Energy Conversion and Management 135 (2017) 21–34

Contents lists available at ScienceDirect

Energy Conversion and Management

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

Dynamic model with experimental validation of a biogas-fed SOFC plant

G. D’Andrea a, M. Gandiglio a, A. Lanzini a,⇑, M. Santarelli a,b
a
Energy Department, Politecnico di Torino, Torino, Italy
b
Energiteknik, Royal Institute of Technology KTH, Sweden

a r t i c l e i n f o a b s t r a c t

Article history: The dynamic model of a poly-generation system based on a biogas-fed solid oxide fuel cell (SOFC) plant is
Received 22 September 2016 presented in this paper. The poly-generation plant was developed in the framework of the FP7 EU-funded
Received in revised form 19 December 2016 project SOFCOM (www.sofcom.eu), which consists of a fuel-cell based polygeneration plant with CO2
Accepted 21 December 2016
capture and re-use. CO2 is recovered from the anode exhaust of the SOFC (after oxy-combustion, cooling
Available online 2 January 2017
and water condensation) and the Carbon is fixed in the form of micro-algae in a tubular photobioreactor.
This work focuses on the dynamic operation of the SOFC module running on steam-reformed biogas.
Keywords:
Both steady state and dynamic operation of the fuel cell stack and the related Balance-of-Plant (BoP)
Biogas
SOFC
has been modeled in order to simulate the thermal behavior and performance of the system.
Dynamic model The model was validated against experimental data gathered during the operation of the SOFCOM
Stack proof-of-concept showing good agreement with the experimental data. The validated model has been
Control used to investigate further on the harsh off-design operation of the proof-of-concept.
Simulation results provide guidelines for an improved design of the control system of the plant, high-
lighting the feasible operating region under safe conditions and means to maximize the overall system
efficiency.
Ó 2016 Elsevier Ltd. All rights reserved.

1. Introduction the stack level focuses on the mathematical description of electro-

chemical processes, chemical reactions, transport phenomena, and
Among CHP generators, fuel cells could be the option with a geometry influence. Investigation of the entire system includes,
higher emphasis toward the maximization of the energy efficiency. instead, studies on the system integration, heat and mass
Furthermore, the coupling with a renewable fuel such as biogas exchange, electrical circuits, and equipment. Intensified research
can determine an interesting synergetic effect, putting together a efforts have been made in the development of SOFC mathematical
sustainable fuel and technology with characteristics of very high models at the system level in recent years, and several dynamic
conversion efficiency. Moreover, it is possible to take advantages models have been reported as well [5–11]. Various mathematical
of other specific aspects of the SOFC-based systems, such as the models based on mass, energy and momentum conservation laws
opportunity of a simplified recovery of CO2 from the cell exhausts, and statistical data-driven models of the SOFC system have been
and thus their operation in the framework of more ambitious poly- proposed [12–15].
generative systems. Most of these algorithms are suitable only for steady state per-
In 2004, Van Herle et al. were already discussing the potential formance analysis, whereas they are not suitable for the study of
for the integration of sewage biogas with high temperature fuel complex integrated power systems in which the monitoring of
cells [1,2]. Over the last years, new research activities aimed to real-time plant performance (on load tracking) and its impact on
the demonstration of the same concept; in fact, the use of fuel cells power quality are required.
in biogas plants can lead to many advantages in terms of environ- Mathematical models of the dynamic behavior of the SOFC have
mental and economic aspects [3,4]. been proposed already in the literature [16–18]. Hall proposed a
In order to analyze the advantages of SOFC-based systems, dynamic model of a single tubular cell [18]. Nehter [16] extended
mathematical models can be developed from micro to macro scale, the dynamic modeling approach toward a two-dimensional tran-
depending on the scales investigated. The analysis of the SOFC at sient model of a multi-cell system consisting of micro-tubular cells
arranged in a cascade configuration. However, large (of approx.
250 K) thermal gradients were calculated for this configuration.
⇑ Corresponding author. Huangfu et al. [19] provided a transient 1-dimensional model of
E-mail address: andrea.lanzini@polito.it (A. Lanzini).

http://dx.doi.org/10.1016/j.enconman.2016.12.063
0196-8904/Ó 2016 Elsevier Ltd. All rights reserved.
22 G. D’Andrea et al. / Energy Conversion and Management 135 (2017) 21–34

Nomenclature

Acronyms Gi mass flow rate of i-stream [kg/s]

ASR Area Specific Resistance hi enthalpy of i-stream [kJ/kg]
CHP Combined Heat and Power Q_ thermal power [kW]
DIR Direct Internal Reforming P_ electrical power [kW]
FU Fuel Utilization T temperature [°C]
INSUL Insulation t time [s]
INTER Interconnector yi mole fraction of i-stream [-]
LPM Lumped Model f constant percentage value [-]
NG Natural Gas VOCV open circuit voltage [V]
PBR PhotoBioReactor Pel electrical power [kW]
PID Proportional Integrative Derivative i current density [A/cm2]
MEA Membrane Electrode Assemblies I total current [A]
SIM Simulation M mass [kg]
SMR Steam Methane Reforming ncell number of cells
S/C Steam to Carbon
SOFC Solid Oxide Fuel Cell Greek symbols
WGS Water Gas Shift q density [kg/m3]
WWTP Waste Water Treatment Plant k thermal conductivity [W/m K]
g thermal efficiency [-]
Symbols
cp specific heat capacity [kJ/kg K]
Vol (or V) volume [m3]

a tubular SOFC that was able to predict the non-uniform distribu- algorithms. The internal stack flow configuration is also key to con-
tions of current density, gas composition and temperature across trol and limit the in-plane thermal gradient of the SOFC. Fardadi
the fuel cell. Padulles et al. [17] provided a mathematical descrip- et al. [25] proposed a modified cross-flow arrangement with
tion of the control system of an SOFC unit. However, the thermal non-uniform air flow feeding. The control unit adjusts both air flow
response of fuel cell was not included in his work. Achenbach [5] rate and temperature resulting in calculated negligible tempera-
formulated a three-dimensional time-dependent model of an SOFC ture variations for the fuel cell, even when large load changes are
stack. Different internal flow configurations and internal methane applied.
reforming were assessed in his model. The time constants of the This paper develops a modeling approach in order to reproduce
stack were calculated showing how load changes entail a response the dynamic behavior of a proof-of-concept poly-generation sys-
time of 5 min. Achenbach [20] extended further the study of load tem based on SOFC: an integrated system, in which the SOFC is
changes on the transient behavior of the SOFC by simulating the the core element of a more complex structure with several pro-
impact of fast load changes. Results showed how a step load cesses running simultaneously.
change produces a temperature overshoot followed by a relaxation As discussed in [26–29], to meet these purposes, 0 D, 1/2/3 D
time whose duration depends on the thermal properties, size and techniques can be used. However, in the overall perspective of a
geometrical and flow configuration of the cell. However, no system integrating several components, already in operation, and
attempts at establishing the effects of a control loop on the cathode especially considering the dynamic approach, 0 D methods proved
air flow rate to moderate the temperature overshoot and relaxation to be sufficient for a suitable simulation and analysis of the system.
time were modeled in the works of Achenbach. Thus, the considered model includes multi-physics phenomena, as
The design of accurate control strategies is a critical aspect for better described below, such as thermal processes (i.e., heat
integrated SOFC systems, as described in Refs. [11,21]. Recently, exchanges and oxy-combustion of the anode exhausts), electro-
Barelli et al. [22] developed a mathematical model of an SOFC stack chemical reactions in the SOFC stack, and chemical reactions
system for which the cathode air flow rate is dynamically adjusted occurring in the fuel processor.
by a PID controller unit. Nearly a constant stack temperature In particular, the model has been validated making use of the
behavior has been calculated even during fast load changes by experimental results obtained from the first-of-a-kind proof-of-
rapidly varying the PID-controlled cathode airflow rate. The simu- concept of this type of systems, developed by the Authors. This
lations were carried out for a 4-cell stack. The fast adapting tran- work has been developed in the framework of the SOFCOM Euro-
sient behavior of the stack temperature might be ascribed to the pean Project (2010–2015), where a poly-generation proof-of-
low thermal mass of the modeled fuel cell stack. Looking to their concept, able to produce electricity from an SOFC, recover heat
dynamic model assumptions, the thermal capacity of the metallic from exhaust streams and capture CO2 for algae growth in a pho-
interconnected has been not included thus yielding a low thermal tobioreactor, has been designed, built and tested.
inertia of the solid stack structure. The validated model has been used to investigate the perfor-
Recently, Barelli et al. have experimentally shown on an SOFC mance of the demonstration plant in specific operating condition.
short stack how the fuel composition affects the amount of cathode In particular, the following off-design conditions have been ana-
air flow rate that is needed to maintain a constant operating tem- lyzed: (i) the impact of a possible malfunction of the coolant air
perature [23]. Wu and Gao [24] developed an extensive control regulation system, (ii) the stack performance under different
methodology for the SOFC that is based a non-linear predictive degrees of direct internal reforming (which could be also con-
and that includes fault diagnostics. They proposed a fault- nected to loss of catalytic activity of the external steam reformer)
tolerant control strategy that makes use of artificial neural network and, finally, (iii) the influence of a sudden current load change.
 G. D’Andrea et al. / Energy Conversion and Management 135 (2017) 21–34 23

2. Methodology 2.2. Lumped models for reformer and SOFC stack

2.1. Mathematical modeling the integrated system A schematic block diagram for the lumped volume model of the
SOFC stack + biogas reformer is shown in Fig. 3 [27,28,34]. The
The SOFCOM [30] proof-of-concept is composed of a biogas block diagram in Fig. 4 summarizes the model implementation.
clean-up unit (for the removal of H2S and siloxanes), gas pre- All the used assumptions were reported in a previous publication
heaters, an evaporator to feed the methane reformer, a catalytic [35].
reactor for steam reforming, an SOFC stack, an oxy-combustor of By applying an electrical load to the cell, the rate of fuel intro-
the anode off-gas stream, a water condenser, a dryer of the residual duced into the system is determined. According to the fuel compo-
CO2 stream, and finally a photo-bio-reactor (PBR) for the CO2 recy- sition (methane content), the Steam-to-Carbon (S/C) ratio is
cling in microalgae (see Fig. 1). The oxy-combustor is installed in imposed, and the vapor flow rate is calculated. The fuel, mixed
order to recover pure CO2 from the anode exhaust stream: in fact, with the high temperature water vapor, reaches the reforming sec-
it is able to convert all the residual fuel (mainly H2 and CO) into tion where the steam methane reforming and water gas shift reac-
H2O and CO2. The resulting oxy-combusted anode exhaust thus tions are considered by imposing chemical equilibrium of the
consists mainly of H2O and CO2. Then, a relatively simple con- mixture. The reformate gas is then flushed into the SOFC anode
denser unit is sufficient to recover a high-purity CO2 stream [31– at a temperature of around 750 °C; here, the chemical and electro-
33]. chemical model determine the fuel conversion rate and the power
The developed model takes into account the above-listed com- production.
ponents at macroscopic lumped-volume scale, in order to analyze The main assumptions of the stack model are:
their transient behavior. In the description and analysis developed
in this paper, the clean-up system, oxy-combustor, water con-
adiabatic control volume;
denser and the PBR have been neglected since the focus is the
the same temperature of cathode and anode outlet streams;
dynamic behavior of the SOFC island.
anode inlet stream composed of a mixture of CH4, CO, CO2, H2,
Fig. 2 shows the strategy adopted. The core component of the H2O;
SOFCOM proof-of-concept, the integrated SOFC stack biogas refor-
stack current density (A/cm2) and fuel utilization are given as
mer, is modeled using a lumped-parameters MATLABÒ code (SOFC input values.
LPM), which solves a set of equation that includes molar balances
(based on the electrochemical reactions within the SOFC), steam Reforming and shifting reactions are taken at equilibrium con-
reforming and water-gas shift chemical reactions based on the dition and as a function of temperature, using the polynomial
equilibrium constants of the two reactions, and the energy balance expression given below (coefficients of the polynomial were taken
of the overall ‘hot box’. The LPM code is nested loop inside a second from Ref. [35]):
model, which is a SIMULINKÒ Dynamic Model. A detailed descrip-
tion of the two models is provided in the following paragraphs. log K p ¼ A  T SOFC;4 þ B  T SOFC;3 þ C  T SOFC;2 þ D  T SOFC þ E ð1Þ

Fig. 1. SOFCOM proof-of-concept layout.

24 G. D’Andrea et al. / Energy Conversion and Management 135 (2017) 21–34

Fig. 2. Model’s strategy adopted.

Fig. 3. Lumped volumes for reformer and SOFC stack.

Fig. 4. Flow diagram of reformer and SOFC stack according to the lumped volume modeling approach.

where 2.3. Dynamic model of the integrated system

– Kp is the equilibrium constant of the reaction;
– TSOFC,i is the temperature at which chemical reactions occurs, in 2.3.1. Overall approach
K; The model reproduces the transient behavior of the integrated
– A, B, C, D, E are polynomial coefficients taken from [35]. system: a schematic block diagram of the dynamic model is shown
in Fig. 5.
It was assumed that only H2 reactant moles are responsible for The main material flows of the system are the anodic streams
the electric flow. A further simplification is the use of an overall (green1 lines, biogas) and the cathodic streams (orange lines, air).
area specific resistance (ASR) parameter, provided by the manufac-
turer as a function of the SOFC working temperature, instead of the 1
For interpretation of color in Fig. 5, the reader is referred to the web version of
polarization terms (ohmic, diffusion and activation losses). this article.
 G. D’Andrea et al. / Energy Conversion and Management 135 (2017) 21–34 25

Fig. 5. Conceptual map of the dynamic model.

The anode line deals with the pre-treatment of the fuel before feed- Balance-of-Plant. Each plant section has been modeled starting
ing the SOFC; the three main components are, from left to right: from information available either from on experiments performed
evaporator, reformer, and electric heater. The flow rate of fuel is during the field activity or from the manufacturer of the compo-
determined according to the electric load applied and the molar nent itself. For auxiliary components such as heaters and heat
composition of the biogas. Biogas and liquid water are mixed in exchangers, some simplifying assumptions were taken.
the evaporator component to reach the desired S/C ratio. In the evap- Each component is modeled by applying the energy conserva-
orator, water is first vaporized before mixing the biogas. The mixed tion equation for open systems (given in the differential form):
humid gas is finally preheated to reach a temperature of about
650 °C.
out
dT x Xn   X j
ðCpx  qx  V x Þ  G_ i  hi ðT in
in out
¼ out
x Þ  hi ðT x Þ þ Ux;k ð2Þ
The pre-heated fuel stream feeds the reformer where both SMR dt i¼1 k¼1
and WGS reactions occur. The methane is almost completely con-
verted to produce a hydrogen and carbon monoxide rich reformate where
gas. The required thermal power for the endothermic reformer
reactions (SMR) was supplied by electrical heaters, located around – Cpx is the heat capacity of xth component, in [kJ/kg K];
the reactor. In a real-size plant, the reformer would be either ther- – qx is the density of xth component, in [kg/m3];
mally integrated with the exothermic SOFC stack, or heated exter- – Vx is the volume of xth component, in [m3];
nally by the fuel cell exhaust [36]. Furthermore, an additional – T out and T in
x x are the outlet/inlet temperatures of xth component,
electrical heater was inserted between the reformer and the SOFC in [K];
in order to provide the thermal power required to compensate for in out
– hi and hi are the inlet/outlet enthalpies of ith compound, in
heat losses along the piping and thus maintain the required inlet
[kJ/mol];
temperature at the SOFC anode inlet.
– G_ i is mass flow rate of ith compound, in [mol/s];
The cathode line provides air (oxygen) to the stack that is
– Ux.k is the incoming/leaving heat flux of xth component, in
required both as a reactant and for the internal stack cooling. An
[kW].
air blower is used to flow ambient air through the pre-heater sec-
tion. The blower speed is controlled by an inverter, while the flow
The integrated reformer SOFC stack model include a dedicated
rate is measured by a rotameter. In the dynamic model, the air
chemical and electrochemical model. In the reformer section, both
blower is assumed to follow the load quickly, so its dynamics has
SMR and WGS occur.
been neglected. Then a heat exchanger heats up the inlet air
The calculation procedure for the reformer is the following:
(450–600 °C) recovering heat from the cathode exhaust, released
out at a high temperature (850 °C).
1. given the reformer core temperature, the equilibrium constants
The nominal inlet temperature of the air is 650 °C. In order to
are evaluated according to [35];
reach this set-point, an electrical heater is installed (again, a not
2. given current density and fuel utilization of the stack and biogas
optimized solution, useful only in case of a proof-of-concept), to
composition, the inlet biogas flow rate is calculated;
provide the extra thermal power needed. The stack core tempera-
3. given the biogas (and steam) inlet molar flow and the equilib-
ture should never go above 860 °C. The thermal control is carried
rium constants for reforming reactions, the reactions products
out by regulating the amount of cathode air fed through the air
are calculated by the equation solver;
blower: a retroactive feedback control calculates and imposes the
4. the endothermic and exothermic heat of SMR and WGS reac-
mass flow rate needed to keep the core temperature below this
tions are evaluated in order to evaluate the thermal balance
critical level.
of the reformer component;
Tables 1–4 give the input parameters to the model for each
5. the gross power required for the SMR is regulated by an exter-
component.
nal PID controller, introduced to keep the reformer at the con-
stant temperature of 800 °C.
2.3.2. Equations and hypothesis
A ‘‘black box” modeling technique is used to model all the The reformer dynamic thermal behavior is finally modeled by
system components, including the auxiliaries of the overall applying the energy conservation equation for open systems:
26 G. D’Andrea et al. / Energy Conversion and Management 135 (2017) 21–34

Table 1
SOFC stack input parameters.

SOFC stack Parameter Value Unit Ref.

MEA properties kMEA 2 W/m K [37,38]
cp MEA 600 J/kg K [37,38]
qMEA 7500 kg/m3 [37,38]
Interconnector properties kINTER 25 W/m K [39]
cp INTER 627 J/kg K [39]
qINTER 7500 kg/m3 [39]
Insulation properties cpINSUL 1130 J/kg K [40]
qINSUL 125 kg/m3 [40]
VolINSUL 0.104 m3 Sunfire
Technical data Rated power 1.9 kW Sunfire
Rated voltage output 58.5 V Sunfire
Number of cells 90 ESC – Sunfire
Weight 80 kg Sunfire
Size 709  436  426 mm2 Sunfire
Max temperature 860 °C Sunfire
Fuel utilization 75% – Sunfire
Fuel composition Reformed NG or Biogas – Sunfire

Table 2
Evaporator section input parameters.

Evaporator Parameter Value Unit Ref.

Evaporator Length 0.260 m [36]
Volume 0.00093 m3 [36]
Density 8000 kg/m3 [36]
Mass 7.48 kg [36]
Heat capacity 0.502 kJ/kg K [36]
Heater evaporator Length 0.390 M [36]
Volume 0.00452 m3 [36]
Density 8000 kg/m3 [36]
Mass 36.13 kg [36]
Heat capacity 0.502 kJ/kg K [36]

Table 3
Air pre-heater section input parameters.

Heat exchanger Parameter Value Unit Ref.

Heat exchanger air preheater Max. flow rate 300 Nl/min Sunfire
Min. flow rate 130 Nl/min Sunfire
Exchange heat flow 4.4 kW Sunfire
Max inlet temp. 830 °C Sunfire
Max outlet temp. 650 °C Sunfire
Electrical air preheater Power 3.5 kW Sunfire
Max inlet temp. 650 °C Sunfire
Max outlet temp. 850 °C Sunfire

out X
5  
dT ref Uref _ ref yref ;out ref ;out
T out
ðC pref  M ref þ C p cat  M cat Þ  out ¼ Gout  i  hi ref i ¼ fCH4 ; CO; CO2 ; H2 O; H2 g
dt i¼1
   
¼ Uref þ Uref
 U WGS T out
 USMR T out þ UH2 ð3Þ ð5Þ
in out ref ref

where C pref ; Mref ; C p cat ; Mcat indicate the mass and heat capacity of UH2 ¼ UH2n  f  g ð6Þ
the reformer and the catalyst contained therein; UH2 is the power where
supplied by the electrical heater to the anode stream;
   
UWGS T out
ref and USMR T out ref represent the energy fluxes produced/ – T out out
ev a and T ref are the reformer and evaporator leaving tempera-

absorbed at reformer temperature; U and U ref

in
ref
are the energy
out
ture, in [K];
fluxes due to the mass flux inlet/outlet from the reformer; – f is a percentage parameter managed by SIMULINKÒ for the
power modulation;
X
5
ref ;ion  out  – g is the electrical efficiency of heater;
Uref _ ref yref ;in
in ¼ Gin   hi T ev a i ¼ fCH4 ; CO; CO2 ; H2 O; H2 g
i¼1
i
– UH2,n is the inlet power controlled by PID feedback (varying f
parameter).
ð4Þ
 G. D’Andrea et al. / Energy Conversion and Management 135 (2017) 21–34 27

Table 4
Reformer section input parameters.

Reformer Parameter Value Unit Ref.

Reformer External height 0.185 m [36]
Internal height 0.175 m
External volume 0.00118 m3
Internal volume 0.00088 m3
Density 7940 kg/m3
Mass 2.360 kg
Heat capacity 0.460 kJ/kg K [39,36]
Heater reformer Height 0.300 m [36]
Volume 0,00726 m3
Density 7940 kg/m3
Heat capacity 0.460 kJ/kg K [39,36]
Mass 57.61 kg
Catalyst Mass catalyst 0.700 kg [36]
Density of bed catalyst 970 kg/m3
Heat capacity - pellet 0.880 kJ/kg K
Heat capacity - gas 3.174 kJ/kg K
Heat capacity - fluid bed 1.912 kJ/kg K
Mass catalytic bed 0.853 kg

The development of the SOFC section follows these guidelines: X

5
an  
Uan _ an yan;out  hi T out
out ¼ Gout  i SOFC i ¼ fCH4 ; CO; CO2 ; H2 O; H2 g
i¼1
1. The inlet anode composition is known.
2. The air flow rate necessary for the stack thermal control is ð11Þ
imposed by a PID controller.
X
5
cat  
Ucat _ cat ycat;in  hi T out
in ¼ Gin  i air i ¼ fCH4 ; CO; CO2 ; H2 O; H2 g
Note that also an ‘‘internal reforming” of possible residual i¼1
methane are taken into account by the simulator. These calcula- ð12Þ
tions are the same used for the reformer section, implemented in
the LPM previously discussed. X cat  
5
Ucat _ cat ycat;out  hi T out
out ¼ Gout  i ¼ fCH4 ; CO; CO2 ; H2 O; H2 g
Knowing the molar fraction of inlet/outlet compounds, the i SOFC
Nernst potential (V0CV) calculation is done, considering the final i¼1
value as the average between two different Nernst potentials: ð13Þ

The first one is evaluated considering the SOFC inlet molar frac-
tion of oxygen, hydrogen, and water. 3. Model validation

The second one is evaluated with all the terms considered at
SOFC outlet conditions. The dynamic model is validated according to two different
approaches:
Finally, knowing the average stack temperature it is possible to
obtain the ASR value and calculate the output voltage of stack with
the first benchmark session is carried out against data provided
the following equations: by the SOFC manufacturer;
V cell ¼ V 0CV  ASRðT stack Þ  i ð7Þ
the second benchmark session is carried out with results col-
lected during the tests performed during the SOFCOM proof-
where of-concept operation.

– ASR, is the area specific resistance, in [X cm2]; Model validation is carried out by initially setting the model in
– i is the current density [A/cm2]. the same initial condition as the experimental tests.

The electric power produced is calculated as follows: 3.1. Model validation with manufacturer’s performance
W el ¼ V cell  i  Acell  ncell ð8Þ
The stack operating conditions at three different operating
2
where Acell is the surface of a single cell (127.8 cm ). points are given in Table 5.
The implemented thermal energy balance is reported below: The stack temperature is detected by six type-N thermocouples
out disposed along the package of the cells. Other four thermocouples
dT SOFC
ðC p;isol  M isol þ C p;MEA  M MEA þ C p;INTER  MINTER Þ  are installed to monitor the input/output anode and cathode tem-
dt
peratures. The main thermocouples used as reference indicate:
¼ Uan
in þ Uout þ Uin þ Uout  W el
an cat cat
ð9Þ

T2001 measure the inlet cathode temperature.
X
5  

T1001 measure the inlet anode temperature.
¼ G_ an
an
U an
in in  yan;in
i  hi T out
pipe i ¼ fCH4 ; CO; CO2 ; H2 O; H2 g
i¼1

T2105, T2107, T2111, T2106, T2108, T2112 are the six thermo-
couples located inside the core of the cell package. Among
ð10Þ
the different temperatures, T2111 (shown in Table 5) is the
28 G. D’Andrea et al. / Energy Conversion and Management 135 (2017) 21–34

Table 5
Reference test data from manufacturer.

I [A] I [A/cm2] V [V] H2 [Nl/min] N2 [Nl/min] Air [Nl/min] T2001 [°C] T1001 [°C] T2111 [°C] Pel [W]
24 0.18 62 20.1 30.1 210–235 650 750 855 1500
29 0.23 58 24.1 36.1 315–330 650 750 855 1680
34 0.27 55 28 42 400 650 750 862 1880

reference thermocouple chosen by the manufacturer for the 3.2. Model validation with experimental results from the proof-of-
performances datasheet and used as ‘stack temperature’ for the concept
validation.
The comparisons are made with the data obtained monitoring
The test was carried out imposing the air flow rates shown in
the SOFCOM prototype at the SMAT WWTP. In particular, three
Fig. 6.
polarization tests are considered:
Then, results have been compared with values from manufac-
turer’s technical datasheet (Table 5).

Polarization Test session#1 & Test session#2: at first, polarization
In Fig. 6 Tsofc is the SOFC stack temperature computed by the
in H2/N2 mix up to 24 A was implemented once completed the
dynamic model and compared with T2111. The maximum allow-
heat-up; then, once stabilized the working point, the shift from
able stack temperature was set to 10 °C lower compared to the
H2/N2 to the fuel reformate was gradually carried out, complet-
manufacturer specification, equal to 860 °C. The vertical black
ing the polarization until reaching the OCV conditions.
dashed lines highlight three main regions where a constant load

Polarization Test session#3: using a mixture of H2/N2 up to 24 A
is imposed (24 A, 29 A, and 34 A).
again.
It is possible to observe a very good correlation between the
obtained results and those provided by the manufacturer. In
Each polarization test was carried out by varying the stack cur-
particular:
rent step-wise from 0 to 24 A using current steps of 0.25 A/min.
Throughout the polarization test, the anode fuel composition and
– The air flow rate which is necessary to keep the fixed tempera-
flow rate was kept constant. The inlet fuel temperature was set
ture set-point is in the range of the manufacturer datasheet.
to 750 °C. The cathode flow rate was gradually adjusted (increased)
– The model tends to slightly underestimate (0.2–0.5 V less) the
as the current load increases to maintain a stack temperature
voltage at 24 and 34 A while showing an excellent correlation
below 860 °C.
at medium load.
– The temperature set point is kept constant by the PID controller
that regulated the amount of cathode air in the stack. 3.2.1. Test session #1
See Fig. 7.

Fig. 6. H2/N2 (40/60%) 24-29-34 A @ FU = 75%. Model results are represented by continuous lines. Reference data from the manufacturer are represented with dashed lines.
 G. D’Andrea et al. / Energy Conversion and Management 135 (2017) 21–34 29

Fig. 7. Test session #1.

Fig. 8. Test session #2.

3.2.2. Test session #2 the test session #1, small deviations from the experimental data
See Fig. 8. are detected in both test session #2 and test session #3. These dif-
ferences stand out only in the first section of the curve (at currents
of the order of 0.01–0.03 A/cm2, so very low currents, that are of
3.2.3. Test session #3 low interest in the real operation of the plant), with the average
See Fig. 9. relative errors below 1.21% anyway. The maximum relative error
The results of the presented analysis show good agreement was about 2.47% (voltage curve comparison), registered during
between experimental and simulated data. the test session #3 at 0.04 A/cm2.
Considering all the temperature charts, the results obtained
from the model fitting well TT302 values. Note that the tempera-
ture obtained by the model represents both the core and the outlet 4. Simulations of harsh off-design conditions
stack temperature (outlet gasses are assumed to be at the same
temperatures stack core). The validated model is useful for predictive purposes: as an
A good data correlation concerning the polarization curves has example, we analyze the behavior of the proof-of-concept under
also been obtained. While the curve overlap is nearly perfect for unconventional situations possibly related to an unexpected
30 G. D’Andrea et al. / Energy Conversion and Management 135 (2017) 21–34

Fig. 9. Test session #3.

Fig. 10. Direct internal reforming: by-pass layout.

malfunctioning of a given plant component. Hence, the following the electrochemical oxidation of hydrogen (and polarization
sections cover specific cases of practical interest and/or abnormal losses) is internally recycled to sustain reforming reactions, thus
working conditions caused by an unforeseen (partial) failure of a reducing the need for external cooling air (and thus reducing the
crucial BoP component. The analyzed off-dosing conditions are: air blower consumption). This has an effect on the regulation of
the cathode air flow in order to assure the correct thermal balance
1. DIR - Direct Internal Reforming operation. This condition can be inside the SOFC modules.
caused by a reduction in the catalytic activity of the reformer Three by-pass ratios of the reforming section (and thus DIR%)
catalyst: in this scenario, the reformer is not able to convert have been evaluated, starting from the zero by-pass condition
100% of methane, which is fed directly to the fuel cell. The same (0%) up to 60% of DIR (above this percentage, the DIR is not practi-
condition can also be a design choice in order to reduce the size cal under the present conditions, due to a sub-cooling of the anode
and the heat requirement of the external reformer. stream at the stack inlet). In stationary working conditions, the
2. Unexpected variations of the electrical load. This condition is operating temperature is stabilized at around 850 °C, varying the
expected especially during load flowing operation; the related cathode air flow rate, while the FU is kept constant at 75%.
risk is to generate stack overheating and thermal stresses on Fig. 11 shows the obtained simulation results, confirming the
the cells. trend discussed above. The air blower PID controller, thanks to
3. Instability or incorrect operation of the air blower. the progressive cooling of the stack due to an increasing DIR, pro-
gressively reduces the flow rate to maintain the working tempera-
4.1. DIR – direct internal reforming ture. It is evident that the transition from a 0% by-pass to a by-pass
of 60% will cause about 25 Nl/min less air.
The purpose of this simulation is to understand the behavior of The cathodic airflow reduction for cooling is important in indus-
the stack resulting from the DIR - Direct Internal Reforming oper- trial applications: in SOFC plants, the electric power required for
ation. DIR operation is obtained by feeding part of the biogas the air cooling (proportional to the airflow rate) is one of the most
directly to the stack anode without any reformation: in the model, energy intensive auxiliaries. It means that a proper management of
this is obtained through the introduction of a by-pass line to let this facility implies an economic saving.
some biogas entering the stack directly without passing into the
reformer (see Fig. 10). It is worth noting that a reformer with a par- 4.2. Load following: ramp rates and related thermal effects
tially deactivated catalyst would result in a similar operating con-
dition, i.e., with a higher fraction of unconverted CH4 (biogas) fed The goal of this trial is to simulate the load following capability
directly to the SOFC. of the plant, and especially to identify the maximum allowable
By directly feeding methane to the fuel cell, a thermal sink is change in the load that can be accepted in order to avoid any pos-
brought inside the stack due to the internal reforming reactions sible SOFC over-heating.
now taking place on the Ni-based anode electrodes. Several load conditions, simulated varying the fuel flow rate
At the stack temperature of around 830–850 °C, the reforming (H2/N2 mix – 40/60%) while guaranteeing an FU equal to 75%, are
of methane is fast directly on the anode electrode; due to the performed. A selection of some meaningful simulations is reported
highly endothermic DIR reaction, part of the heat generated by in Table 6 with the related starting conditions.
 G. D’Andrea et al. / Energy Conversion and Management 135 (2017) 21–34 31

Fig. 11. DIR simulation with CH4/CO2 (60/40%) - I = 24 A @FU = 75%.

Table 6
Load following simulations.

T SOFC [°C] T anode inlet [°C] T air inlet [°C] Air [Nl/min]
Set-Point 825 750 650 PID regulated
Current target [A] Time interval [min] Current ramp [A/min]
Simulation A 14 ? 34 125 0.16
Simulation B 14 ? 34 62.5 0.32
Simulation C 14 ? 34 0 Instantaneous

All simulations start with a constant load equal to 24 A. Follows
Sim B – the required speed load application leads to a slight
a progressive linear load reduction to 14 A in order to show the PID overheating of the SOFC, reaching the maximum acceptable
regulation of the air flow (remember that its function is to keep the stack temperature (860 °C).
stack temperature below 850 °C). After, Sim A, Sim B, and Sim C
Sim C – the most extreme case (sudden current jumping
differ in the ramp rate (A/min) to reach the final target of 34 A from 14 A to 34 A) should be strictly avoided as it leads
(a current increase of 20 A is then simulated). Results are shown an overrun of 15 °C above the maximum design stack
in Fig. 12. temperature.
Both in the first (constant load) and in the second (descending
ramp) part, the air PID controller manages to maintain the previ- Some horizontal lines for Sim B and C are noticeable in the air
ously established set point (850 °C). graph: they represent the upper saturation limit for the PID con-
On the contrary, different dynamics is involved in the subse- troller, which corresponds to the maximum airflow rate managed
quent load variation (14–34 A): by the air blower.
Please note that about three and a half hours are necessary to

Sim A – no over-heating problems occur and the current ramp bring the stack to its optimum working condition in the Sim C. This
applied is well handled by the PID controller. is, however, unacceptable as a hot spot for this time length can lead
32 G. D’Andrea et al. / Energy Conversion and Management 135 (2017) 21–34

Fig. 12. Load following simulations – temperature vs. airflow vs. current.

to a thermo-mechanical degradation of the stack components A 50 Nl/min reduction of the air flow rate results in a stack tem-
(especially cells and sealing layers). perature that exceeds the maximum temperature allowed by
Finally, according to the previous calculation, the maximum the manufacturer (860 °C). A further reduction of the air flow
applicable load ramp results to be equal to 0.30 A/min, which cor- rate to 175 Nml/min results to a stack temperature of 950°, well
responds approximately to 13 W/min in terms of electric power. above the limiting value provided by the manufacturer.

4.3. Malfunction of cathode air blower: reduced air flow at constant 5. Conclusion
load
This work presents findings from the SOFCOM project, an EU-
The purpose of this simulation is to verify how a malfunctioning funded applied research project devoted to demonstrating the
of the air blower during the plant operation would affect the stack technical feasibility, efficiency and environmental advantages of
behavior. CHP plants based on the SOFC technology and fed by different
The simulations are down in two different load conditions: types of locally produced biogenous primary fuels. This work
medium load (12 A) and nominal load (24 A). In both cases, the fuel focuses especially on the performance of a biogas-fed SOFC stack.
feed is H2/N2 40/60 vol.% and FU is equal to 75%. For both condi- The aim of the paper was to develop a dynamic model of the
tions, the airflow rate was progressively reduced, from 250 Nl/ fuel cell stack in order to predict its behavior under unconventional
min to 175 Nl/min. The simulations results are shown in Fig. 13. (or fault) operating conditions which might occur during the
We observe that: power plant lifetime. The research focus has been directed on the
design and simulation of a robust control system of the fuel cell

At medium-load operation (12 A), any further reduction of the stack, which is able to avoid the damaging of the stack (e.g.,
air flow rate does not lead to any harmful condition for the fuel overheating).
cell – the stack temperature rises by 10–15 °C maximum. As the Since the load-response of an SOFC system is strongly influ-
temperature increases, internal losses will decrease, leading to a enced by the auxiliaries included in the Balance of Plant – as it
slightly increased power output (about 100 W). was also confirmed by the real operation of the proof-of-concept

At high current load (24 A), a critical stack operating condition – a simulation model has been implemented to predict the thermal
might be reached depending on the available stack air flow rate. performances of the fuel cell stack that also includes the reformer
 G. D’Andrea et al. / Energy Conversion and Management 135 (2017) 21–34 33

Fig. 13. Malfunctioning of the cathode air blower (i.e., the available air flow rate reduces at constant load).

unit and the heat-exchangers. Realistic thermal data (e.g., thermal allowed one to avoid a stack damage; ramping up the current
capacity) for the stack components have been used for the dynamic above this threshold will lead to a too high thermal gradient
simulations by relying on information provided by the SOFC stack across the fuel cell. One of the expected consequences of a too
manufacturer. high thermal gradient is the cracking of the glass-ceramic seal-
The dynamic model has been validated using both the manufac- ing, which joins the fuel cell to the interconnect thus providing
turer’s available performance data as well as experimental data gas tightness among anode and cathode volumes. A shown by
gathered directly for the operation of the proof-of-concept in a real simulations provided in Fig. 12, an instantaneous load-
wastewater treatment plant. following of the fuel cell of the electric load has been shown
The validated model has been used to study the plant behavior as being not feasible since a temperature overshoot of the solid
under harsh operating conditions of the SOFC stack due to the possi- stack structure would occur.
ble malfunctioning of selected components (e.g., a deactivated refor-
A malfunctioning of the cathode air blower has also been simu-
mer, an ill-regulated air blower, etc.). The main results show that: lated. When working at high current loads, dangerous situations
might occur in the case of a sudden and unexpected reduction

The Direct Internal Reforming (DIR) of biogas fuel in the fuel cell in the cathode air flow rate. At partial-load operation (12 A), a
stack has been simulated up to 60% of direct CH4 conversion sudden air reduction by 20% would not result in any stack dam-
into the fuel cell anode. DIR operation can be seen as an addi- age since the stack temperature remains quite constant as well
tional means to control the stack thermal behavior: by feeding as the power output. At high-load operation (24 A), a sharp
more biogas directly to the fuel cell, endothermic reforming increase of the stack temperature would occur instead upon a
reactions are promoted which can sink the internal heat gener- sudden reduction in air flow rate. Simulations (Fig. 13) show a
ation of the stack thus limiting the risk of stack overheating. large overshoot of the stack temperature well above the safety
Also, by increasing the DIR ratio, a reduced air flow rate to the limit specified by the manufacturer.
stack is required. We calculated how, going from 0% to 60% of
DIR, the air flow rate is reduced by 14% (Fig. 11). A reduced cath- Our findings, which have been calibrated and validated on a
ode air flow rate would also entail a lower parasitic loss from commercial stack module, show how critical is the stack thermal
the air blower: by assuming an overall pressure drop of management for the stable and durable operation of a fuel cell
200 mbar in the overall SOFC system and an isentropic effi- plant. We have verified again that fast load following is problem-
ciency of 50%, operation at DIR = 60% would also save about atic for the real-life operation of an SOFC stack. In addition, the
14% of the electric power supplied to the air blower. availability of the cathode air flow rate is essential to avoid a stack

The cathode air flow rate is efficiently controlled by the PID reg- overheating. The immediate open-circuit condition should be
ulation; the PID controller can well manage the stack tempera- forced to the system in the case of a sudden reduction of the air
ture set point within the safety limits provided by the flow rate due to a malfunctioning of the cathode air blower. Finally,
manufacturer. However, careful attention should be paid to direct internal reforming of biogas into the SOFC stack represents
the applied current load. Especially, it is important to control an effective means to reduce the parasitic loss of the air blower.
(limit) the current ramp rate in the case of load-following oper- Simulation results provide guidelines for an improved design of
ation. A dangerous stack overheating might occur when a sud- the control system of the plant, highlighting the feasible operating
den increase of the electric load occurs. Results show that a region under safe conditions and the means to maximize the over-
current ramp-up rate above 0.30 A/min could be the maximum all system efficiency.
34 G. D’Andrea et al. / Energy Conversion and Management 135 (2017) 21–34

Acknowledgments [19] Huangfu Y, Gao F, Abbas-Turki A, Bouquain D, Miraoui A. Transient dynamic

and modeling parameter sensitivity analysis of 1D solid oxide fuel cell model.
Energy Convers Manage 2013;71:172–85.
The research leading to these results has received funding from [20] Achenbach E. Response of a solid oxide fuel cell to load change. J Power
the European Union’s Seventh Framework Program (FP7/2007- Sources 1995;57:105–9.
[21] Aguiar C, Adjiman N, Brandon J. Anode-supported intermediate temperature
2013) for the Fuel Cells and Hydrogen Joint Technology Initiative
direct internal reforming solid oxide fuel cell. J Power Sources
under grant agreement number 278798 ‘SOFCOM’. 2005;147:136–47.
[22] Barelli L, Bidini G, Ottaviano A. Solid oxide fuel cell modelling: electrochemical
performance and thermal management during load-following operation.
Energy 2016;115:107–19.
References
[23] Barelli L, Bidini G, Cinti G, Ottaviano A. SOFC regulation at constant
temperature: experimental test and data regression study. Energy Convers
[1] Van Herle J, Membrez Y, Bucheli O. Biogas as a fuel source for SOFC co- Manage 2016;117:289–96.
generators. J Power Sources 2004;127:300–12. [24] Wu X, Gao D. Fault tolerance control of SOFC systems based on nonlinear
[2] Van herle J, Maréchal F, Leuenberger S, Membrez Y, Bucheli O, Favrat D. Process model predictive control. Int J Hydrogen Energy 2016:1–21.
flow model of solid oxide fuel cell system supplied with sewage biogas. J [25] Fardadi M, McLarty DF, Jabbari F. Investigation of thermal control for different
Power Sources 2004;131(1–2):127–41. SOFC flow geometries. Appl Energy 2016;178:43–55.
[3] Venkatesh G, Abdi R. Economic e environmental analysis of handling biogas [26] Singhal SC, Kendall K. High-temperature solid oxide fuel cells: fundamentals,
from sewage sludge digesters in WWTPs (wastewater treatment plants) for design, and applications; 2003.
energy recovery: case study of Bekkelaget WWTP in Oslo (Norway). Energy [27] Grew KN, Chiu WKS. A review of modeling and simulation techniques across
2013;58:220–35. the length scales for the solid oxide fuel cell. J Power Sources 2012;199:1–13.
[4] De Arespacochaga N, Valderrama C, Peregrina C, Hornero A, Bouchy L, Cortina [28] Kupecki JMJ, Jewulski J. Multi-level mathematical modeling of solid oxide fuel
JL. On-site cogeneration with sewage biogas via high-temperature fuel cells: cells. Clean Energy 2012.
benchmarking against other options based on industrial-scale data. Fuel [29] Tanaka T, Inui Y, Urata A, Kanno T. Three dimensional analysis of planar solid
Process Technol 2015;138:654–62. oxide fuel cell stack considering radiation. Energy Convers Manage
[5] Achenbach E. Three-dimensional and time-dependent simulation of a planar 2007;48:1491–8.
solid oxide fuel cell stack. J Power Sources 1994;49(1–3):333–48. [30] SOFCOM Project, SOFC CCHP with Poly-fuel Operation and Maintenance.
[6] Hall RG, Colclaser DJ. Transient modeling and simulation of a tubular solid http://cordis.europa.eu/project/rcn/101098_en.html [Last accessed on
oxide fuel cell. IEEE Trans Energy Convers 1999;14(3):749–53. December 23, 2016].
[7] Ota T, Koyama M, Wen C, Yamada K, Takahashi H. Object-based modeling f [31] Lanzini a, Kreutz TG, Martelli E, Santarelli M. Energy and economic
SOFC system: dynamic behavior of micro tubular SOFC. J Power Sources performance of novel integrated gasifier fuel cell (IGFC) cycles with carbon
2013;118:430–9. capture. Int J Greenh Gas Control Jul. 2014;26:169–84.
[8] Petruzzi L, Cocchi S, Fineschi F. A global thermo-electrochemical model for [32] Curletti F, Gandiglio M, Lanzini A, Santarelli M, Maréchal F. Large size biogas-
SOFC system design and engineering. J Power Sources 2003;118:96–107. fed Solid Oxide Fuel Cell power plants with carbon dioxide management:
[9] Pfafferodt M, Heiderbrecht P, Stelter M, Sundmacher K. Model-based technical and economic optimization. J Power Sources Oct. 2015;294:669–90.
prediction of suitable operating range of a SOFC for an auxiliary power unit. [33] Gandiglio M, Lanzini A, Santarelli M, Leone P. Design and balance-of-plant of a
J Power Sources 2005;149:53–62. demonstration plant with a solid oxide fuel cell fed by biogas from waste-
[10] Aguiar P, Adjiman C, Brandon N. Anode-supported intermediate temperature water and exhaust carbon recycling for algae growth. ASME DC – J Fuel Cell Sci
direct internal reforming solid oxide fuel cell. J Power Sources Technol 2014;11(June):31003.
2004;138:120–36. [34] Larminie J, Dicks A. Fuel cell systems explained, vol. 93; 2001.
[11] Xi H, Sun J. Proceedings of ASME international mechanical engineering [35] Massardo FLAF. Internal reforming solid oxide fuel cell- gas turbine combined
congress and exposition; 2005. cycles (IRSOFC-GT): Part A—cell model and cycle thermodynamic analysis. J
[12] Bove R, Lunghi P, Sammes NM. SOFC mathematic model for systems Eng Gas Turbines Power 2000;122/2.
simulations–Part 2: definition of an analytical model. J Hydrogen Energy [36] Gianotti Pret M, Ferrero D, Lanzini A, Santarelli M. Thermal design, modeling
2005;30:189–200. and validation of a steam-reforming reactor for fuel cell applications. Chem
[13] Recknagle KP, Williford RE, Chick LA, Rector DR. Three-dimensional thermo- Eng Res Des Dec. 2015;104:503–12.
fluid electrochemical modeling of planar SOFC stacks. J Power Sources [37] Salogni A, Colonna P. Modeling of solid oxide fuel cells for dynamic
2003;113(1):109–14. simulations of integrated systems. Appl Therm Eng 2010;30:464–77.
[14] Arriagada J, Olausson P, Selimovic A. Artificial neural network simulator for [38] Cao H, Li X, Deng Z, Li J, Qin Y. Thermal management oriented steady state
SOFC performance prediction. J Power Sources 2002;112:54–60. analysis and optimization of a kW scale solid oxide fuel cell stand-alone
[15] Wu X-J, Zhu X-J, Cao G-Y, Tu H-Y. Modeling a SOFC stack based on GA-RBF system for maximum system efficiency. Int J Hydrogen Energy 2013;38
neural networks identification. J Power Sources 2007;167(1):145–50. (28):12404–17.
[16] Nehter P. Two-dimensional transient model of a cascaded micro-tubular solid [39] Vdm T. Crofer22H material data sheet no. 4050, no. 4050; 2008.
oxide fuel cell fed with methane. J Power Sources 2006;157(1):325–34. [40] Microtherm. FREEFLOW-free pouring granules of high temperature insulation
[17] Padulles J, Ault GW, McDonald JR. An integrated SOFC plant dynamic model for 1000 °C. http://www.microthermgroup.com/high/EXEN/assets_db/ITEMS
power systems simulation. J Power Sources 2000;86(1–2):495–500. KEYWORDS3/items/documents/MSDS_MicrothermFreeFlow_EN_REACH2011
[18] Hall DJ, Colclaser RG. Transient modeling and simulation of a tubular solid 0310_v3.pdf [Last accessed on December 23, 2016].
oxide fuel cell. IEEE Trans Energy Convers 1999;14(3):749–53.