
along the cell axis and reproduces a co-flow planar configuration of the cell. The cell is
discretized axially in a number of unit elements; in each of them, the electrochemical,
mass and energy equations are solved. The adopted discretization method is the first-
order backward finite difference (BFD1).
The cell is schematized as depicted in figure 3.1, with the pen structure, the channels
on both electrode sides that bring the flow, and the interconnects which enclose the
channels.
33
SOEC model
Fuel electrode and sweep electrode designations are used here instead of cathode and
anode, to avoid confusion. Abbreviations for fuel and sweep gas flows in the model
equations are f u and sw, respectively.
The assumptions under which the model was developed are:
uniform cell voltage along the cell axis: the electrodes are considered as equipo-
tential surfaces due to high electrical conductivity of the electrode materials;
lumped pen temperature at each unit element (only resolved in the streamwise
direction);
cell constituted by one single channel;
negligible pressure gradient across the fuel and sweep channels;
mass transport occurs as equimolar counter diffusion at constant pressure.
Fluidynamic conditions in the channels are specified thereafter, when discussing the
energy balance section.
For writing the model equations, the reaction is considered in this direction
1
H2 + O2 −→ H2 O (3.1)
2
Hence to simulate the model in the electrolyzer mode, and not in the fuel cell one, a
negative current density must be specified.
Nernst potential The Nernst voltage is the theoretical minimum potential required
to dissociate water by electrolysis, below which electrolysis cannot proceed [7]. To derive
it, Nernst’s law is applied to the hydrogen oxidation reaction.
1
∆G RTpen xH2 · xO2 2 RTpen psw
VN ernst =− + ln + ln (3.3)
2F 2F xH 2 O 4F p0
∆G is the Gibbs free energy change of reaction; p0 is the standard pressure (1 atm).
34
Chapter 3
Ohmic overpotential Ohmic losses are due to the resistance to ionic and electronic
transport through cell layers. The overpotential scales directly with current density by
Ohm’s law.
Req,tot [Ωm2 ] is the equivalent ohmic resistance, which varies depending on the geometry,
the materials and operating conditions. It is given by
where Rsw ,Rf u ,Rel are the sweep electrode, fuel electrode and electrolyte resistances,
respectively; Rc is called contact resistance and it is a correction term introduced to
account for other possible resistance contributions, like interfacial contact between cell
layers [78]. The resistances related to the pen structure are calculated as
δi
Ri = ; i = sw, f u, el (3.6)
σi
where δi is the thickness, while σi the conductivity of the i layer. Since the materials for
the electrodes are characterized by high electronic conductivity, it is the ionic resistance
of the electrolyte that dominates.
Sweep electrode
βa,sw F ∆Vact,sweep βc,sw F ∆Vact,sweep
j = j0,sw exp − exp − (3.9)
RTpen RTpen
For each electrode, βa and βc are the symmetry factors in the anodic and cathodic
direction. j0,f u and j0,sw are the exchange current densities, expressed as
35
SOEC model
Fuel electrode
βa,f u 1− βa,f u
2 2
p H2 p H2 O
p∗H p0
∗ 2
j0,f u = jH 2 1 (3.10)
2
p H2
1+ p∗H
2
where
9.6 · 104
p∗H2 5
= 2.1362 · 10 exp − [atm] (3.11)
RTpen
Sweep electrode
βa,sw
2
p O2
p∗O
∗ 2
j0,sw = jO 2 1 (3.12)
2
pO2
1+ p∗O
2
where
200 · 103
p∗O2 8
= 4.9 · 10 exp − [atm] (3.13)
RTpen
p∗H2 and p∗O2 are equilibrium pressures and, for a LSM/YSZ interface, they are repre-
sented in Arrhenius form ([79],[39]); the pre-exponential value is in [atm], the value
inside the exponential is in [kJ/mol].
∗ and j ∗ are related to parameters of the pen structure such as TPB length, site
jH 2 O2
fractions and equilibrium constants of the elementary reactions. However, these param-
eters are difficult to measure and ask for more thorough models, therefore empirical
formulations are typically used. Therefore, to describe the temperature dependence of
exchange current density, the following Arrhenius expression is employed
Fuel electrode
∗ Eact,f u
jH = γf u exp − (3.14)
2
RTpen
Sweep electrode
∗ Eact,sw
jO = γsw exp − (3.15)
2
RTpen
Pre-exponential factor and activation energy are assigned empirically to fit the measured
polarization data.
Research is still facing the issue of accurately model activation overpotentials. An-
other simpler functional form, widely adopted in literature, is reported hereunder, with
the expression of the Butler-Volmer relation and the corresponding exchange current
density, for each electrode.
Fuel electrode
βa,f u 2F ∆Vact,f uel βc,f u 2F ∆Vact,f uel
j = j0,f u exp − exp − (3.16)
RTpen RTpen
36
Chapter 3
m n
pH2 pH2 O Eact,f u
j0,f u = γf u exp − m, n = const (3.17)
p0 p0 RTpen
Sweep electrode
βa,sw 2F ∆Vact,sweep βc,sw 2F ∆Vact,sweep
j = j0,sw exp − exp − (3.18)
RTpen RTpen
k
pO2 Eact,sw
j0,sw = γsw exp − k = const (3.19)
p0 RTpen
Nevertheless, it is decided to discard this model in favor of the first presented one. In
fact, when implemented to simulate the cell in electrolysis mode, it proves to underesti-
mate the contribution of activation losses, especially at high current density, where the
temperature effect is remarkable. This evidence is in agreement with literature findings
[38] [39], for which a major deviation from experimental data is observed by adopting
this activation submodel.
RTpen xH2 · xH2 O,T P B
∆V conc,f uel = ln (3.21)
2F xH2 ,T P B · xH2 O
RTpen xO2
∆V conc,sweep = ln (3.22)
4F xO2 ,T P B
xi is the mole fraction of species i in the bulk gas above the electrode surface, while
xi,T P B is the one at the TPB. The computation of the species molar fraction at cell
reaction sites is performed as done by [18]. For a reactant species, the path for diffusion
involves the diffusion from the bulk flow composition to the cell surface layer, then the
diffusion through the cell porous electrode to cell reaction sites; for product species, the
37
SOEC model
path is in the opposite direction. Fick’s law is implemented to derive the mole fraction
at the surface electrode and at the TPB.
Fuel electrode
RTf u j Hf u ch
xi,sur = xi ∓ · ; i = H2 , H2 O (3.23)
2F pf u Dbin,f uel 2
RTpen j
xi,T P B = xi,sur ∓ · δf u (3.24)
2F pf u Def f,i
Sweep electrode
RTsw j Hsw ch
xO2 ,sur = 1 + (xO2 − 1) · exp · (3.25)
4F psw Dbin,sweep 2
RTpen j
xO2 ,T P B = 1 + (xO2 ,sur − 1) · exp · δsw (3.26)
4F psw Def f,O2
The expressions for the diffusivities of the i species are reported below. In the channels,
only molecular diffusion takes place; binary diffusivities are calculated using an Aspen
procedure, given the species and the operating conditions (in terms of temperature and
pressure of the anodic and cathodic streams). In case of a mixture of N species, the
following mixing rule is applied to the molecular diffusion coefficients
1 − xi
Dmix,i = PN xj (3.27)
j6=i Di,j
Instead, in the porous structure, also the Knudsen diffusion (equation 3.28) is taken into
account besides the molecular one, since they are both significant. Molecular diffusion
is dominant when the pore size is much larger than the mean free path of the molecular
species, viceversa Knudsen diffusion becomes important when the pore size is much
smaller than the mean free path of the molecular species. Thus, for this case, the
diffusion process between the electrode surface and the TPB will be governed by both
molecule-molecule interaction and by molecule-pore wall interaction [56].
r
2 8RTpen
DKn,i = rpore (3.28)
3 πM Mi
Finally an effective diffusivity is calculated, by combining the two diffusion mechanisms
and by accounting for the tortuosity (τ ) and porosity (φ) parameters.
1 τ 1 1
= + (3.29)
Def f,i φ Dmix,i DKn,i
38
Chapter 3
Mass balance The composition of the anodic and cathodic streams evolves along the
cell due to the steam reduction and oxygen oxidation. This composition variation along
the channels is tracked by formulating the material balance for each species on both fuel
and sweep sides, and the boundary conditions at the cell inlet. The dependent variables
will be taken as the species concentrations. Although total moles are not conserved, the
individual species balance is equivalent to a mass balance [14].
Fuel electrode
∂Ci ∂(uf u Ci ) νi rred
=− + (3.30)
∂t ∂x Hf u ch
The differential equation comprises terms relative to storage, bulk flow and reaction.
The storage term is the time derivative of the species concentration, the bulk flow term
is the spatial derivative of the species concentration multiplied by the bulk fluid velocity
u, the reaction term is the reaction rate r times the stoichiometric coefficient νi . r is
linked to the local current density through Faraday’s law
j
r= (3.31)
2F
where 2 refers to the number of electrons transferred per electrochemical reaction, F is
the Faraday’s constant. The air channel mass balance is derived in a similar fashion.
39
SOEC model
Energy balance The derivation of the energy equations is simplified by some assump-
tions, that will be specified in this paragraph. As already said, a lumped temperature
in the direction transverse to the bulk flow is considered.
Kpen,f u [W/m2 /K] is the convective heat transfer coefficient between pen and fuel flow,
Kpen,sw between pen and sweep flow. These parameters are assumed to be equivalent
also for convection over the interconnects; they are derived assuming constant Nusselt
number and hydraulic diameter [75].
kpen and kint [W/m/K] are the thermal conductivities of pen and interconnects, respec-
tively; they are set at a constant value. kpen is calculated based on the anode, cathode
and electrolyte average thermal conductivity.
ρpen [kg/m3 ] is the density, while cp,pen [kJ/kg/K] is the heat capacity of the pen
structure; they are fixed values [75]. The same assumption holds for ρint and cp,int .
Table 3.1 lists the utilized values of the aforementioned quantities.
Fuel channel
∂Cf u hf u ∂Cf u uf u hf u j
=− − (hH2 − hH2 O )
∂t ∂x 2F Hf u ch
(3.37)
Kpen,f u Kpen,f u
+ (Tpen − Tf u ) + (Tint,f u − Tf u )
Hf u ch Hf u ch
The formulation contains the enthalpy flux of the bulk flow, the thermal energy associ-
ated with mass diffusion of reactants and products due to the electrochemical reaction
and the convective heat transferred to the solid phase (pen and interconnect).
40
Chapter 3
The same method as outlined above can be used to perform the energy balance in the
sweep channel.
Sweep channel
∂Csw hsw ∂Csw usw hsw j Kpen,sw
=− − hO2 + (Tpen − Tsw )
∂t ∂x 4F Hsw ch Hsw ch
(3.38)
Kpen,sw
+ (Tint,sw − Tsw )
Hsw ch
Pen structure
∂Tpen ∂ 2 Tpen j j jV
ρpen cp,pen =kpen 2
+ (hH2 − hH2 O ) + hO2 −
∂t ∂x 2F δpen 4F δpen δpen
Kpen,f u Kpen,sw
+ (Tf u − Tpen ) + (Tsw − Tpen ) (3.39)
δpen δpen
4 4 4 − T4
1 σ(Tpen − Tint,f u ) 1 σ(Tpen int,sw )
− 1 1 − 1 1
δpen ε + ε − 1 δpen ε + ε − 1
pen int pen int
The thermal fluxes along the solid parts of the cell are modelled using Fourier’s law of
heat conduction. Terms on the right hand side of the equation, in order of appearance,
are relative to:
The heat capacity of the gas phase species present in the porous media of the fuel and
sweep electrodes are supposed to be negligible compared to the heat capacity of the solid.
Furthermore, the heat radiation between the solid parts and gas streams is neglected
[41].
Interconnect/fuel electrode
4 4
∂Tint,f u ∂ 2 Tint,f u kpen,f u 1 σ(Tpen − Tint,f u )
ρint cp,int = kint + (Tfu − Tint,f u ) + (3.40)
∂t ∂x2 δint δint ε 1 + ε 1 − 1
pen int
Interconnect/sweep electrode
4 − T4
∂Tint,sw ∂ 2 Tint,sw kpen,sw 1 σ(Tpen int,sw )
ρint cp,int = kint 2
+ (Tsw −Tint,sw )+ 1 1 (3.41)
∂t ∂x δint δint ε + ε − 1
pen int
These last two energy balances include conduction through the interconnect material
itself, convection with the fuel or sweep streams and radiation with the pen structure.
To solve the energy balances equations, different boundary conditions must be ap-
plied. To determine the temperature distribution of the fuel and sweep streams, the
inlet temperature of the gases to the channels is needed. The second order term in the
energy equation for pen and interconnects implies the definition of boundary conditions
at both tip and end of the cell. In particular adiabatic conditions can be used with good
approximation [41], since the cell is thin relative to its length.
As far as the momentum balance is concerned, it is not included in the model, since
it has a weak impact on the cell performance. This is one of the outcomes of the work by
[33]: they demonstrated that the cell efficiency differs by less than 1% and the gas outlet
temperature difference is about 5 K, hence the adoption of constant physical properties
and gas velocities yields an accurate prediction of the overall cell performance.
Input data for the simulation are:
cell geometry;
temperature, pressure and composition of the fuel and sweep streams;
current density or voltage;
flow rates or utilization of the inlet streams.
Steady-state simulations are run. Any time derivative (storage term) in the mass and
energy equations obviously vanishes.
42
Chapter 3
materials employed to manufacture it are the conventional ones: the cell is constituted
by a NiO/YSZ fuel electrode, a YSZ electrolyte and a LSM/YSZ sweep electrode. Ex-
periments were carried out by putting the cell in a furnace, to keep it at a uniform and
constant temperature. To reproduce the experimental points with the model, it is nec-
essary to simulate an isothermal operation of the cell, so its temperature is fixed equal
to the temperature of the inlet streams. Three polarization curves are fitted, related
to three set of experiments, performed at different temperatures and same pressure.
The characteristics of the inlet streams for each test and the operating conditions are
indicated in table 3.2.
Table 3.2: Operating conditions and characteristics of the inlet streams of the experi-
mental tests
The value of the active area, since it is not explicitly reported, is computed from the
information of the steam utilization factor at a specified current density. The result is
1.71 cm2 . In the model, the mathematical way to describe the active area is
In this specific case, the presence of interconnects is not considered. Therefore from the
value of the active area, it is derived the length and width of the simulated cell, with the
same size proportions of the utilized electrode material, for which the width halves its
length. It is found in fact a cell with an equivalent rectangular shape able to reproduce
the behaviour of the real button cell. Input data to the model are summarized in table
3.3.
The validation is performed by manually changing the exchange current density
parameters in the model and monitoring the variation of each simulated polarization
curve with respect to the experimental points. Looking at figure 3.2, it is observed a
good agreement between numerical results and measured data; consequently the model
input parameters succeed in accurately matching the simulated cell performances with
43
SOEC model
Cell geometry
δf u [µm] 310
δsw [µm] 10
δel [µm] 10
L [m] 0.01852
W [m] 0.00926
Hf u ch [m] 0.001
Hsw ch [m] 0.001
Material electrical properties
σf u [S/m] 107 / Tpen exp(1150/Tpen )
σsw [S/m] 4.6 · 106 / Tpen exp(1100/Tpen )
σel [S/m] 3.6 · 107 / Tpen exp(-8 · 104 /RTpen )
Kinetic parameters
βa 0.5
βc 0.5
Exchange current density fitted parameters
Eact,f u [kJ/mol] 87.4
Eact,sw [kJ/mol] 88.75
γf u [A/m2 ] 3.50428 · 108
γsw [A/m2 ] 1.69850 · 108
Electrodes’ microstructural parameters
φ 0.35
τ 3
rpore [µm] 0.15
44
Chapter 3
1.6
1.4
1.2
1
Voltage [V]
0.8
Exp 750
0.6 Exp 850
Exp 950
0.4 Mod 750
Mod 850
0.2 Mod 950
0
0 0.5 1 1.5 2 2.5 3 3.5 4
2
Current density [A/cm ]
Figure 3.2: Model validation. Dashed lines are the simulated polarization curves; ex-
perimental data are marked with dots
the real one, without adding any contact resistance. The relative error on the cell voltage
values, defined as
Vexp − Vmod
e= (3.43)
Vexp
varies within a narrow range and reaches a maximum of 5% at high current density
for the 850 case. The simulations for deriving the polarization curves were obtained
by splitting the cell length in 500 sections. This discretization was chosen according to
the analysis of the grid independency. It was decided to accept a relative tolerance of
0.05% on a number of significant model variables. The following plot (figure 3.3) shows
the decreasing trend of the relative error on some significant calculated variables, as the
model is run with an increasing number of unit elements.
45
SOEC model
Figure 3.3: Results of the grid independency analysis for the cell of the model validation
46
Chapter 4
Model results
The model described in chapter 3 allows to generate results useful for the optimization
of the cell operation within a plant. For understanding the electrochemical performance
and thermal behaviour of the SOEC, parameters like temperatures, operating pressure,
flow rate and composition of the inlet streams, play a key role. Thus, the effects of
these parameters are studied here through numeric simulations at the cell level. For
all the results, the cell geometry of the model validation is used, with the addition of
interconnects on the anode and cathode sides, of thickness 0.2 mm.
47
Model results
1.6
1.5
1.4
1.3
Voltage [V]
1.2
1.1
1
Air
0.9 Steam
Oxygen
0.8
0.7
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
2
Current density [A/cm ]
Figure 4.1: Comparison of polarization curves with air, steam and oxygen as sweep gases
cases, but with different magnitude. The reason lies in the fact that the thermal power
required for the electrolysis reaction (equation 2.7) is linked to the Nernst potential, and
increases as the oxygen molar fraction of the sweep gas goes down. Hence, at a given
value of current density, the curve related to oxygen shows the maximum mean cell tem-
perature, whereas the lowest one is reached by adopting steam as sweep gas. In addition
steam is characterized by the highest heat capacity, therefore its flowing reduces more
the cell temperature at low current density but dampens the pen temperature variation
at higher current density.
A higher pen temperature leads to mitigate overpotentials. In fact the conductivity
of the electrolyte (its functional form is indicated in table 3.3) improves, which is re-
sponsible for the ohmic losses. The reaction kinetics and mass transport are enhanced
by temperature so both activation and concentration losses experience a decrease.
At a fixed current density, the ohmic overpotential simply depends on the pen tem-
perature, so it has a lower value in the oxygen case. The dominant loss for all cases is
the activation overpotential, in particular the one at the sweep electrode. Activation loss
is greatly affected by the composition of the streams, in addition to the cell temperature
48
Chapter 4
0.3 0.3
Ohm Ohm
Act fuel el Act fuel el
0.25 0.25
Act sweep el Act sweep el
Polarization losses [V]
0.15 0.15
0.1 0.1
0.05 0.05
0 0
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
2 2
Current density [A/cm ] Current density [A/cm ]
(a) Air (b) Steam
0.3
Ohm
Act fuel el
0.25
Act sweep el
Polarization losses [V]
Conc fuel el
0.2 Conc sweep el
0.15
0.1
0.05
0
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
2
Current density [A/cm ]
(c) Oxygen
Figure 4.2: Separated overpotentials versus current density for each sweep gas case: air
(a), steam (b) and oxygen (c)
49
Model results
775
Air
770 Steam
Average cell temperature [°C]
Oxygen
765
760
755
750
745
740
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
2
Current density [A/cm ]
Figure 4.3: Cell temperature variation with current density for the three sweep gas cases
influence. It is useful to report the behaviour of the exchange current density relative
to the sweep electrode as function of the oxygen molar fraction (figure 4.4). To stress
the concentration influence, the cell temperature is kept fixed. It can be stated that, for
the air and oxygen cases, an increase in the oxygen molar fraction generates an increase
in the activation loss (the exchange current density decreases). Consequently, the air
and oxygen cases present two opposing effects: considering air, it is characterized by
lower oxygen partial pressure, which is beneficial for the activation overpotential, but
also lower temperature, which is is instead detrimental for it; with oxygen, it is the other
way around. As a result, the activation loss is higher for air or oxygen depending on
which of the two effects dominates. As regards the cell functioning with steam, the fea-
ture of a very low oxygen partial pressure in the sweep channel adds to the low average
cell temperature, resulting in a higher activation loss.
The benefit of flowing oxygen is the absence of diffusion overpotential. This loss
enlarges as the oxygen dilution increases, so it is greater for the steam case.
As regards the Nernst potential, its trend is shown in figure 4.5. Since the Nernst
potential depends on the oxygen partial pressure, the maximum VN ernst occurs when
50
Chapter 4
3500
3000
Sweep exchange j [A/m 2 ]
2500
2000
1500
1000
500 750 °C
780 °C
0
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
O2 molar fraction [-]
Figure 4.4: Exchange current density at sweep electrode versus O2 molar fraction at 750
and 780
oxygen is selected as sweep gas. As the oxygen content in the sweep gas lowers, the
reversible potential decreases accordingly. The current density has no impact on the
magnitude of the reversible voltage; the small variations observed in figure 4.5 for air
and oxygen are due to a temperature effect and a concentration effect. As current density
increases, VN ernst tends to raise because the O2 concentration on the sweep side and the
H2 concentration on the fuel side are increasing in the cell. Instead at high current den-
sity, the temperature effect overcomes the increase in the oxygen concentration, thereby
reducing the Nernstian voltage. For the steam case, in the low current density portion,
a gradient of VN ernst is noted, which is the reflection of the O2 concentration gradient
in the first part of the channel: as soon as the reaction starts, oxygen is produced, thus
the composition of the sweep mixture (H2 O, O2 ) in the channel changes considerably
at the beginning of the cell length. This voltage gradient would smooth if a larger flow
rate of steam were selected.
All the losses increase with current density. In fact ohmic losses are directly pro-
portional to the current density, activation losses follow the Butler-Volmer relation,
51
Model results
1.05
1
Nernst potential [V]
0.95
0.9
0.85
0.8
Air
0.75 Steam
Oxygen
0.7
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
2
Current density [A/cm ]
Figure 4.5: Nernst potential versus current density for the three sweep gas cases
concentration losses increase as more hydrogen is produced (at constant inlet fuel flow
rate, steam utilization raises), thus mass diffusion is more hindered.
It is finally evaluated the efficiency of the system, according to the definitions pro-
vided in equations 2.10 and 2.11. As a general consideration, the efficiency has a mono-
tonic trend and decreases with current density elevation, due to the higher electric con-
sumption of the cell. It can be noticed that the steam case retains the best ηSOEC,el ,
on the grounds that it shows the lowest voltage level. If the thermal energy contribu-
tion is considered, then the steam case exhibits a slight reduced SOEC efficiency. As
regards ηSOEC , its variation among the three cases for each simulated current density
is nonetheless narrow. Comparing the air and oxygen cases, the SOEC efficiency is a
the result of a trade off between the electric share, which is greater for oxygen, and the
thermal share, whose magnitude is higher for the air case, instead.
52
Chapter 4
1 1.4
Air Air
0.95 Steam Steam
1.3
1.2
0.85
0.8 1.1
0.75
1
0.7
0.9
0.65
0.6 0.8
0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
2 2
Current density [A/cm ] Current density [A/cm ]
(a) (b)
Table 4.1: Input variables for simulation of streamwise molar fraction at fuel and sweep
channels in figure 4.7
of H2 O decreases, while the molar fraction of H2 increases to satisfy the mass balance.
Analogously, at the sweep electrode O2 is produced, so its molar fraction increases while
the one of H2 O decreases.
53
Model results
Figure 4.7: (a) Molar fraction distribution at the fuel channel and variation with steam
utilization; (b) Molar fraction distribution at the sweep channel and variation with sweep
mass flow rate
54
Chapter 4
At fixed current density and composition of the fuel mixture (figure 4.7a), the higher
the steam utilization, the lower the inlet fuel flow rate to the channel. So at the exit of
the cell, hydrogen will be less diluted with the unreacted steam.
Looking at the sweep channel, if the sweep flow rate is large (0.01 kg/h), then
the molar fraction of oxygen does not change substantially from inlet to outlet. The
reduction in the sweep flow rate brings the oxygen fraction to grow (see figure 4.7b).
The variation in the oxygen molar fraction is more evident at 0.0001 kg/h, when the
oxygen flow produced by the electrochemical reaction is of the same order of magnitude
of the inlet steam; this variation would be enhanced if a higher value of current density
was chosen, owing to the greater oxygen production.
780
775
770
T fuel 0.3 A/cm 2
Temperature [°C]
765
T sweep 0.3 A/cm 2
760 T pen 0.3 A/cm 2
T fuel 0.9 A/cm 2
755 T sweep 0.9 A/cm 2
740
735
0 0.002 0.004 0.006 0.008 0.01 0.012 0.014 0.016 0.018
Cell length [m]
Figure 4.8: Temperature profiles of pen, fuel and sweep streams at 0.3 A/cm2 and 0.9
A/cm2 - Tin streams 750 , U f 0.6, msweep 0.01 kg/h
Figure 4.8 shows the temperature profiles relative to the anodic and cathodic streams
and to the solid structure, at two different values of current density. At low current
density, the behaviour of the SOEC is endothermic because the endothermicity of the
electrolysis reaction prevails over the heating associated with polarization losses. As a
result, a decrease in the plotted temperatures is observed.
55
Model results
A higher current density contribute to boosting irreversibilities within the cell, there-
fore both streams, as well as the pen structure, are heated up by the cell heat generation.
The solid temperature results from the pen energy balance (equation 3.39); at the
open circuit voltage, it is not equal to the inlet streams’ temperature (750 in this
case) due to the heat transfer with interconnects. For the 0.3 A/cm2 case, the intercon-
nects’ temperature makes Tpen to drop to 742 , whereas at 0.9 A/cm , interconnects 2
contribute to raise the cell temperature to 764 . The streamwise variation of the in-
terconnects’ temperature is small, as depicted in figure 4.9.
780
775
Temperature interconnects [°C]
770
765
T int fuel 0.3 A/cm 2
760
T int sweep 0.3 A/cm 2
755 T int fuel 0.9 A/cm 2
T int sweep 0.9 A/cm 2
750
745
740
735
0 0.002 0.004 0.006 0.008 0.01 0.012 0.014 0.016 0.018
Cell length [m]
Figure 4.9: Temperature profiles of interconnects at 0.3 A/cm2 and 0.9 A/cm2
As soon as the fuel flow enters the channel and reaches the active sites, the reaction
takes place, inducing a temperature gradient in a very narrow portion of the cell. This
temperature gradient strictly depends on the fuel composition. Looking at figure 4.10, it
can be noted that at high current density, a decrease in the hydrogen molar fraction leads
to a decrease in the fuel temperature step at the beginning of the cell; at low current
density an increase in the fuel temperature gradient is seen. If the fuel mixture is less
rich in hydrogen, then the Nernst voltage decreases and more heat is required to sustain
the electrolysis reaction. Furthermore the heat production by irreversibilities is lower,
56
Chapter 4
752
T fuel 0.5 H 2
750
T sweep 0.5 H 2
748 T pen 0.5 H 2
T sweep 0.3 H 2
744
T pen 0.3 H 2
742
740
738
736
734
732
0 0.002 0.004 0.006 0.008 0.01 0.012 0.014 0.016 0.018
Cell length [m]
(a) 0.3 A/cm2
780
775
770
Temperature [°C]
765
T fuel 0.5 H 2
760
T sweep 0.5 H 2
T pen 0.5 H 2
755
T fuel 0.3 H 2
750 T sweep 0.3 H 2
T pen 0.3 H 2
745
0 0.002 0.004 0.006 0.008 0.01 0.012 0.014 0.016 0.018
Cell length [m]
(b) 0.9 A/cm2
Figure 4.10: Variation of temperature profiles with fuel composition at (a) 0.3 A/cm2
and (b) 0.9 A/cm2
57
Model results
since both fuel activation and fuel concentration overpotentials reduce. The behaviour
of the cell basically is more endothermic, so the cell is ’colder’. The same considerations
can be drawn when decreasing the utilization factor of steam. Temperature gradients
in a SOEC stack (along the length and across the various layers) need to be carefully
monitored to avoid any thermally induced fractures on its ceramic components, with a
consequent cell failure [12].
0.45
Uf 0.4
0.4 Uf 0.6
Uf 0.8
Current density [A/cm2 ]
0.35
0.3
0.25
0.2
0.15
0.1
0 0.002 0.004 0.006 0.008 0.01 0.012 0.014 0.016 0.018
Cell length [m]
Figure 4.11: Current density profiles and variation with steam utilization
Finally it is reported the evolution of the current density along the axial direction
(4.11). Within the first millimeter nearly, there is a steep increment of j, then it has a
decreasing trend coherently with the reduction of steam concentration at the fuel side
from inlet to outlet. The shape of the first part of the curve can be explained by the
examination of the overpotentials’ distribution, in figure 4.12(a).
At the very beginning of the cell, as soon as the reacting steam is consumed, overall
the losses get smaller and the current density appreciably increases in order to keep a
constant cell voltage. In particular activation and concentration overpotentials relative
to the sweep electrode experience a reduction due to the increase in the O2 molar fraction.
Such effect is nonetheless mitigated by the Nernst potential (figure 4.12(b)), that rapidly
58
Chapter 4
0.25
Ohm
Act fuel el
0.2
Act sweep el
Conc fuel el
Overpotential [V]
Conc sweep el
0.15
0.1
0.05
0
0 0.002 0.004 0.006 0.008 0.01 0.012 0.014 0.016 0.018
Cell length [m]
(a)
1.1
1.05
1
Nernst potential [V]
0.95
0.9
0.85
0.8
0.75
0.7
0 0.002 0.004 0.006 0.008 0.01 0.012 0.014 0.016 0.018
Cell length [m]
(b)
Figure 4.12: (a) Separated losses contribution and (b) Nernst potential in the streamwise
direction - j 0.3 A/cm2 , Tin streams 750 , U f 0.6, msweep 0.01 kg/h
59
Model results
grows at the very beginning of the cell length. In the plot, the three different curves
are parameterized with respect to the steam utilization factor. It can be remarked that
the higher Uf , the higher the diffusion losses at the fuel side, since the concentration of
steam in the channel is lower and the reaction is limited by mass diffusion. Furthermore
the activation overpotential at the fuel electrode is greater as a result of the hydrogen
concentration profile, while at the sweep electrode the activation loss distribution along
the cell is caused by the combined effect of temperature variation and oxygen molar
fraction. Consequently in the first half of the cell, the coupling between the lower
sweep activation and the higher Tpen enables current density to increase; VN ernst helps
in maintaining a constant voltage due to its higher value at high Uf . In the second part
of the cell, activation at sweep electrode and concentration loss at the fuel electrode are
prevailing and, along with the Nernst voltage, act such that current density decreases
more at high Uf ; the drop in the current density is lessened by the decreasing trend of the
pen temperature. The total hydrogen output is the same for all cases, since the average
current density is kept constant; the difference is that a larger amount of hydrogen is
generated at the inlet of the cell for the high Uf case. It is reasonable not to operate
the cell at very high steam utilization factor because it is essential to avoid a significant
increase in the cathode concentration overpotential caused by steam starvation near the
cell outlet.
The cell is simulated in isothermal condition at 750 . This is done in order to stress
and better understand the effect of concentration produced by the sweep flow rate and
thus to separate the temperature effect. The sensitivity is performed considering the
following input conditions (table 4.2).
Variable Range
j [A/cm2 ] [0 1.5]
msweep [kg/h] [5e-7 0.1]
Fixed parameter Value
Fuel composition [mol] 0.1 H2 0.9 H2 O
Uf 0.6
Sweep type steam
60
Chapter 4
As a result, maps related to QSOEC , Qreact and Qloss , as previously defined in equa-
tions 2.9, 2.7 and 2.8, are obtained.
0.1 0.1
2
0.05
0
0.11
0.08
0.02
0.06
0
0.1
0.02
0.11
0.01 0.13
0.005
5
-0.2
-0.2
-0.3
-0.1
-0.4
0 0.02
-0.5
-0.6
-0.0
0.06
0.1
Qsoec [W/cm 2 ]
1e-3
-0.4
1e-4
0.08
-0.6
0.02
0.06
1e-5
0
-0.8
-0.8
-0.05
0.06
0.02
-0.6
-0.5
-0.1
-0.4
-0.3
-0.2
-1
-1
5e-7
0 0.3 0.6 0.9 1.2 1.5
2
Current density [A/cm ]
61
Model results
0.6 0.1
0.05
0.6
0.5
0.02
0.06
0.1
0.15
0.2
0.25
0.3
0.35
0.4
0.4
0.02
0.5
5
5
0.5 0.01
Sweep flow rate [kg/h] 0.005
0.4
Qreact [W/cm2 ]
1e-3
0.3
1e-4
0.45
0.4
0.35
0.3
0.2
0.25
0.2
0.15
0.1
0.06
0.02
1e-5
0.1
0 5e-7
0 0.3 0.6 0.9 1.2 1.5
2
Current density [A/cm ]
(a)
0.1
0.05
1
1
0.02
0.9
0.8
0.15
0.7
0.06
0.6
0.3
0.4
0.5
0.2
0.1
0.02
0.01
0.005
Sweep flow rate [kg/h]
0.8
Qloss [W/cm2 ]
1e-3
0.6
1e-4
0.4
1
0.9
0.8
0.7
0.6
0.5
0.15
0.4
0.3
0.06
0.2
0.1
0.02
1e-5
0.2
0 5e-7
0 0.3 0.6 0.9 1.2 1.5
2
Current density [A/cm ]
(b)
62
Chapter 4
Concerning Qreact (figure 4.14(a)), it can be noted that, at fixed current density it
increases with the sweep flow rate. This trend reflects the substantial influence of the
Nernst potential on the oxygen concentration. With the addition of more and more
steam at the sweep channel, the produced oxygen gets more diluted, its molar fraction
reduces and, therefore, VN ernst exhibits a decrement. This effect is stronger for the
larger simulated sweep flow rates, that is why Qreact more rapidly increases with j in
the upper region of the plot. The concentration effect is mitigated as current density
is raised, because of the greater amount of product O2 from the reaction. This means
that, for a certain sweep flow rate, VN ernst slightly increases as current density increases.
Despite this, Qreact shows an almost linearly increase with current density, at fixed sweep
flow rate. Comparing the two maps in figure 4.14, it can be observed that Qloss reaches
greater values, meaning that it increases faster than Qreact : this is due to the quadratic
dependence on current density, as the cell voltage depends on current density through
polarization losses.
Dwelling now on Qloss (figure 4.14(b)), fixing the sweep, it obviously increases as
j goes up because losses are enhanced; fixing j, it slightly decreases and then starts
increasing. In fact, as more steam is utilized as sweep, concentration overpotential
at the sweep electrode increases, always due to the oxygen dilution effect. On the
other hand, following the trend of the exchange current density previously discussed,
activation overpotential reduces until 0.01 kg/h value, then starts increasing. So when
both overpotentials vary in the same direction, the growth of Qloss in the upper region
of the map is observed.
Therefore, by combining the trends of Qreact and Qloss , the map of QSOEC can be
more evidently justified. Starting from the no sweep case, the isolines go in the direction
of higher current densities following the thermal inertia provided by a greater mass flow
of steam; the inversion detected at high sweep is due to a more rapid rise of Qloss coupled
to a more rapid rise of Qreact as well, which act such that the balance between the two
thermal fluxes occur at lower value of current density. If the sweep flow were increased
even more, then the concentration effect would not emerge anymore and only the inertia
effect would act, bringing the QSOEC isoline again to higher current density values.
To figure out the temperature impact on the magnitude of the net thermal flux, two
maps of QSOEC are likewise built at 700 and 800 , with the other conditions as the
ones listed in table 4.2. It can be clearly seen in figure 4.15 that a higher cell temperature
enhances the endothermic behaviour of the cell, coherently with the thermodynamic
considerations (section 2.1.1). The cell is better performing from an electrochemical
63
Model results
0.1
0
0.06
0.04
0.02
0.02
0
-0.2
08
0.01
0.
0.06
Sweep flow rate [kg/h]
-0.05
-0.4
0.02
-0.1
-0.2
-0.3
-0.4
-0.5
-0.6
-0.8
-1
-1.2
Qsoec [W/cm ]
2
1e-3
-0.6
0
-0.8
1e-4 0.0
4
-1
0.02
-1.2 1e-5
0
-1.4
-0.05
0.02
-0.8
-0.6
-0.5
-0.4
-0.3
-0.2
-0.1
-1.4
-1
-1.2
-1.6 5e-7
0 0.3 0.6 0.9 1.2 1.5
2
Current density [A/cm ]
(a)
0.1
0.05
0
0.12
0.1
0.1
0.02
0.06
0.1
0.1
0.16
0.12
4
0.18
0.01
0 0.005
Sweep flow rate [kg/h]
-0.3
-0.2
0 0.02
-0.1
-0.0
0.06
0.1
6
Qsoec [W/cm 2 ]
0.1
-0.1 1e-3
-0.2
1e-4 0.14
0.12
-0.3
0.1
0.02
0.06
0.1
1e-5
0
-0.4
-0.4
-0.05
0.06
0.02
-0.1
-0.2
-0.3
0.1
-0.5
-0.5
5e-7
0 0.3 0.6 0.9 1.2 1.5
2
Current density [A/cm ]
(b)
64
Chapter 4
It is interesting to study the behaviour of the cell under pressure. It is expected that
a rise in the operating pressure allows for a reduction in concentration and activation
overpotentials. At the same time, the Nernst law predicts a reversible voltage elevation.
The simulations done confirm what has been stated above. To make a comparison with
the results of section in terms of polarization curves and losses and temperature profile,
input data of the 750 validation case are fixed and only the operating pressure is
varied. Polarization curves in figure 4.16 are obtained at increasing values of operating
1.5
1.4
1.3
1.2
Voltage [V]
1.1
0.9 1 bar
3 bar
0.8 5 bar
10 bar
0.7
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
Current density [A/cm 2 ]
pressure; the details of Nernst voltage and overpotentials are presented in figures 4.17
65
Model results
and 4.18.
0.3 0.03
Fuel, 1 bar
Fuel, 3 bar
Sweep, 3 bar
0.2 0.02
0.15 0.015
Fuel, 1 bar
Sweep, 1 bar
0.1 0.01 Fuel, 3 bar
Sweep, 3 bar
0.05 0.005
0 0
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
Current density [A/cm 2 ] Current density [A/cm 2 ]
(a) (b)
0.04
0.035
Ohmic overpotential [V]
0.03
0.025
0.02
0.015
0.01
0.005 1 bar
3 bar
0
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
Current density [A/cm 2 ]
(c)
Figure 4.17: Comparison of activation (a), concentration (b) and ohmic (c) overpoten-
tials at 1 bar and 3 bar
It can be noticed that, at low current density, the higher Nernst voltage is dominant
over the loss abatement: this is a drawback for the electrolysis process because it leads to
increase the energy requirement for water splitting. The sensitivity of the Nernst voltage
diminishes with pressure elevation, owing to the logarithmic nature of the pressure term
66
Chapter 4
Figure 4.18: VN ernst variation with pressure. Values refer to a current density of 0.5
A/cm2
67
Model results
765
1 bar
3 bar
Average cell temperature [°C]
760 5 bar
10 bar
755
750
745
740
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
2
Current density [A/cm ]
Figure 4.19: Average cell temperature variation with current density, at different values
of pressure
Observation on the model Like in a similarity approach, the reported trends and
behaviors of the cell are representative of all sizes, given the same input conditions and
by scaling proportionally the flow rates. However, small differences (of some degrees)
in the pen temperature profile are observed when increasing the scale of the system.
This could be attributed to an approximation inherently employed in the heat exchange
submodel of the SOEC, for which it is neglected the thermal exchange between adjacent
elements. This kind of simplification could give rise to differences in the cell behavior,
sensitive to the scale of the physical problem to cope with. In fact the reactant and
sweep flow rates as well as the electric energy at stake are proportional to the cell active
area, whereas the thermal fluxes are linked to the resistances of pen and interconnects’
transversal areas, that increase only with the channel width.
68
Chapter 5
Plant model
The plant is conceived to supply hydrogen for a small refueling station in California.
Therefore a hydrogen throughput of 150 kg/d is set as target. The SOEC stack, within
the plant, is fed with steam as sweep gas; operation without the presence of a sweep
gas is considered as well. Two plant layouts are then proposed, depending on whether a
sweep gas is employed or not. The implementation of the plant model is done on Aspen
Plus. Figures 5.1 and 5.2 represent the two plant schematics.
5.1 Description
The streams entering the electrode compartments of the SOEC stack are preheated
to the design temperature (750 ) in a single stream. The feed water passes first
through an economizer and an evaporator (EE), so as it is partly evaporated: in this
fashion, the enthalpy content of the hot streams exiting the SOEC is exploited. Then
the stream is directed to the solar plant, where the evaporation is completed and steam
is further superheated to the design temperature. By way of a splitter, the amount of
steam constituting the sweep gas flow is set; the remainder of steam, with the addition
of the hydrogen reflux, creates the fuel flow to the fuel channel. In the SOEC, the
electrolysis reaction takes place and the output flows are a mixture of H2 and H2 O
at the fuel electrode, and steam enriched with O2 at the sweep electrode side. Both
streams represent the hot streams in the heat exchange section. At the exit of the
economizer, they are further cooled in two different condenser. The hydrogen reflux
(from the fuel stream) is partitioned before the condenser, it is re-compressed through a
H2 recirculation blower and re-heated in a separated heat exchanger (HR) before sending
it upstream the SOEC unit. Inlet hydrogen is in fact required in order to maintain
69
Plant model
reducing conditions on the nickel cermet electrode. Two adiabatic flashes allow for the
condensate separation, in order to get the hydrogen and oxygen flows. On the one
hand, the valuable product streams are sent each one to an intercooled compression
unit, where they are further purified by condensation of the residual water and attain
the storage pressure of 200 bar. On the other hand, the condensate is recirculated, so
that it is decreased the supply of make-up water. A feed water pump, electrically driven,
is needed to recover the plant pressure drops.
70
Chapter 5
The absence of a sweep gas lets simplify the plant arrangement, with great potential
for cost reduction. The oxygen separator is not necessary anymore since a pure oxygen
stream exits the sweep channel of the SOEC unit. Moreover the two condensers are
removed as a result of the lower temperature levels achieved at the outlet of the econo-
mizing section. The reflux of hydrogen is split after the hydrogen separator, ensuring a
single phase stream approaching the H2 blower.
It is assumed that each cell has a geometric area of 550 cm2 , which is the area achievable
by Fuel Cell Energy, SOFC manufacturer. As already noted (equation 3.42), in the
model, the mathematical way to describe the active area is
Since our model is one dimensional, it is assumed that the reaction occurs at the interface
between the electrodes and the electrolyte. Actually the active area is different from the
geometric area. The active area is defined as the sum of all the active sites where the
electrochemical reaction takes place. It has an irregular shape and it lies on a 3D surface;
so the reaction takes place in the volume of the electrode, at the triple phase boundary.
To more accurately take into account the active area, it is decided to use the value of
the specific area provided by [51]. The specific area [m−1 ] is defined as
m2 active area
Aspecif ic = (5.2)
m3 overall electrode catalyst
So by multiplying by the electrode volume, it is found a value of 594 cm2 of active area.
Assuming a square cell, it is calculated the equivalent length and width, in order to be
consistent with the mathematical formulation of the model. However the weakness of
this method is that we have a perturbation in the temperature field of the cell, as we are
increasing the cell length. Given that the difference in length is small, the procedure is
nonetheless accepted. The resulting cell has the same features of the cell in [51] (same
thicknesses, same electrical properties), but with a different active area. Similarly to
what has been done for the cell employed in the model validation, a grid independency
analysis has been performed for the present cell (figure 5.3). At least 800 unit elements
need to be considered in order to satisfy the constraint on the relative error (equal or
lower than 0.05%). Nevertheless on grounds of computation time, for running plant
simulations, a SOEC submodel with 50 nodes is used, inevitably raising the error to
71
Plant model
Figure 5.3: Results of the grid independency analysis for the cell employed in the plant
about 1% (with respect to the 800 nodes case). For each operating condition, the
needed number of cell is obtained by iterative process, with the objective of producing
the target hydrogen output.
To simulate the thermal dishes, a heater block is used, so the value of net duty that
the Aspen simulation returns is the overall thermal power that the solar plant has to
provide.
The following percentage pressure drops (with respect to the inlet pressure of the
block), gathered in table 5.1, are considered.
72
Chapter 5
73
Plant model
pressor, entailing lower electricity consumption. Finally the process can take advantage
of saving oxygen as a valuable commodity: to do this, the separation of oxygen from
steam is obtained simply by condensation of the steam itself. If air was employed instead,
a pressure swing adsorption (PSA) unit would be necessary as air separation technology
for oxygen purification, which is more energy intensive.
Unlike fuel cells, a sweep gas is not necessarily required since oxygen is being produced
in electrolytic operation. Therefore it is interesting to study a no sweep gas scenario,
where pure oxygen evolves from the sweep electrode side of the SOEC unit. Benefits
both from a system efficiency and economic viewpoints are expected. More technical
and safety issues have to be managed when processing pure oxygen; by the way research
[59] indicates that pure oxygen can be safely handled at high temperature, if proper
materials are used.
10-3
4 3
2 0.0035
1 0.0
0.00025
0.00 03 15
0
0.001
0
3 0.5 0.0
02 5
2 0.0
5
02
01
00
0.0
0.
015
02
Sweep flow rate [kg/h]
05
0.1
1
2
00
-0.0
.00
-0.0
-0.
-0
Qsoec [W/cm ]
2
1
0.01
0
25
0 00
0.0
1e-3
-1
-0.0025
-0.0005
-0.0015
-0.001
-0.002
1e-4
0
-2
-3 1e-5
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
2
Current density [A/cm ]
Figure 5.4: Map of QSOEC for the cell employed in the plant; Uf was fixed to 0.6
As already performed in the Model results (section 4.3), it is reported in figure 5.4 the
plot of the net thermal flux versus current density, with the conditions of the anode and
cathode streams of the plant simulation. The resulting QSOEC should be interpreted as
the net thermal energy that not only supports the electrolysis reaction (and contributes
to the H2 production) but is redistributed among all the layers of the cell and hence
contributes to the temperature change (between inlet and outlet) of pen, interconnects
74
Chapter 5
and the enthalpy variation of the fuel and sweep flows. This map turns out to be useful
in support of the choice of a representative sweep gas flow rate for the plant process.
As first guess, it is considered a SOEC operation at the thermoneutral voltage,
whose value is about 1.285 V . Being aware that a higher current density is desirable
for boosting the hydrogen production rate, a sweep gas flow rate of 0.5 kg/h per cell is
selected, featuring a current density of about 0.7 A/cm2 at null QSOEC .
Concerning the no sweep gas case scenario, it is simulated by executing the SOEC
submodel with a very little steam flow rate per cell (10−5 kg/h, at the extreme bottom
of the map).
The temperature gradient across the cell length should be lower than 100 , not
to run into damages due to thermal stresses [49]. An upper limit value in the operat-
ing current density is thus set by this constraint, beside durability considerations (cell
degradation increases at high j, resulting in a shorter lifetime).
75
Plant model
Figure 5.5: DNI map relative to USA. The chosen installation site is pointed out with
a grey marker [5]
System Advisor Model [57], the most recent hourly data of Direct Normal Irradiance
[W/m2 ] and ambient temperature [ ] of the installation site are collected and visualized
in figures 5.6 and 5.7. These are fundamental information for the definition of the solar
field performance.
The input data necessary for the dimensioning are the power plant requirements, in
terms of total electric power Pel,tot and overall thermal power Qth,tot :
P el,tot includes the electric consumption by the SOEC stack, both the intercooled
compression units, the feed water pump, the H2 recirculation blower. Also the
electric power required by auxiliaries (recirculation pumps for the condensers and
intercoolers) is taken into account and it is assumed as 0.8% of the exchanged
thermal power.
Q th,tot is the heater duty, which is equal to the mass flow rate of steam times its
enthalpy variation.
1. Solar only: the dimensioning is aimed at providing the annual energy requirement
76
Chapter 5
1000
900
800
20 700
600
500
400
15
300
Hour of day
200
100
10 0
0
Jan Feb Mar Apr May Jun Jul Aug Sep Oct Nov Dec
Figure 5.6: Heat map of the Direct Normal Irradiance of Lancaster [57]
30
25
20
Dry bulb temp (C)
15
10
0
Jan Feb Mar Apr May Jun Jul Aug Sep Oct Nov Dec
77
Plant model
of the plant, only by exploiting the solar energy source. Calculations are performed
on a yearly energy basis; it is supposed that Pel,tot and Qth,tot are mean powers.
2. Hybrid : the dimensioning is aimed at assuring the plant power needs in a specific
design condition; to fulfill the overall plant duty in the other conditions, it is
supposed an integration with a natural gas source. Calculations are performed on
a power basis.
This method starts from the information of the yearly DNI of the selected site, which
is available on the solar map. The value for Lancaster is 2989 kW h/m2 /y. Some
assumptions are then made, concerning the efficiency and the size of the dish collectors.
The selected values are listed in table 5.2. References [25] and [46] are used.
78
Chapter 5
The solar field shade factor accounts for the reciprocal shading among the mirrors,
generated by the sun position variation during each time. It is defined as
T otal shaded area
SF = (5.3)
T otal aperture area
The SF depends on many factors, such as the available land area, the total number of
dishes and the layout of the solar field (dishes could be arranged in a more or less packed
way). The value reported in table 5.2 is a yearly average and will be checked only a
posteriori. Rigorously, given a certain geometry of the solar field layout, an iterative
procedure on SF would be necessary to find out the dimensioning that exactly matches
the utilized value of Shade Factor.
Moreover, it is supposed not to produce below a certain value of solar irradiance, so as to
avoid a heavy efficiency penalty; a lower DNI limit is therefore fixed to 200 W/m2 . Given
the DNI distribution of Lancaster during the year, it is found that a DNI curtailment
by 5% has to be introduced.
An effective Annual Direct Normal Irradiation ADN Ief f [kW h/m2 /y] is obtained
by taking into account the abovementioned shade factor and DNI curtailment, as in
equation 5.4.
ADN Ief f = ΣDN I · (1 − SF ) · (1 − 5%) (5.4)
The useful specific electric energy [kW h/m2 /y] is computed as
The ratio between the two returns the required total aperture area. Knowing the aper-
ture of one electric dish (113.1 m2 ), it is possible to derive the total number of electric
dishes. The procedure is analogous for the thermal dish. However when the total aper-
ture area is found, only one thermal dish would be built with that aperture, unless the
upper bound (400 m2 ) is hit. In that case, a number of thermal dishes with the same
aperture area is considered, leading to equally divide the thermal energy demand of the
plant among them.
It should be noted that the procedure does not give information on the dispatchability
of solar power. Therefore it can be accounted as a speculative discussion, that tries to
give an idea of the size of the solar field necessary for producing a fully renewable
hydrogen. To deepen the feasibility of the solar only strategy, the dynamic behaviour of
the plant should be explored, together with the inevitably intermittent characteristic of
the solar source.
79
Plant model
The dimensioning starts from the selection of the design conditions. It is supposed that,
at the peak sun irradiance, the solar field is up to deliver the overall plant power re-
quirement. The values assumed for the nominal condition are indicated in table 5.3.
Following what have been done in the solar only approach, the same electric dish diam-
Table 5.3: Assumptions for hybrid approach. Values refer to the nominal condition
eter is chosen and the same upper bound on the thermal dish aperture is set. At design
conditions, no shading is supposed to occur.
Concerning the electric dish part, it is computed the power produced by one dish as
The number of needed dishes is simply obtained by dividing the total electric power
plant requirement (Pel,tot ) by the above term.
For the thermal dish part, first it is calculated the maximum power that can be deliv-
ered by a dish of 400 m2 . If the resulting value is higher than the thermal requirement
by the plant, this means that the aperture is oversized and only one thermal dish is
necessary. Therefore, the thermal dish is dimensioned exactly to provide Qth,tot and the
needed aperture is derived, consequently. In case the resulting value is not enough to
make a single thermal dish, then the thermal power plant demand is split equally among
the dishes, thus characterized by same aperture area.
The strategy involves a hybridization: in fact, the delta of power that for every
hour the solar field cannot provide due to lower DN I or absence of solar radiation,
is indeed supplied by natural gas. In particular a hybrid solar MGT is considered, in
which a burner is placed after the solar absorber and is activated in lack of enough solar
energy to bring the process fluid to the design TIT; in this way, the MGT never operates
at partial load. Therefore, the removal of penalties arising from part-load operation
80
Chapter 5
implies an enhanced energetic performance with respect to the solar only case. The
cycle efficiency will be affected only by ambient temperature variations.
For the thermal demand, a natural gas fired boiler provides thermal energy in support
of the thermal receiver. A boiler efficiency equal to 0.95 is assumed.
Within both approaches, a best case scenario is evaluated (table 5.4), in which an
improvement in the solar efficiencies is considered. For the hybrid approach, only the
nominal solar to electric efficiency is enhanced, while the design value for the solar to
thermal efficiency is kept constant. Values related to solar to electric efficiencies are from
reference [26], which investigates the behaviour of a solar MGT endowed with a ceramic
expander, allowing a greater maximum temperature (TIT 1100 ) to be achieved, with
consequent improvement in the cycle efficiency. Hence it is evident the positive impact
of this solution from a thermodynamic viewpoint.
Table 5.4: Best case efficiencies for solar only and hybrid approach
Natural gas calculation To determine the consumed natural gas, the following pro-
cedure is set up, starting from the hourly data of Direct Normal Irradiance and ambient
temperature of the chosen location. A detailed discussion is made on the electric part,
exploiting information on the performance of a parabolic dish present in [25] for the
base case and [26] for the best case scenario; an analogous procedure is employed for the
thermal part (with reference [46]), in order to compute the yearly amount of natural gas
to be fed to the boiler. Calculations are performed on an hourly basis.
Thanks to the natural gas addition, the MGT mounted on the electric dish always
works at full load. So the thermodynamic cycle ’feels’ the same solar irradiance, equal
to DN Idesign , but is modified by ambient temperature. The cycle efficiency ηcycle is
directly computed from the solar to electric efficiency, given that
81
Plant model
The performance maps, present in [25] for the base case and [26] for the best case,
are useful to determine by interpolation the corresponding solar to electric efficiency
variation with ambient temperature. Consequently, it is possible to compute on the one
hand the total heat input to the cycle Qsolar+ng,cycle [kW h], required by the hydrogen
production plant in order to generate the target hydrogen yield.
A fraction of this thermal energy is provided by the solar source, depending on the
available sun irradiance. The natural gas source is in charge of supplying the remaining
fraction. The input thermal energy from the solar source Qsolar,cycle [kW h] is expressed
as
Qsolar,cycle = DN Ief f · Atot,el · ηopt · ηth (5.10)
qloss
ηth = 1 − (5.11)
DN I · ηopt
The optical efficiency is assumed to be constant; also the specific heat loss qloss is taken
as a constant as first approximation, even though it varies with the receiver temperature.
The effective sun irradiance DN Ief f , indicated in equation 5.10, accounts for the
DN I reduction due to shading among solar concentrators, therefore a lower amount of
energy can be extracted from the solar source. It is simply calculated as
DN Ief f = DN I · (1 − SF ) (5.12)
The yearly fuel energy [kW h/y] is obtained by the difference of the aforementioned
thermal energies, whose yearly values are found by summation for each i hour of the
year.
Qf uel = ΣQsolar+ng,cycle,i − ΣQsolar,cycle,i (5.13)
Eventually, knowing the Lower Heating Value of natural gas, the yearly amount of
natural gas consumed in the burner of the MGT is derived.
In table 5.5, the natural gas specifications are reported.
Shading calculation The estimation of the Shade Factor is done by using a tailored
Matlab ® routine. The essential steps to run the computation are the definition of the
sun position and the solar field layout.
82
Chapter 5
Table 5.5: Molar composition, molar mass and Lower Heating Value of natural gas
Figure 5.8: Representation of Zenith and Azimuth angles [6]. The convention for Az-
°
imuth definition is 0 North, 90 East°
The sun position is uniquely defined by Zenith and Azimuth angles (figure 5.8) and
it constitutes a vital information to represent the double-axes tracking movement of the
dish.
The layout of the solar field is determined by fixing the geometry of the single dish, in
terms of aperture shape (circular, in this case) and size, and the mutual position between
dishes. Specifically, regularly-spaced layouts, investigated in [20], can be described by
four independent parameters:
Aspect Ratio (AR), ratio of spacing between collector columns with respect to
the spacing between collector rows;
83
Plant model
Offset (or Stagger, S), relative offset of adjacent collector columns as a fraction of
the spacing between layout rows;
Rotation (θ ), anticlockwise rotation of the array;
r
Ground Cover Ratio (GCR), ratio of the collector aperture area to the reserved
land area.
84
Chapter 5
85
Plant model
86
Chapter 6
Plant results
This chapter presents the results of the plant simulations. Simulations are divided into
two main sub-groups, according to the choice of the operating pressure. Within each
subset, two cases are explored, depending on whether steam is used or not as sweep
gas. A series of performance maps are obtained and discussed. Best efficiency points are
selected for each case and afterwards the economic feasibility of these points is assessed.
The electrochemical reaction carried out by the SOEC implies the generation of about
1190 kg/d of oxygen in addition to 150 kg/d of hydrogen. This is another valuable
product, with many industrial and commercial applications. The multi-product feature
of the plant enhances its value and has to be accounted for when evaluating its economic
impact.
The feed water pump is in charge of regulating the pressure at which every unit
works, in order to overcome all the plant pressure drops and avoid sub-atmospheric
operation. The resulting discharge pressure of the pump is 1.192 bar; the electrolyser
will operate at a pressure of 1.11 bar, slightly higher than the ambient one.
A set of performance maps is determined, by simulating the process with different
conditions of the SOEC. Specifically two key parameters are varied in reasonable ranges:
current density j [A/cm ], in the range [0.2 0.8] to capture all the possible
2
87
Plant results
would imply a greater mass flow of steam evolving within the plant, leading to
higher duty requirement and a consequent efficiency penalty.
The number of cells involved for each value of current density is showed in the bar
plot below (figure 6.1).
Figure 6.1: Number of cells obtained for each current density. Values refer to the 60%
Uf simulation case
88
Chapter 6
20 0.9
18 0.8
14
0.6
12
0.5
10
0.4
8
0.3
6
4 0.2
2 0.1
0 0
40 45 50 55 60 65 70 75 80
Steam utilization [%]
(a)
160 0.15
120
100
0.11
80
0.09
60
40
0.07
20
0 0.05
40 45 50 55 60 65 70 75 80
Steam utilization [%]
(b)
Figure 6.2: Mass flow rate and hydrogen content versus steam utilization for (a) H2
recirculation stream and (b) inlet fuel stream to SOEC. Values refer to the 0.8 A/cm2
simulation case
89
Plant results
content in that specified stream. As a result, a lower flow rate of hydrogen needs to
be recirculated, that is why a ’plus one’ cell is necessary at high Uf . Moreover, since
the molar fraction of H2 in the fuel mixture exiting the SOEC becomes greater at high
Uf , the overall mass flow of reflux experiences a reduction. To support the explanation,
in figure 6.2 (a), the trends of the reflux flow rate and its hydrogen molar fraction are
reported as function of the steam utilization, at 0.8 A/cm2 ; figure 6.2 (b) shows the
mass flow rate of the fuel stream at the inlet of the cell and its hydrogen molar fraction
at the same j.
Firstly, a cell voltage map is provided (figure 6.3), to get aware of how the cell
performs from an electrochemical viewpoint.
1.32 80
1.32
1.315
1.31
1.305
1.29
1.3
1.29
1.28
1.31
5
5
1.28
70
Steam utilization [%]
1.3
Voltage [V]
1.29 60
1.27
5
1.28
1.28
1.29
1.295
1.3
1.305
1.31
1.315
1.32
1.27
50
1.28
1.27
1.26 40
0.2 0.3 0.4 0.5 0.6 0.7 0.8
2
Current density [A/cm ]
The isoline at near 1.285 V describes the thermoneutral condition of the cell, in which
the heat generated by internal losses balances the heat sink of the endothermic steam
reduction.
The SOEC unit works with a variable inlet temperature at the fuel electrode. In
fact, while the temperature of the superheated steam is fixed to 750 by the solar pre-
heating, the addition of the recirculation stream perturbs its temperature, according to
two main factors:
the temperature and flow rate of the recirculation stream: as already noted, it has
90
Chapter 6
750 80
748
746
744
742
740
750
745
70
740
60
736
750
744
742
740
746
748
735
734
50
73
738
2
730
40
0.2 0.3 0.4 0.5 0.6 0.7 0.8
2
Current density [A/cm ]
820 80
800
820
680
700
720
740
760
780
800
780 70
Steam utilization [%]
760
Tfuel,out [°C]
740 60
800
720
780
760
740
720
700
680
700 50
680
660 40
0.2 0.3 0.4 0.5 0.6 0.7 0.8
2
Current density [A/cm ]
91
Plant results
the thermal behaviour of the cell: depending on the operating current density, the
net heat generation of the cell can be positive, zero or negative and so the cell
could exhibit an endothermic, isothermal and exothermic functioning respectively.
The inlet and outlet temperature maps of the fuel stream are shown in figures 6.4 and
6.5. Going to higher j, the maximum outlet temperatures are detected. This has a direct
consequence on the reflux, whose temperature gradually increases (keeping an approach
temperature difference of 30 ) till reaching the superheated steam temperature exiting
from the thermal dish. There is no need to bring the recirculation flow to a temperature
greater than 750 , so when the fuel flow downstream of SOEC is characterized by a
temperature equal or higher than 780 , the recirculation outlet temperature from HR
is set to be a constant value (750 ). This explains the inlet fuel temperature plateau at
high j. A greater Uf brings greater temperatures both at SOEC inlet (due to the smaller
and hotter reflux stream) and at SOEC outlet, because the exothermic behaviour of the
SOEC is enhanced. The isothermal condition takes place in between 0.4 and 0.5 A/cm2 ,
depending on the steam utilization factor.
80 0.69
0.6
0.68 8
0.68
70 0.67
Steam utilization [%]
0.66
I law efficiency [-]
0.67
0.66
60 0.66
0.65
0.64
0.65
0.64
0.64
0.63
0.62 50
0.63
0.62
0.62
0.61
0.6 0.6 0.61
40
0.2 0.3 0.4 0.5 0.6 0.7 0.8
2
Current density [A/cm ]
92
Chapter 6
Figure 6.6 represents the I law efficiency of the plant, which is given by
mH2 ,target LHVH2
ηI = (6.1)
Pel,tot + Qth,tot
This definition evaluates the hydrogen throughput with respect to the total power (elec-
tric plus thermal) required by the process; it does not take into account the solar energy
conversion executed by the dish system. The value of LHVH2 is 119.954 M J/kg.
It can be noted that it has a monotonic trend and increases with both current density
and steam utilization. The result can be explained by looking at the maps related to
the total electric demand and thermal demand from the plant (figure 6.8).
Figure 6.7: Electric power contributions. Values refer to the 0.8 A/cm2 , 80% Uf simu-
lation case
The former scales with current density because the SOEC absorbs more and more
electricity to increase its hydrogen production rate; as displayed in figure 6.7, it is the
SOEC unit that possesses the largest share of electric power consumption (83.6%) fol-
lowed by the intercooled compressor of hydrogen (11.2%) and oxygen (5.1%).
As for the steam utilization effect, it can be seen that the total electric power decreases
with Uf elevation, because the negative variation of the electric consumption of the
equipment (except for SOEC), mainly the intercooled compressor of H2 , overtakes the
increasing power absorbed by SOEC.
Viceversa, the latter is minimized at the highest current density. In fact at high j,
due to the exothermic behaviour of the SOEC, the outlet streams from the sweep and
fuel channels are hotter and more heat can be transferred to the feed water in the heat
93
Plant results
272 80
260
0.5 26
1
2
26
26
3
26
270
70
Steam utilization [%]
268 26
4
Pel,tot [kW]
2
26
266 3 5
60 26 26
6
26
4
264 26 7
26
50 265
8
262 26
266 9
26
267 27
0
260 40
0.2 0.3 0.4 0.5 0.6 0.7 0.8
2
Current density [A/cm ]
(a)
80
80 44 40
75
70 48
Steam utilization [%]
70 44
Qth,tot [kW]
65 52
48
60 60 56
52
55 60
56
50 64
50 60
68
45 64
72
68
40
80 76 72
40
0.2 0.3 0.4 0.5 0.6 0.7 0.8
2
Current density [A/cm ]
(b)
Figure 6.8: Power plant requirements for the no sweep gas case: (a) total electric de-
mand, (b) total thermal demand
94
Chapter 6
exchange section, so that the stream going to the solar section is characterized by a
higher vapor fraction. As expected, the lowest power level (either electric or thermal)
occurs at the highest Uf because of the lower amount of working fluid that the plant is
processing. For the 80% steam utilization case, the variation with current density of the
feed vapor fraction at the outlet of EE is depicted in figure 6.9.
0.8
Feed vapor fraction before dish [-]
0.7
0.6
0.5
0.4
0.2 0.3 0.4 0.5 0.6 0.7 0.8
Current density [A/cm 2 ]
Figure 6.9: Feed vapor fraction variation at the outlet of EE with current density. Values
refer to the 80% Uf simulation case.
It is interesting to derive the efficiency map relative to the SOEC functioning (figure
6.10), given the definition in equation 2.10. Compared to the I law efficiency, it exhibits
indeed an opposite trend with current density, meaning that the highest SOEC efficiency
point is at the lowest value of current density. In spite of a worse electrochemical
performance of the SOEC, it is convenient from a thermodynamic perspective to operate
at higher current density within the plant. As a result the best I law efficiency point is
found at 80% Uf , 0.8 A/cm2 , with a value of 69.6%.
As far as the solar dish segment is concerned, the variability in the total number of
required concentrators is narrow to changing SOEC functioning. It can be deduced that
the component that has more impact in the economic evaluation of the proposed plant
is the SOEC stack.
95
Plant results
0.815 80
0.81
0.805 70
0.815
0.805
0.81
5
Steam utilization [%]
0.79
0.8
85
SOEC efficiency [-]
0.7
0.8
0.7
0.795
60
0.79
0.785
50
0.78
0.805
0.795
8
5
0.81
0.79
0.7
0.78
0.8
75
0.775
0.7
40
0.2 0.3 0.4 0.5 0.6 0.7 0.8
2
Current density [A/cm ]
Solar only A total of 43 solar dishes is computed, one of which is thermal. For this,
the aperture area varies proportionally to the total thermal power required, as displayed
in figure 6.11.
The solar to hydrogen efficiency, whose related map is provided in figure 6.12, is
expressed as
mH2 ,target LHVH2
ηsun−to−H2 ,year = (6.2)
Pel,tot Qth,tot
+
ηtot,sun−to−el,year ηtot,sun−to−th,year
where ηtot,sun−to−el,year and ηtot,sun−to−th,year are the conversion efficiencies of the electric
dish and thermal dish respectively, based on the effective Annual Direct Normal Irradi-
ation ADN Ief f . This efficiency is representative of the yearly energetic performance of
the plant.
Coherently with the previous discussion, best efficiency points appear at the highest
steam utilization. However, unlike the I law efficiency, a maximum of 12.5% is found at
a medium value of current density (0.4 A/cm2 ), because the two contributions of power
at the denominator are weighted differently, based on the value of conversion efficiency
featured by electric and thermal dishes. At high current density, the increment in the
electric power requirement slightly prevails over the thermal power reduction.
96
Chapter 6
Figure 6.11: Thermal dish aperture area as function of Uf , for j equal to 0.2 A/cm2 and
0.8 A/cm2
In a best case scenario (see table 5.4), the improvement in the solar conversion effi-
ciencies obviously translates into the attainment of a higher solar to hydrogen efficiency
(figure 6.13): an upgrade by 30% is observed, compared to the base case values. The
required number of electric dishes and the aperture area of the single thermal dish re-
duce, accordingly. Table 6.1 summarizes the number of dishes for both scenarios. The
percentage value in brackets denotes the reduction in the aperture area of the thermal
dish, with respect to the base case.
Table 6.1: Number of dishes for solar only approach in base case and best case scenarios.
Values refer to the best ηsun−to−H2 ,year point
97
Plant results
0.125 80
0.125
0.124
Solar to hydrogen efficiency [-]
0.124
0.123 70 0.124
Figure 6.12: Solar to hydrogen efficiency map for solar only base case
80 0.179
0.178
Solar to hydrogen efficiency [-]
0.174
0.175 0.175
60
0.172
0.173 0.173
0.17
50 0.171
0.171
0.168 0.169
0.169
40 0.167
0.2 0.3 0.4 0.5 0.6 0.7 0.8
2
Current density [A/cm ]
Figure 6.13: Solar to hydrogen efficiency map for solar only best case
98
Chapter 6
Hybrid A total of 14 solar dishes is obtained, one of which is thermal. Again the
aperture area variation with the operating conditions is reported in figure 6.14.
Figure 6.14: Thermal dish aperture area as function of Uf , for j equal to 0.2 A/cm2 and
0.8 A/cm2 . Comparison with solar only values is portrayed.
99
Plant results
0.152 80
0.152
0.151
Solar to hydrogen efficiency [-]
0.15 0.151
70
Figure 6.15: Solar to hydrogen efficiency map for hybrid best case
0.216 80
0.216
0.216
0.214
Solar to hydrogen efficiency [-]
0.212 70 0.214
0.214
Steam utilization [%]
0.208 60
0.21 0.21
0.206
0.208 0.208
50
0.204 0.206 0.206
0.204 0.204
0.202
0.202
40 0.202
0.2 0.3 0.4 0.5 0.6 0.7 0.8
2
Current density [A/cm ]
Figure 6.16: Solar to hydrogen efficiency map for hybrid best case
100
Chapter 6
efficiency point, the resulting hybrid efficiency is 18.4% in the base case and 26.4% in
the best case.
The number of dishes for both cases is summed up in table 6.2.
Table 6.2: Number of dishes for hybrid approach in base case and best case scenarios.
Values refer to the best ηsun−to−H2 ,design point
An additional pressure drop contribution needs to be accounted for, owing to the presence
of the condensers. So the discharge pressure of the pump is slightly higher compared to
the no sweep case, and it is equal to 1.216 bar. The electrolyser block operates at 1.132
bar.
Similarly to the no sweep gas case, a multiple sensitivity analysis is carried out by
changing the steam utilization in the same range (40 to 80%) and current density from
0.3 to 1 A/cm2 .
Since we are imposing the same hydrogen throughput, the total amount of cells does
not change from the previous case, except for small differences, due to the redesigned
withdrawal of the H2 recirculation stream. The numbers are listed in table 6.3, for
completeness.
Table 6.3: Number of cells. Values refer to the 60% Uf simulation case
Current density [A/cm2 ] 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
N cells 938 703 561 467 400 350 311 279
The voltage map is firstly introduced (figure 6.17). It can be observed that, at a given
current density, the voltage level reached by the SOEC is lower than in the no sweep
case, meaning that the cell is better performing from an electrochemical standpoint. The
isoline at near 1.285 V identifies the thermoneutral voltage operation. This occurs at a
greater current density (in the range [0.6 0.8] A/cm2 depending on Uf ) compared to the
no sweep case, consistently with the results of the sweep sensitivity analysis carried out on
101
Plant results
80
1.3
1.18
1.2
1.22
1.24
1.26
1.28
1.3
1.31
3
1.3
1.3
5
2
70
1.16
Steam utilization [%]
1.25
Voltage [V]
60
1.2
1.31
1.3
1.285
1.26
1.24
1.22
1.18
1.2
50
1.16
1.15
40
0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
2
Current density [A/cm ]
the SOEC. It can be noticed that the current density at which the thermoneutral voltage
occurs increases as the steam utilization decreases. In fact the hydrogen generated at
the fuel electrode is more diluted with steam and so the Nernst potential undergoes a
reduction. The heat flux associated to the electrolysis reaction (Qreact , see equation
2.7) tends to increase and it is necessary to go to a higher loss condition to offset that
reaction thermal flux. Basically working with a lower Uf brings about the same effect of
employing a larger amount of sweep gas (the produced oxygen becomes more diluted).
The graph below (figure 6.18) offers the variation of Qreact with the steam utilization
at the thermoneutral voltage. The resulting average current density is also indicated in
labels for each Uf case.
The decrease in the operating voltage of the cell implies a reduced electric power
absorbed by the SOEC and this reflects in the trend of the total electric power required
by the plant, as portrayed in figure 6.19.
The trend of Pel,tot with Uf is opposite with respect to the no sweep situation, since the
SOEC electric consumption positive variation prevails over the negative contribution of
the other equipment (apart from the auxiliary of the sweep gas condenser, that sums
to the SOEC). This is visualized in figure 6.20. Moreover the variability with Uf of the
auxiliaries’ electric demand is transferred to the condensers, whose presence relieves the
102
Chapter 6
Figure 6.18: Reaction thermal flux and corresponding average current density (in labels)
at thermoneutral as function of Uf
80
265
26
242
244
246
250
248
252
254
256
258
262
260
260
70
240
Steam utilization [%]
255
Pel,tot [kW]
250 60
260
262
258
256
254
252
250
248
246
242
244
245
50
240
240
235 40
0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
2
Current density [A/cm ]
103
Plant results
80
60
Figure 6.20: Electric power variation with Uf . Values refer to the 1 A/cm2 simulation
case
Solar only The variability in the required number of dishes is slightly higher as com-
pared to the no sweep case. Therefore in the following table 6.4, the minimum and
maximum values of electric dish and thermal dish numbers are indicated, for both base
and best case scenarios. In each row, the first value refers to the 0.3 A/cm2 40% Uf
simulation case, while the second value refers to the 1.0 A/cm2 80% Uf simulation case.
104
Chapter 6
400 80
350
152
132
70
340
300
260
240
220
202
192
182
172
162
300
Qth,tot [kW]
142
60
250
200
50
15
2
380
340
300
260
240
220
202
192
182
17
16
150
2
40
0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
2
Current density [A/cm ]
80
0.5
0.5
0.36
0.38
0.4
0.42
0.44
0.46
0.48
0.52
70
Steam utilization [%]
I law efficiency [-]
0.45
60
0.4
0.5
0.48
0.46
0.36
0.44
0.38
0.4
0.42
50
0.35
40
0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
2
Current density [A/cm ]
105
Plant results
The aperture area range for the single thermal dish is also specified. The reduction by
12.5% of the thermal solar field area in some cases brings about the same number of
thermal dishes but with lower aperture, in other cases reduces the number of thermal
dishes, that will be characterized by an increased aperture.
Table 6.4: Number of dishes for solar only approach in base case and best case scenarios
The maps related to the solar to hydrogen efficiency, as defined in equation 6.2, in
the base case and best case are provided below in figure 6.23. Due to the larger share
of thermal power demand, the resulting values are lower than in the no sweep case.
Hence, the Qth,tot decrease prevailing over the Pel,tot increase at greater j, shifts the best
efficiency point at the highest current density (1 A/cm2 ): 11.4% of ηsun−to−H2 ,year is
attained in the base case, 16% in the best case.
Hybrid The total number of solar dishes is reported in table 6.5. The number of
electric dishes is close to the no sweep scenario, while the main difference is in the
thermal dish, whose total aperture is greater than the corresponding no sweep case.
Table 6.5: Number of dishes for hybrid approach in base case and best case scenarios
In the same fashion of the solar only approach, it is shown in figure 6.24 the solar to
hydrogen efficiency (as defined in equation 6.3) for the base and best case scenarios. For
the same reason explained in the solar only approach, the best efficiency point is again
located at the highest simulated current density and steam utilization, both in the base
and best scenarios. The thermal power decrease at high j and Uf dominates, there is
no trade off. The value of ηsun−to−H2 ,design achieved in the base case is 13.9%.
106
Chapter 6
0.114 80
0.102
0.104
0.106
0.108
0.109
0.11
0.112
0.111
0.112
0.113
Solar to hydrogen efficiency [-]
70
0.11
0.1
0.102
0.104
0.106
0.108
0.109
0.11
0.1
0.104
12
11
50
0.102
0.1 40
0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
2
Current density [A/cm ]
(a)
0.16 80
0.136
0.14
0.144
0.146
0.148
0.15
0.152
0.154
0.156
0.155
Solar to hydrogen efficiency [-]
70
Steam utilization [%]
0.15
60
0.145
0.1
58
0.136
0.14
0.144
0.146
0.15
0.148
0.15
0.1
0.1
56
54
0.14 50
2
0.135
40
0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
2
Current density [A/cm ]
(b)
Figure 6.23: Solar to hydrogen efficiency maps for the solar only approach: (a) base
case, (b) best case
107
Plant results
80
0.138
0.139
0.126
0.128
0.132
0.13
0.133
0.134
0.137
0.135
0.136
0.136
Solar to hydrogen efficiency [-]
70
Steam utilization [%]
0.134
0.132
60
0.13
0.
13
8
0.1
0.126
0.128
0.13
0.132
0.133
0.134
0.13
0.1
0.128
37
36
50
5
0.124
0.126
0.124
40
0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
2
Current density [A/cm ]
(a)
80
0.19
0.192
0.188
0.162
0.166
0.17
0.174
0.178
0.18
0.182
0.184
0.186
0.19
Solar to hydrogen efficiency [-]
0.185
70
Steam utilization [%]
0.18
0.175 60
0.17
0.1
0.162
0.166
0.174
0.178
0.17
0.18
0.18
0.1
0.1
88
84
86
50
2
0.165
0.16
40
0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
2
Current density [A/cm ]
(b)
Figure 6.24: Solar to hydrogen efficiency maps for the hybrid approach: (a) base case,
(b) best case
108
Chapter 6
The improvement in the solar to electric efficiency of the best case brings the highest
ηsun−to−H2 ,design to a value of 19.2%, which is 27.6% higher than the base case value.
This point features the lowest aperture area of the thermal dish, equal to 170 m2 .
It is finally reported the resulting hybrid efficiency, as defined in equation 6.4, for
the best ηsun−to−H2 ,design point: 15.9% in the base case, 21.6% in the best case.
As previously pointed out in the Model results (section 4.4), it is advantageous to operate
the cell under pressure owing to the decrease in the irreversibilities within the SOEC
itself. As a consequence, the SOEC absorbs a lower amount of power and there is
the potential for lower electricity consumption within the plant. In this respect, it is
interesting to study the pressurized operation of the proposed plant.
Simulations are run by fixing the operating pressure of the SOEC to 5 bar. This
choice is meaningful in the wake of the SOEC model results. As demonstrated in figure
4.16, at increasing operating pressure above 5 bar, polarization curves collapse and so
a further pressure elevation does not yield a substantial benefit from an electrochemical
viewpoint.
The same scheme of the atmospheric operation section is kept and design performance
maps are derived. Apart from the cell behaviour, the plant can take advantage of
the pressurized functioning: the equipment is more compact, the piping is smaller and
less bulky and the compression section is assisted due to the lower compression ratio
required. In addition to this, from the thermodynamic standpoint, the evaporation and
condensation of water are favored at higher pressure so a benefit in the heat exchange
is expected.
As expected, the performance of the SOEC is enhanced under pressure, as can be seen in
figure 6.25, which represents the SOEC efficiency. The positive influence of pressurized
operation is more evident at higher current density, where the Nernst potential increase
with pressure elevation has lower impact on the cell electrochemistry.
Consequently, also the I law efficiency improves, whose map is shown in figure 6.26:
the highest value of 73.5% is achieved at 0.9 A/cm2 80% Uf . This result correctly
matches with the reduced electric and thermal power requirements. On the one hand,
the electric consumption of the SOEC and the compressing units decreases, on the other
109
Plant results
0.815 80
0.81
81
15
0.
0.8
70
0.805
5
80
0.
8
0.
0.8
5
60
79
0.
0.795
1
0.8
50
0.79
05
0.8
0.8
79
0.
0.785
40
0.4 0.5 0.6 0.7 0.8 0.9
2
Current density [A/cm ]
80 0.73
0.72 0.72
0.72
70 0.71
Steam utilization [%]
0.7
I law efficiency [-]
0.71
0.7
0.7
0.68 60 0.69
0.69
0.68
0.68
0.66 0.67
50 0.67
0.66
0.66
0.65
0.64 0.65
0.64
40
0.4 0.5 0.6 0.7 0.8 0.9
2
Current density [A/cm ]
110
Chapter 6
80
830
810
820
750
760
770
820
780
790
800
70
740
Steam utilization [%]
800
Tfuel,out [°C]
780
60
81
0
800
790
780
770
760
760
750
50
730
740
740
40
0.4 0.5 0.6 0.7 0.8 0.9
2
Current density [A/cm ]
a lower enthalpy difference has to be carried out by the thermal dish. In fact the feed
stream prior to the thermal dish is characterized by a higher saturation temperature
(153 ) compared to the atmospheric situation (103.5 ). The major improvement
takes place where more steam is involved, so at low j and Uf .
From the map reported in figure 6.27, it is worth observing that downstream the
SOEC, the outlet temperatures of the fuel stream are lower, as a result of the different
thermal operation of the SOEC under pressure. Therefore, it is possible to work at
higher current density, that is why 0.9 A/cm2 was also simulated.
Solar only Related maps of solar to hydrogen efficiency are in figure 6.28. The range
of variability of this parameter is from 12.5% to 13.2% in the base case, while from 17.7%
to 18.9% in the best case, thus higher values are achieved under pressure.
Hybrid This case features the lowest possible solar field section and it results therefore
interesting as a cost effective solution. The number of electric dishes is listed in table
6.6; as for the single thermal dish, a reduction in the aperture area is detected, up to
3.3% at low steam utilization.
Maps of solar to hydrogen efficiency in figure 6.29 show increased values with respect to
the atmospheric case, with a range of variability between 15.2% and 16.1% in the base
case, and between 21.3% and 22.9% in the best case.
111
Plant results
0.132 80 0.132
0.131
Solar to hydrogen efficiency [-]
0.131 0.131
0.13 70
Steam utilization [%]
0.129 0.13 0.13
0.128 60
0.129 0.129
0.127 0.128
0.128
0.126 50 0.127
0.127
0.126
0.125 0.126
0.125
40 0.125
0.4 0.5 0.6 0.7 0.8 0.9
2
Current density [A/cm ]
(a)
80
0.188
0.188
0.188
Solar to hydrogen efficiency [-]
0.186 70
Steam utilization [%]
0.186 0.186
0.184
60
0.184 0.184
0.182
0.182 0.182
0.18 50
0.18 0.18
0.178
0.178 0.178
40
0.4 0.5 0.6 0.7 0.8 0.9
2
Current density [A/cm ]
(b)
Figure 6.28: Solar to hydrogen efficiency maps for the solar only approach: (a) base
case, (b) best case
112
Chapter 6
0.16 80
0.16 0.16
Solar to hydrogen efficiency [-] 0.159
0.158 70
0.159 0.159
0.228 80 0.228
0.228
0.226
Solar to hydrogen efficiency [-]
70 0.226
0.224 0.226
Steam utilization [%]
0.222 0.224
0.224
60
0.22 0.222 0.222
0.216 0.216
0.214
40
0.4 0.5 0.6 0.7 0.8 0.9
2
Current density [A/cm ]
(b)
Figure 6.29: Solar to hydrogen efficiency maps for the hybrid approach: (a) base case,
(b) best case
113
Plant results
Table 6.6: Number of electric dishes for solar only and hybrid approach in base case and
best case scenarios. Values refer to the best efficiency point
The trade off between the electric power increase and the thermal power decrease at
high j leads to locate the best solar to hydrogen efficiency point at 0.6 A/cm2 80% Uf
simulation case. The following table 6.7 gathers the values of solar efficiency for the best
efficiency point.
Table 6.7: Solar efficiencies for solar only and hybrid approach in base case and best
case scenarios. Values refer to the best efficiency point
For the present case, equivalent considerations can be drawn on the positive implications
of a pressurized plant operation. Hence the maps associated to the previously defined
efficiencies are just exposed.
The I law efficiency map (figure 6.30) reports greater values compared to the ambient
case: an improvement of about 5% is observed for all the simulated points. The best
efficiency point is always placed at 1 A/cm2 , 80% Uf , with a value of 56.7%, that is
5.4% higher than the ambient case value.
Solar only Figure 6.31 represents the solar to hydrogen efficiency maps, in the base
and best scenarios: the same trend is observed compared to the atmospheric case, but
greater values are achieved. The location of the best efficiency point does not change
from the atmospheric case.
Hybrid Similarly to the solar only discussion, the maps associated to the solar to
114
Chapter 6
0.56 80
0.54
0.5
0.53
0.46
0.47
0.48
0.49
0.5
6
0.5
0.51
0.52
0.54
5
70
0.5
0.45
60
0.5
0.48
0.5
4
0.5
0.5
0.5
0.49
0.48
0.46
0.47
3
2
1
0.46 50
0.44
0.44
40
0.5 0.6 0.7 0.8 0.9 1
2
Current density [A/cm ]
hydrogen efficiency (figure 6.32) reveal the improvement of such efficiency, compared to
the atmospheric situation.
An overview table 6.8 is offered, which contains the different efficiency results for oper-
ation of the SOEC at 1 A/cm2 , 80% Uf , in comparison to the corresponding values of
the atmospheric scenario.
Table 6.8: Comparison of solar efficiencies between pressurized and atmospheric opera-
tions. Values refer to the best efficiency point
Finally, table 6.9 summarizes the number of involved solar dishes relative to the best effi-
ciency point, for all the cases. The negative number in brackets stands for the reduction
of Nel compared to the ambient case.
As for the thermal dishes, the percentage reduction of aperture area varies between 1.4%
115
Plant results
0.122 80
0.116
0.12
0.117
0.121
0.118
0.119
0.121
Solar to hydrogen efficiency [-]
0.12 70
Steam utilization [%]
0.119
0.118
60
0.117
0.115
0.1
0.11
0.1
0.1
2
0.1
0.116
17
19
18
50
6
0.115
0.114
40
0.5 0.6 0.7 0.8 0.9 1
2
Current density [A/cm ]
(a)
80
0.17
0.168
0.16
0.162
0.17
0.164
0.166
0.168
Solar to hydrogen efficiency [-]
70
Steam utilization [%]
0.166
0.164
60
0.162
0.158
0.
16
0.16
0.16
8
0.1
0.1
0.16
66
6
50
4
2
0.158
0.156
40
0.5 0.6 0.7 0.8 0.9 1
2
Current density [A/cm ]
(b)
Figure 6.31: Solar to hydrogen efficiency maps for the solar only approach: (a) base
case, (b) best case
116
Chapter 6
80
0.147
0.148
0.142
0.143
0.144
0.145
0.146
0.148
Solar to hydrogen efficiency [-]
70
60
0.144
0.
14
0.14
0.1
0.14
0.1
7
0.
0.14
14
44
0.142
45
50
6
2
3
1
0.14
40
0.5 0.6 0.7 0.8 0.9 1
2
Current density [A/cm ]
(a)
0.205 80
0.20
0.19
0.192
0.20
0.194
0.2
0.196
0.198
4
Solar to hydrogen efficiency [-]
0.2 70
Steam utilization [%]
0.195 60
0.188
0.
20
0.2
2
0.19
0.19
0.1
0.1
0.19
0.19
98
96
50
4
2
0.185
40
0.5 0.6 0.7 0.8 0.9 1
2
Current density [A/cm ]
(b)
Figure 6.32: Solar to hydrogen efficiency maps for the hybrid approach: (a) base case,
(b) best case
117
Plant results
Table 6.9: Number of dishes for solar only and hybrid approach in base case and best
case scenarios. Values refer to the best efficiency point
The multi-dish configuration intrinsically raises the issue of how to choose the mutual
position of solar dishes, with the objective of optimizing the energy production on behalf
of the solar field itself. A rigorous evaluation of the solar field layout and shading
implications is beyond the scope of this work. However it is worth understanding the
mechanisms behind the design of a solar field and, after that, proposing possible dish
arrangements.
Hybrid It is considered the size of the solar field obtained in the no sweep atmo-
spheric case, at the best solar to hydrogen efficiency point (15.2%). In the framework
of regularly-spaced arrangements, since the instance does not deal with a large number
of solar dishes (14), the simplest design layout is chosen, namely rectangular, with null
offset and rotation. To make a decision on the value of the Aspect Ratio, a sensitivity
analysis on this parameter is performed.
Accordingly, in figure 6.33, it is reported the trend of the mean Shade Factor, Ground
Cover Ratio and total land area, as function of the Aspect Ratio. In particular the N-S
spacing is kept fixed to 15 m, allowing neighboring dishes to be 3 m distant. Conversely,
the E-W spacing is left to raise, up to a value of 75 m, corresponding to AR equal to
5. In fact, when θr = 0, arrays are arranged with larger collector separation oriented
East-West, due to longer shadows cast in the E-W direction when the sun is at low
angles [20].
As can be noticed, a trade-off is established: for low values of AR, dishes are closer
in relation to each other with consequent land saving, but to the detriment of a more
prominent shading effect. Viceversa, by increasing the spacing between concentrator
118
Chapter 6
0.6 12000
SF
GCR
0.5 Total land 10000
0.3 6000
0.2 4000
0.1 2000
0 0
1 1.5 2 2.5 3 3.5 4 4.5 5
AR [-]
Figure 6.33: Mean Shade Factor, Ground Cover Ratio and total land area as function
of Aspect Ratio
columns, the interaction among them progressively drops, resulting in a lower value of
SF ; nevertheless, the GCR experiences a reduction, causing a greater land occupation.
The GCR here is computed by considering the minimum possible land area reserved by
the dishes.
Downline of this analysis, it is considered a fine choice to opt for AR equal to 1.7. The
complete characteristics are available in table 6.10 and the aerial view of the resulting
solar field is in figure 6.34.
In the corresponding best scenario, the number of dishes further reduces to 10. Aim-
ing for efficiency optimization, a new arrangement is then proposed, in which 9 dishes are
positioned at the vertices of a regular polygon and one (the thermal dish, in this case) is
placed in the center. Figure 6.35 presents the aerial view and the 3D representation at
one specific time instant. Reminding a blossom shape, this design will be nicely named
pistil layout.
Results show that, keeping the same value of GCR of the base scenario, the mean SF
improves to 4.6%. So this kind of arrangement can be accounted as a good alternative
to the rectangular one.
Concerning the sweep gas case, since the solar field size is not heavily modified, an
analogous setting is maintained and these are the results (gathered in table 6.11).
Again it can be noted that the pistil layout brings about a lessening in the mean
119
Plant results
50
40
30
y [m]
20
10
-10
0 10 20 30 40 50 60 70 80
x [m]
Figure 6.34: Aerial view of the rectangular solar field layout, for the selected no sweep
base case
Table 6.10: Results of shading calculation for the hybrid base case, in absence of sweep.
Values refer to the best efficiency point
Parameter Hybrid
Base
N dishes 14
Layout Rectangular
AR 1.7
Offset 0
θr 0
Total land [m2 ] 4057
GCR 0.376
SF mean [%] 5
solar field shading; equivalently it is possible to say that keeping the same value of SF ,
a lower area of the terrain is required in the best configuration.
Solar only On the grounds that a larger number of solar dishes (43) is involved,
an offset of 0.5 is added, creating the so called diamond layout. This allows raising
120
Chapter 6
30
20
10
y [m]
-10
-20
-30
-30 -20 -10 0 10 20 30
x [m]
(a)
z
= 65.00° - = 125.00° shade
= 5.31%
10
z [m]
5
0
20
20
0
0
-20
y [m] -20
x [m]
(b)
Figure 6.35: (a) Aerial view of the solar field layout in the best scenario and (b) its 3D
°
representation at Zenith 65 and Azimuth 125 °
121
Plant results
Table 6.11: Results of shading calculation for the sweep atmospheric case. Values refer
to the best efficiency point
Parameter Hybrid
Base Best
N dishes 15 11
Layout Rectangular Pistil
Total land [m2 ] 4551 3380
GCR 0.385 0.385
SF mean [%] 5.5 5.1
140
120
100
80
y [m]
60
40
20
Figure 6.36: Aerial view of the diamond solar field layout, for the selected no sweep base
case
Given these settings, the resulting mean SF is 6.6%. This is in line with the value of
average annual Shade Factor assumed in the solar only dimensioning procedure, therefore
this solar field layout could be considered as a suitable design solution. The minimum
required land area is 13775 m2 , leading to find a GCR of 0.36.
Imagine that dishes are located in a random way, as in figure 6.37 for instance. In
122
Chapter 6
160
140
120
100
y [m]
80
60
40
20
0
0 20 40 60 80 100 120 140 160 180 200
x [m]
Figure 6.37: Example of random layout, for the selected no sweep base case
response to this, it is obtained about the same Shade Factor (6.7%) but at the expense
of a larger soil occupation, as demonstrated by the reduced value of GCR (0.237) with
respect to the previous regular configuration. In addition to this, the solar field would be
highly unbalanced from the point of view of shading: some isolated dishes would be well-
performing while some other dishes, very close to each other, would be characterized by
a high overlap fraction and thus by a worse energy conversion efficiency. This example
is offered to stress that it is essential to study the suitable layout of the solar field,
especially when the availability of land is a limiting factor.
123
Plant results
Table 6.12: Comparison of energy efficiency for different solar hydrogen technologies
In [15] the solar tower technology was again used for large scale hydrogen production;
the system was though made more complex by the adoption of fluoride molten salt as
heat transfer fluid and a supercritical carbon dioxide cycle in the bottoming power block
section. In addition to this, a Thermal Energy Storage (TES) system was integrated
in order to keep a continuous operation, enabling the decoupling of the thermodynamic
cycle from the intermittency of the solar source.
It should be considered that the results are dependent on the amount of solar irradi-
ation in the selected site. By means of a SAM simulation, it is possible to estimate the
average annual solar to hydrogen efficiency of a photovoltaic system coupled to a PEM
electrolyser as if it were installed in the same site selected for this study (Lancaster).
The resulting average PV efficiency is 12.2%; this number is then multiplied by 70%,
considered as the state of the art PEM efficiency, so as to find out a solar to hydrogen
efficiency of 8.5%, which is at least 3 percentage points lower than the efficiency reported
in this study.
124
Chapter 7
Based on the results of plant simulations, four different design points are selected, each
one featuring the best solar to hydrogen efficiency within the considered base case sce-
nario. These are briefly summarized in the following table 7.1 and labeled with capital
letters, for the sake of clarity.
Table 7.1: Selected design operating points for economic analysis. The listed
ηsun−to−H2 ,design correspond to the hybrid base case nominal value
Characteristics A B C D
Sweep gas - - • •
Pressurized - • - •
j [A/cm2 ] 0.4 0.6 1 1
Uf [%] 80 80 80 80
ηsun−to−H2 ,design [%] 15.2 16.1 13.9 14.9
It should be noted that the base case scenario is dealt with for this analysis, due to the
deployment within the solar system of conventional MGTs (no blade cooling is applied
and TIT is limited to 850-900 ), which has already penetrated a market segment,
especially for micro-cogeneration applications. On the other hand, the ceramic MGT
development is still marked by a considerable degree of uncertainty.
125
Preliminary economic analysis
It is adopted a bottom-up model, whose logical sequence is schematized in figure 7.1 and
which has the LCOH determination as final goal.
Figure 7.1: Logical sequence of the adopted economic model. (Adapted from [72])
The following cost functions are employed for the computation of capital cost of the
equipment and related installation. Firstly, the SOEC system cost is provided specific
to the electric power absorbed. Assuming a production volume of some tens of M W
per year, the reference cost CSOEC,ref [36] is 1800 e/kW , considering a reference power
density Pdens,ref of 2800 W/m2 . So it is necessary to scale that reference cost according
to the power density that characterizes the selected case, while keeping constant the
system cost specific to the active area.
CSOEC,ref · Pdens,ref
CSOEC [e/kW ] = (7.1)
Pdens
The contributions of electrolyser block (59%), enclosures (7.9%), inverter (17.9%), trans-
port (3.7%) and foundations (11.6%) are accounted separately as a fraction of the SOEC
system cost (just outlined in brackets for each mentioned item).
Compressing units are mapped as volumetric compressors and the associated cost
function is expressed as [64]:
126
Chapter 7
Also for the economic modelling of the pump, the same reference [64] is employed.
0.2724
Ppump [kW ]
Cinv,pump [e] = 808.19 · (7.3)
0.98
2 ])
Cinv,HE [e] = 2.78 · 103.6788+0.4412·log(AHE [m (7.4)
It should be specified that, for simulation purposes, multi-flow heat exchange units were
applied; however in real manufacturing, the cooling of the anodic and cathodic streams
will be treated separately in different heat exchangers.
The cost associated to condensers is based on data available in [76],
0.67
Qcond [M W ]
Cinv,condenser [e] = C0 · (7.5)
Q0
where the parameters relative to the reference component, C0 and Q0 , are equal to 26.4
M e and 470 M W , respectively.
The estimation of the cost of flash separators is done with the assistance of Aspen
Process Energy Analyzer tool and includes the cost of manufacturing and setting.
As far as the solar field is concerned, the utilized specific costs are summarized in
table 7.2; values for the thermal engine and boiler are from [19], for the solar dish part
from [26]. It should be noted that the land cost [57] already considers the expenses for
Table 7.2: Specific costs for the equipment related to the solar field section
the site preparation; moreover the cost related to the solar MGT is derived from the
previous value of the hybrid MGT, assuming an avoided cost of the burner equal to 40
e/kW [61].
By summation of the abovementioned cost items, the Total Direct Plant cost (TDP)
is obtained. To get the Total Plant cost (TP), two steps are then necessary:
127
Preliminary economic analysis
Figure 7.2: California natural gas price. Values are in dollars per thousand cubic feet
[3]
The fixed O&M (CO&M,f ixed ) comprises maintenance material and labor cost, man
salaries (for both administrative and technical divisions), property taxes and insurance
(1% of TP). In particular the maintenance-related expenses are accounted as a fraction
of the equipment capital cost and are assumed as 4% for SOEC, 3% for the solar field
section, 1.7% for compressing units and condensers and 1.5% for separators. The SOEC
replacement cost is computed separately, assuming the stack lifetime to 7 years. It is
supposed that the system could operate autonomously and periodic interventions are
scheduled; each involved operator is presumed to be paied 40.24 e/h; the administrative
and support section constitutes 10% of the total labor cost.
Moving to variable O&M (CO&M,var ), two feedstock cost items appear, referred to
water and natural gas utilization. The cost of natural gas is site dependent, therefore
in this study it is considered its industrial price for the California state, whose trend is
128
Chapter 7
shown in figure 7.2. It is picked the most recent value available from the graph and,
manipulating the information with the necessary unit conversions, a value of 0.2867
e/kg is found. The specific cost for raw water is 0.1593 e/m3 [55].
Further economic assumptions are listed in table 7.3.
LT
X
CO&M,P V [e/y] = (CO&M,f ixed + CO&M,var ) · (1 + r)−t (7.8)
t=0
LT
X
Ctax,P V [e/y] = Ctax · (1 + r)−t (7.9)
t=0
LT
X mH2 LHVH2
EH2 ,P V [M W h/y] = · heq · (1 − ∆η · (t − 1)) · (1 + r − i)−t (7.10)
3600
t=0
129
Preliminary economic analysis
It is reported in table 7.4 a summary of the costs for each selected plant operating
condition, and the resulting LCOH, referred to as base case scenario and labeled with
BS.
Cost A B C D
CSOEC [e/kW ] 971.0 641.9 377.2 385.4
TP [Me] 1.113 1.010 1.177 1.096
CO&M,f ixed [Me/y] 0.071 0.062 0.067 0.064
CO&M,var [Me/y] 0.105 0.104 0.122 0.114
LCOH [e/kg] 7.0 6.5 7.45 6.98
eCO2 [kgCO2 /kgH2 ] 20.2 20.1 23.6 22.1
It should be remarked that specific carbon dioxide emissions are around 20 kgCO2 /kgH2
for the presented cases, which are greater than typical values from steam methane re-
forming (around 10 kgCO2 /kgH2 ). This is due to the fact that nearly 64% of the thermal
energy input is provided by the natural gas source, to ensure the target volume pro-
duction. If a TES was incorporated, the system would take advantage of the possibility
to exploit the solar source to a greater extent and, consequently, CO2 emissions could
experience a substantial reduction. The lowest values of LCOH occur for cases B and D,
confirming that operation under pressure is a promising technical solution not only from
an efficiency standpoint but also for making the present technology more economically
viable. B takes advantage of the reduced plant complexity which characterizes the no
sweep layout, whereas in D, the employment of a larger number of components is offset
by the higher operating current density. To support these results, the breakdown of
TDP is illustrated in figures 7.3 and 7.4.
130
Chapter 7
Figure 7.3: Breakdown of TDP for B economic case. Details relative to the solar field
are pointed out in the bar plot aside
Figure 7.4: Breakdown of TDP for D economic case. Details relative to the solar field
are pointed out in the bar plot aside
131
Preliminary economic analysis
Generally, the major contribution to the total investment cost is from the solar dish
section (above 70%). The sweep gas case (D) presents a larger share of cost related
to compressing and heat exchange units (20.2%), due to the additional presence of the
condensers (and the O2 separator). At the same time, the SOEC part weighs less (7.7%)
than the no sweep case (13.9%), because a higher current density (and thus a higher
power density) brings about a lowering in the specific cost of the electrolyser.
Going into the details of the solar field section, the mirrors constituting the parabolic
concentrators represent the dominant cost item, followed by the thermal receiver. The
reason lies in the early development stage of these components, especially the receiver,
for which research has recently become more active. To name a few, [13] designed a solar
volumetric receiver based on open cell ceramic foams, which is able to operate in highly
non uniform irradiation with material temperatures and stresses below the acceptable
threshold; [62], by means of a 3D numerical model, delved into the complex heat transfer
mechanisms occurring within a modified cavity receiver for its thermal optimization.
Concerning the reflector, a number of solutions with different specifications, in terms of
geometry, facets’ shape and number, tracking control, are currently available and many
manufacturing techniques are tested [48].
Featuring the minimization of the thermal duty required to the solar field section, the
no sweep gas case (B) shows reduced values in the solar-related cost items.
An optimistic cost scenario, labeled as OS, is taken into account with these features:
1. the reference investment cost for the SOEC is halved, 900 e/kW ;
2. the specific cost of the solar concentrators is reduced by 30%, 175 e/m2 ;
3. the thermal receiver capital cost is half of the BS value, 65 e/kWt .
This reflects a medium- to long-term outlook, in which the cost associated to SOEC,
mirrors and receiver, being the least mature components, could ramp down to a larger
extent. In particular for the SOEC, the reference value is consistent with FCH 2 JU
2030 target vision [22]. An analogous table 7.5 is built for the OS scenario.
It can be observed that the LCOH decreases, as expected, achieving a minimum value
of 5.3 e/kg in B: compared to the BS value, the LCOH has a decrement by 18.5%, but
still, it does not experience a steep reduction. This behaviour can be explained by
looking at figure 7.5, which visualizes the breakdown of LCOH. The major contribution
is represented by the capital cost (about 40%), immediately followed by the variable
O&M costs, while fixed O&M costs are only by 16.4% affecting the resulting LCOH.
The great impact of variable O&M acts such that the LCOH decrease is dampened. It
should be noted that the share of variable O&M would be even larger if the system
132
Chapter 7
Cost A B C D
CSOEC [e/kW ] 485.5 321.0 188.6 192.7
TP [Me] 0.820 0.752 0.920 0.853
CO&M,f ixed [Me/y] 0.050 0.045 0.052 0.050
CO&M,var [Me/y] 0.105 0.104 0.122 0.114
LCOH [e/kg] 5.6 5.3 6.3 5.9
was applied in another country, due to the natural gas price site dependency. The tax
portion is minor with respect to the other contributions and it is a note of caution in
this preliminary assessment, that not always is present when facing diverse economic
analysis in literature.
A last LCOH evaluation is done for the best performing case B, but applying the
solar only strategy. The equivalent hours are 2683 h and the total aperture area is 4703
m2 , leading to a land occupation of 13098 m2 , which is 3.2 times higher than the hybrid
soil area. The results, reported in table 7.6, indicate that the oversizing of the solar
segment weighs more and more on the capital expenses and is therefore detrimental
from an economic perspective. It should be pointed out that the variable O&M now are
133
Preliminary economic analysis
close to be null, meaning that the water consumption constitutes a minor cost in the
overall economic model.
Table 7.6: Results of economic analysis, for solar only approach, in BS and OS scenarios
Cost BS OS
TP [Me] 1.946 1.455
CO&M,f ixed [Me/y] 0.100 0.073
CO&M,var [Me/y] negligible
LCOH [e/kg] 22.8 16.9
An overview table 7.7 is subsequently offered, which gathers the results of LCOH for
different solar hydrogen production solutions proposed in literature. The work of [52]
devoted on a comprehensive review on renewable hydrogen production is utilized to build
the summary.
As a whole, solar hydrogen technologies are still not competitive, showing greater
values of LCOH with respect to the typical cost of hydrogen from steam methane re-
forming. Electrolysers are highly modular systems: while this makes the technology
very flexible with respect to hydrogen production capacity, it also limits the effects of
economies of scale, as even large electrolysers are based on identically sized cells and
134
Chapter 7
stacks.
The cost of hydrogen in this study is in line with the outcomes present in literature;
these results are considerably influenced by the utilized economic model and financial
assumptions, along with the scale of the production system, the grid connection and the
presence or not of a storage. Among the concentrating solar systems, the parabolic dish
technology to date is more expensive but, at the same time, it potentially offers much
room for improvement both from an efficiency and economic point of view.
135
Conclusion
136
Conclusion
In this study, the coupling between SOEC and a parabolic dish system for solar hydro-
gen production was investigated. The proposed solution positions itself perfectly in a
future scenario of near zero emission hydrogen production processes, with efficiency and
sustainability characteristics.
Firstly, it was essential to delve into the SOEC functioning, by means of a one di-
mensional cell model, to figure out its electrochemical and thermal behavior to changing
input operating conditions. The model validation was successful in accurately matching
the simulated cell performances with the real one, featuring a relative error well below
5%. In essence, temperatures, pressure, fuel composition, gas flow rates and current
density deeply influence the cell performance. An evaluation on the sweep gas type
shows that the deployment of a non oxygen-containing species, like steam, could benefit
from a reduced cell voltage at a given current density, leading to a higher SOEC electric
efficiency. In the case of steam, the combination of the Nernst potential decrease and the
high heat capacity lets enlarge operation in endothermic mode, with an advantageous
decrement of the electric power absorbed by the SOEC itself.
The derivation of cell internal profiles was useful to reveal how the evolution of po-
larization losses and their magnitude directly reflects on the current density distribution
and the thermal characteristic of the SOEC. Moreover, in the wake of a step varia-
tion of the steam utilization factor, the SOEC reported a less symmetric functioning
with respect to hydrogen production and enhanced its temperature gradient with Uf
elevation.
As a result of a sensitivity analysis on the sweep gas flow rate, a map of the cell net
thermal flux was obtained, which clearly identified the different thermal conditions of
the SOEC. In particular, the thermoneutral operation was found to occur progressively
at higher current densities as steam, employed as sweep gas, was increasing; at high
sweep flow rate, a more rapid rise of the heat generated by polarization losses led to a
slight reduction in the current density at the thermoneutral.
137
Conclusion
As last observation on the SOEC model, pressurized cell operation was simulated:
starting from a certain value of current density (in correspondence of the intersection
point on a V − j plane) the decrease in the overpotentials succeeded in reducing the cell
voltage, with consequent efficiency improvement.
Afterwards, the overall plant model was developed on Aspen Plus, in which the
SOEC unit was incorporated as a submodel. In pursuit of solar to hydrogen efficiency
optimization, different operating nominal conditions of the SOEC within the plant were
examined, in terms of pressure, current density, steam utilization and sweep gas flow
rate (steam), with the objective to produce a hydrogen throughput of 150 kg/h. The
following considerations could be drawn: the overall system conversion efficiency could
be enhanced by
the absence of sweep gas: this condition leads to minimize the duty required from
the thermal solar dish section, that has therefore the task to superheat only the
steam constituting the H2 O/H2 mixture. The aperture area of the thermal dish is
minimized, accordingly, with saving in its manufacturing cost. In case safety issues
arise from processing pure oxygen, the flow rate of sweep gas should be minimized
so as to fulfill the imposed safety margin.
high steam utilization: similarly to the previous explanation, this implies a lower
mass flow of steam evolving within the plant, leading to lower duty requirement.
For the system, the thermodynamic benefit of excess steam (smaller Nernst po-
tential) is outweighed by the penalties associated with handling excess steam and
incomplete heat recuperation [58].
pressurization: when increasing the SOEC operating pressure to a value of 5 bar,
not only the electrochemical performance of the SOEC itself is improved, but also
the plant can take advantage of the pressurized functioning, since the electricity
consumption linked to the compression blocks reduces as well, and the evaporation
and condensation of water are thermodynamically favored.
medium or high current density, depending on the amount of processed sweep gas.
With null or very low steam flow rate, a trade-off is given between the electric
power increase and the thermal power decrease at high j, which leads to locate the
best solar to hydrogen efficiency point at 0.4 A/cm2 in the atmospheric scenario,
0.6 A/cm2 in the pressurized scenario. The presence of a certain flow rate of steam
as sweep (corresponding to a condition in which total electric and thermal power
demands from the plant are of the same order of magnitude), shifts the best func-
tioning to the highest current density (1 A/cm2 ). The overheating generated by
138
internal irreversibilities becomes useful heat, contributing to relieving the thermal
energy that the thermal dish has to deliver to the H2 production section of the
plant.
In spite of a lower solar to hydrogen efficiency, operation with sweep gas becomes inter-
esting for its capability to attain higher current density, without incurring in dangerous
thermal gradients between inlet and outlet of the stack. In view of dynamic study of
the plant, the SOEC stack could work with a variable flow rate of the anodic and ca-
thodic streams following the solar energy availability (under high irradiance, a great
amount of thermal energy can be generated by the thermal dish and more steam could
be processed).
Hence it is possible to state that the most suitable set of operating conditions is
always a trade-off between many aspects, connected not only to the SOEC stack but to
all the balance of plant involving it. Going from atmospheric to pressurized operation,
the average solar to hydrogen efficiency, in the solar only approach, increases from 12.5%
to 13.2% for the no sweep gas case; from 11.4% to 12.2% for the sweep gas case. In a
best case scenario, characterized by the improvement in the solar to electric and solar to
thermal efficiencies of solar dishes, the abovementioned system efficiencies raise to 17.9%
and 18.9% for the no sweep case; to 16% and 17.2% for the sweep case, respectively. The
proposed system, thus, attains great performances in comparison to other possible solar
hydrogen solutions, based on photovoltaics coupled to LTE, or other CSP technologies
coupled to SOEC.
Focusing on best solar to hydrogen best efficiency points, the Levelized Cost Of
Hydrogen was determined, provided a set of economic assumptions. Results showed
that pressurization and high current density operation together move in the direction of
a more affordable solar hydrogen generation. The LCOH turns out to be 6.5 e/kg and
6.98 e/kg for the pressurized no sweep and sweep gas economic cases, respectively. These
values lower to 5.3 e/kg and 5.9 e/kg when an optimistic scenario (OS) is considered,
in which electrolyser, mirror and thermal receiver are sensibly envisaged to experience a
major cost reduction. Due to the minor technological maturity, the system proposed in
this study is still not competitive if compared to the conventional steam reforming process
for hydrogen production. Specific carbon dioxide emissions are around 20 kgCO2 /kgH2 for
the presented cases, which are greater than typical values from steam methane reforming.
If a TES was incorporated, the hybrid system could exploit the solar source to a greater
extent and, consequently, CO2 emissions could experience a substantial reduction. The
reported LCOH are in line with the results of other solar hydrogen solutions. Among
139
Conclusion
the concentrating solar systems, the parabolic dish technology to date is more expensive,
nevertheless, it potentially offers much room for improvement both from an efficiency
and economic point of view.
Off-design analysis, that will be the subject of future works, could deepen the fea-
sibility of a totally renewable solar hydrogen production, by examining the coupling of
the solar energy distribution with typical hydrogen demand profiles. The operation at
higher current density and pressure has to be further evaluated and needs to be accom-
panied by development of testing activities on cells and demo projects, for proof of the
concepts addressed in this study. On the modelling front, a further development in the
heat exchange model of the cell is envisaged, so as to achieve greater accuracy. It will be
also implemented a submodel concerning the solar dish functioning, with the purpose of
studying the optical and thermal performance of the solar thermal receiver.
As a whole, the results of this thesis could provide guidance for further development
of small scale solar hydrogen production and for cleverly reinforcing the integration
between the solar system and the electrolysis technology.
140
Bibliography
[7] Notes from the course of ’Electrochemical energy conversione and storage’, 2017
Politecnico di Milano.
[9] The California Fuel Cell Revolution, California Fuel Cell Partnership. Technical
report, 2018.
[10] A. Abánades, C. Rubbia, and D. Salmieri. Thermal cracking of methane into Hy-
drogen for a CO2-free utilization of natural gas. International Journal of Hydrogen
Energy, 38(20):8491–8496, 2013.
141
[13] L. Aichmayer, J. Spelling, and B. Laumert. Preliminary design and analysis of a
novel solar receiver for a micro gas-turbine based solar dish system. Solar Energy,
162:248–264, 2015.
[14] K. J. Albrecht and R. J. Braun. The effect of coupled mass transport and internal re-
forming on modeling of solid oxide fuel cells part I: Channel-level model development
and steady-state comparison. Journal of Power Sources, 304(December):384–401,
2015.
[15] A. A. AlZahrani and I. Dincer. Design and analysis of a solar tower based integrated
system using high temperature electrolyzer for hydrogen production. International
Journal of Hydrogen Energy, 41(19):8042–8056, 2016.
[17] M. Binotti. Linear Fresnel Reflectors: Study of the Technology and Steps toward
Optimization. PhD thesis, Politecnico di Milano, 2013.
[18] S. Campanari and P. Iora. Definition and sensitivity analysis of a finite volume
SOFC model for a tubular cell geometry. Journal of Power Sources, 132(1-2):113–
126, 2004.
[19] S. Campanari and E. Macchi. Technical and Tariff Scenarios Effect on Microturbine
Trigenerative Applications. Journal of Engineering for Gas Turbines and Power,
126(July 2004):747–757, 2004.
[20] J. Cumpston and J. Pye. Shading and land use in regularly-spaced sun-tracking
collectors. Solar Energy, 108:199–209, 2014.
[22] FCH 2 JU. Multi - Annual Work Plan. Technical report, 2018.
142
[24] K. Ghaib and F. Z. Ben-Fares. Power-to-Methane: A state-of-the-art review. Re-
newable and Sustainable Energy Reviews, 81(August 2017):433–446, 2018.
[25] A. Giostri. Preliminary analysis of solarized micro gas turbine application to CSP
parabolic dish plants. Energy Procedia, 142:768–773, 2017.
[32] IEA. Technology Roadmap - Hydrogen and Fuel Cells, 2015. Technical report.
[34] IRENA. Hydrogen from renewable power: Technology outlook for the energy tran-
sition. Number September. 2018.
[36] B. D. James and D. DeSantis. Manufacturing Cost and Installed Price - Analysis
of Stationary Fuel Cell Systems. Technical Report September, 2015.
143
[37] S. H. Jensen, P. H. Larsen, and M. Mogensen. Hydrogen and synthetic fuel pro-
duction from renewable energy sources. International Journal of Hydrogen Energy,
32(15 SPEC. ISS.):3253–3257, 2007.
[38] P. Kazempoor and R. J. Braun. Model validation and performance analysis of regen-
erative solid oxide cells: Electrolytic operation. International Journal of Hydrogen
Energy, 39(6):2669–2684, 2014.
[40] P. Kazempoor and R. J. Braun. Hydrogen and synthetic fuel production using
high temperature solid oxide electrolysis cells (SOECs). International Journal of
Hydrogen Energy, 40(9):3599–3612, 2015.
144
[47] J. C. Mah, A. Muchtar, M. R. Somalu, and M. J. Ghazali. Metallic interconnects
for solid oxide fuel cell: A review on protective coating and deposition techniques.
International Journal of Hydrogen Energy, 42(14):9219–9229, 2016.
[55] M. Moradi and M. Mehrpooya. Optimal design and economic analysis of a hybrid
solid oxide fuel cell and parabolic solar dish collector, combined cooling, heating
and power (CCHP) system used for a large commercial tower. Energy, 130:530–
543, 2017.
145
[58] J. E. O’Brien. Thermodynamic Considerations for Thermal Water Splitting Pro-
cesses and High Temperature Electrolysis. Proceeding of the 2008 International
Mechanical Engineering Congress and Exposition, pages 639–651, 2008.
[60] C. Perkins and A. W. Weimer. Likely near-term solar-thermal water splitting tech-
nologies. International Journal of Hydrogen Energy, 11(7):1587–1599, 2004.
[62] K. S. Reddy, T. S. Vikram, and G. Veershetty. Combined heat loss analysis of solar
parabolic dish - modified cavity receiver for superheated steam generation. Solar
Energy, 121:78–93, 2015.
[65] M. Romero and A. Steinfeld. Concentrating solar thermal power and thermochem-
ical fuels. Energy & Environmental Science, 5(11):9234, 2012.
[68] M. Seitz, H. von Storch, A. Nechache, and D. Bauer. Techno economic design of a
solid oxide electrolysis system with solar thermal steam supply and thermal energy
146
storage for the generation of renewable hydrogen. International Journal of Hydrogen
Energy, 42(42):26192–26202, 2017.
[70] SolarPACES, GreenPeace, and ESTELA. Solar Thermal Electricity - Global Out-
look 2016. Technical report, 2016.
[71] SOPHIA. Public summary of report on most suitable schemes for solar energy
integration. Technical report, 2015.
[77] A. Ursua, L. Gandia, and P. Sanchis. Hydrogen Production From Water Electroly-
sis: Current Status and Future Trends. Proceedings of the IEEE, 100(2):410 –426,
2012.
147
efficiency, distributed-scale electrical energy storage. Journal of Power Sources,
283:329–342, 2015.
148