i n t e r n a t i o n a l j o u r n a l o f h y d r o g e n e n e r g y x x x ( 2 0 1 6 ) 1 e1 4

Available online at www.sciencedirect.com

ScienceDirect

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

Mathematical modeling and dynamic Simulink

simulation of high-pressure PEM electrolyzer
system

Tevfik Yigit a,b, Omer Faruk Selamet a,b,*

a
Mechatronics Engineering Department, Nigde University, Nigde 51245, Turkey
b
Dr. T. Nejat Veziroglu Clean Energy Research Center, Nigde University, Nigde 51245, Turkey

article info abstract

Article history: PEM electrolysis based hydrogen generator system model is developed by using Simulink in
Received 16 March 2016 MATLAB. The model covers the following system components: PEM electrolyzer stack,
Received in revised form water pump, cooling fan, storage tank, water tank, power supply, control unit and sensors.
29 May 2016 In simulation, PEM electrolyzer stack model is the most complex component and consists
Accepted 4 June 2016 of four parts; anode and cathode modules, membrane and finally voltage calculations
Available online xxx module. Efficiency drops and voltage losses in the PEM electrolysis stack are estimated. The
overpotentials are investigated as anode and cathode activation, electronic and ionic
Keywords: ohmic resistances. In addition, the loss of the other components is considered. The model
PEM electrolyzer is tested against the dynamic changes and it responded with quick outputs. The cell and
Hydrogen generation stack behavior under different conditions (i.e. ranging of temperature and pressure) are
Simulink model examined. The loss of each system component at different current densities is added to the
Dynamic model simulation. The study shows that the loss of stack dominates the losses of other compo-
Mathematical model nents at higher current densities. The simulation counts that the stack and system model
System simulation can run with different cases and scenarios. It is run for the scenario that it consumes
constant power for a high pressure operation. Results show that current drawn by the
stack decreases as the voltage increases because of the increasing pressure.
The model is also compared with the experimental results and the model is found to be
consistent with the experimental data.
© 2016 Hydrogen Energy Publications LLC. Published by Elsevier Ltd. All rights reserved.

electrolysis as a hydrogen production technique is recently

Introduction drawing researcher's attention. Among other electrolysis
techniques, it has many advantages such as high working
On earth, hydrogen is the most abundant chemical element, current densities, production at high pressure, pure gas gen-
but not easily accessible to its monatomic form. The pure eration and compact design. PEM electrolysis is relatively new
hydrogen is needed to be produced with the least energy and promising technology. Since PEM electrolysis has many
consumption possible [1]. Being one of the most environ- aspects remain unstudied, it's being investigated by more re-
mentally friendly, efficient and independent from searchers by time, as expected [2,3].
hydrocarbon-based fuels, Proton Exchange Membrane (PEM)

* Corresponding author. Mechatronics Engineering Department, Nigde University, Nigde 51245, Turkey.
E-mail address: [email protected] (O.F. Selamet).
http://dx.doi.org/10.1016/j.ijhydene.2016.06.022
0360-3199/© 2016 Hydrogen Energy Publications LLC. Published by Elsevier Ltd. All rights reserved.

Please cite this article in press as: Yigit T, Selamet OF, Mathematical modeling and dynamic Simulink simulation of high-pressure PEM
electrolyzer system, International Journal of Hydrogen Energy (2016), http://dx.doi.org/10.1016/j.ijhydene.2016.06.022
2 i n t e r n a t i o n a l j o u r n a l o f h y d r o g e n e n e r g y x x x ( 2 0 1 6 ) 1 e1 4

There are numerous experimental studies investigating

the PEM electrolyzers in different aspects. However, PEM electrolyzer system and model
modeling a PEM electrolyzer is also necessary. Modeling a
cell or stack has uttermost importance in understanding the The simplified schematic view of a PEM electrolyzer system is
operational behavior of a PEM electrolyzer [4]. Many mathe- shown in Fig. 1. The system consists of a 100 cm2 PEM elec-
matical models/calculations exist for characterizing the cell trolyzer cell/stack, an AC-DC power supply, a cooling fan, a
component(s), using own code [5], employing statistical water circulating pump, a water vessel, a hydrogen storage
methods [6] available software [7,8]. Gorgun has proposed a tank, a controller, a display and sensors. Simulink model of
PEM electrolyzer stack model containing 3 cells at the at- the system is depicted in Fig. 2. The power supply is thought to
mospheric pressure using Simulink/MATLAB [9]. Awasthi be the source for the energy required for cracking the water in
et al. investigated the effects of changing the temperature the electrolysis process. Besides, all the other components
and pressure on the cell performance and polarizations [10]. such as controller, pump, fan and sensors which need electric
Lee et al. developed a PEM electrolyzer model to investigate power are connected to this power supply. The controller and
the effect of the temperature, the water flow rate and the the sensors have very low power consumption. The energy
flow field on the performance of a PEM electrolyzer and consumption of commercial compatible components has
compared the results with an experimental study [11]. Han been taken into consideration by assigning an average con-
et al. developed an electrochemical model which is able to stant value to the calculations. The controller collects the
compute the diffusion overpotential in a PEM electrolyzer values of the current, the temperature, the pressure and the
[12]. Abdin et al. also developed a PEM electrolyzer model water level in the vessel and decides whether everything is
that employs Simulink to represent the ohmic resistances in neat for starting the system. The cooling fan is responsible for
electrolyzers with an electrical circuit model [13]. They also removal of heat dissipation from the stack. The controller also
investigated the effect of temperature and pressure on PEM decides the fan speed according to the stack temperature. The
electrolyzer performance and compared their results with model includes a clock, thus, the transient response of the
the published experimental ones. system can also be collected.
Most of the modeling studies are mainly focused on the
physics or the behavior of a cell or a stack. The behavior of a
PEM electrolyzer depending on the operating conditions PEM electrolysis stack model
should be investigated when designing a system. Moreover,
building a system using the optimized parameters of a cell/ The Simulink model of the main component in the system,
stack is also quite important. Sometimes, the operating pa- the PEM electrolyzer stack, is given in Fig. 3. It consists of four
rameters of the cell/stack may subject to change according to modules and each module has its own model which can be
the other system components properties. Obviously, the run separately, but, all the modules are linked to each other.
properties of other components such as power supply, water These modules are the anode, the cathode, the membrane and
pump, water tank, cooling system, control unit, gas managing the voltage. The model is based on electrochemical, electrical
subsystem etc. should also be considered when deciding and flow calculations. It also can optionally work in atmo-
system components and the range of operating parameters. spheric or pressurized conditions.
Literature has various system modeling studies regarding fuel Before modeling the system, a flow diagram is formed to
cells [14e20]. However, a study dedicated to system modeling design mathematical calculations and interactions between
including a PEM electrolyzer is scarce and as far as author's each module in Simulink. In the model, the flow of water at
knowledge. Dale et al. [21] modeled a commercial PEM elec- the anode inlet and outlet, electro-osmotic drag and diffusion
trolyzer stack in a system containing balance of plant, but of water are calculated dynamically in real time. Hydrogen
they didn't include the system components to the model. Such ions evolved from the anode catalyst layer pass to the cath-
a system should continuously decide on the operating con- ode, combine and produce H2 gas at the catalyst layer. The
ditions according to environmental conditions and previous hydrogen production rate depending on the current density is
operating conditions while it is running. calculated in the model.
Furthermore, designing and customizing another system The behavior of the PEM electrolyzer considering different
with higher or lower hydrogen flow rate, the stack and the conditions is obtained for different cases. A production sce-
other components behavior in the system may change. nario for higher pressure or temperature are planned and
Instead of performing all the experiments over and over for simulated on the model.
each case, a Simulink model can help to estimate the losses
and required parameters such as the heat dissipation, the Anode module
water flow rate and the working current density for failsafe
operation. The water cracking and the oxygen evolution reaction take
Even though well-established simulations [9,10,22,13,12,23] place in the anode compartment. The half-cell reaction is
are available in the literature, there is a limited number of given as;
system modeling related to PEM electrolysis in the literature.
2H2 O/O2 þ 4Hþ þ 4e (1)
Therefore, a basic PEM electrolyzer system model is
introduced using Simulink/MATLAB in this study. The system Oxygen evolves as gas, the hydrogen ions travel to the
components modeled by using Simulink/MATLAB is taken cathode through the membrane, while the electrons are
after a laboratory scale in house made PEM electrolyzer. driven by the power supply to the outer circuit. The total

Please cite this article in press as: Yigit T, Selamet OF, Mathematical modeling and dynamic Simulink simulation of high-pressure PEM
electrolyzer system, International Journal of Hydrogen Energy (2016), http://dx.doi.org/10.1016/j.ijhydene.2016.06.022
 i n t e r n a t i o n a l j o u r n a l o f h y d r o g e n e n e r g y x x x ( 2 0 1 6 ) 1 e1 4 3

Fig. 1 e Simplified schematic view of the PEM electrolyzer system with which the dynamic model is developed.

Fig. 2 e Dynamic model of the PEM electrolyzer system. It consists of the PEM electrolyzer stack and balance of plant
components.

Please cite this article in press as: Yigit T, Selamet OF, Mathematical modeling and dynamic Simulink simulation of high-pressure PEM
electrolyzer system, International Journal of Hydrogen Energy (2016), http://dx.doi.org/10.1016/j.ijhydene.2016.06.022
4 i n t e r n a t i o n a l j o u r n a l o f h y d r o g e n e n e r g y x x x ( 2 0 1 6 ) 1 e1 4

Fig. 3 e Simulink model of a PEM electrolyzer stack. It consists of four main modules; the anode, the cathode, the membrane
and the voltage calculations. Each module shares information with each other.

oxygen flow rate is equal to the generated oxygen subtracted

by the output oxygen flow rate, the oxygen inlet is assumed to

NO2 $Ru $ Tel
be equal to zero. Water is supplied continuously to the anode PO2 ¼
yan
compartment, meanwhile a small amount of water dragged to
the cathode by the hydrogen ions (electro-osmotic drag, N_ H2 Oeod

NH2 Oan $Ru $Tel
and diffusion, N_ H O ). Water transport mechanism is PH2 Oan ¼ (3)
2 diff yan
described in detail later in membrane section.
The oxygen generation rate and water consumption rate Pan ¼ PO2 þ PH2 Oan
can be written as;
The molar rates of the species can be calculated from the
N_ O2 ¼ N_ O2ai  N_ O2ao þ N_ O2g ratio of partial pressures to the anode pressure;

PO2
N_ H2 Oan ¼ N_ H2 Oai  N_ H2 Oao  N_ H2 Oeod  N_ H2 Odiff (2) yO2 ¼ (4)
Pan

The total pressure of the anode is the sum of the partial The amount of oxygen generated in the anode can be
pressures of the oxygen and the water, which are derived from calculated in real time from Faraday's equation;
the ideal gas law;

Please cite this article in press as: Yigit T, Selamet OF, Mathematical modeling and dynamic Simulink simulation of high-pressure PEM
electrolyzer system, International Journal of Hydrogen Energy (2016), http://dx.doi.org/10.1016/j.ijhydene.2016.06.022
 i n t e r n a t i o n a l j o u r n a l o f h y d r o g e n e n e r g y x x x ( 2 0 1 6 ) 1 e1 4 5

where, Vc is the cathode volume in cubic meters. The molar

n$I rate of the hydrogen can be calculated as;
N_ O2g ¼ (5)
4$F
PH2
The Simulink model of anode section of PEM is shown in yH2 ¼ (9)
Pca
Fig. 4.
The amount of the hydrogen generated by the chemical
reaction can be calculated in real time from Faraday's
Cathode module
equation;
The cathode is the compartment where the hydrogen gas n$I
evolves. The reaction takes place in cathode is given as; N_ H2g ¼ (10)
2$F

4 Hþ þ 4 e/H2 (6) The saturation pressure is required to be defined for the

relative humidity calculations in the membrane module.
The hydrogen ions combine with each other and electrons, However, it is depended only on the temperature. Saturation
which are coming from the power supply to generate pressure (Psat) is defined as [24].
hydrogen gas. The produced hydrogen gas accumulates in the
cathode, if the working pressure is set to higher than the at- Psat ðPaÞ ¼ 2846:4 þ 411:24 T  10:554 T2
mospheric pressure. Hydrogen is taken out to the storage tank þ 0:16636 T3 ðT is in  CÞ (11)
when desired pressure is reached. The amount of the
hydrogen flow rate is equal to the generated hydrogen sub- The Simulink model of the cathode section of the PEM
tracted by the amount of hydrogen taken out from the outlet. electrolyzer is shown in Fig. 5.
The hydrogen and the water in the cathode can be calculated
as;
Membrane module
N_ H2 ¼ N_ H2ci  N_ H2co þ N_ H2g
The membrane is considered where the hydrogen ions and
the water can be transported in the model. The Simulink
N_ H2 Oca ¼ N_ H2 Oci  N_ H2 Oco  N_ H2 Oeod þ N_ H2 Odiff (7)
model of the PEM electrolyzer membrane is shown in Fig. 6.
where N_ H2ci is flow rate of hydrogen transferred from anode to Two essential events related to water are available in the
the cathode, N_ H2co is hydrogen flow rate at cathode outlet. membrane module. These are the diffusion and the electro-
Diffusion (N_ H2 Odiff ) and electro-osmotic drag (N_ H2 Oeod ) of water is osmotic drag of the water. LaConti and Swette reported that
described in details in the membrane module section. The 3.5e4 molecules of water are transported with each proton to
partial pressures of the hydrogen and the water can be the cathode, while the hydrogen ions traveling through the
calculated from the ideal gas law and the cathode pressure is membrane [25]. Later Li et al. [26] offered a model for the water
equal to the sum of the partial pressures of the hydrogen and transportation through the membrane;
the water;

nd $i$MH2 O $A

 N_ H2 Oeod ¼ (12)
NH2 $Ru $Tel F
PH2 ¼
yca where, i is the current density, MH2 O is the molecular weight of
the water, A is the active area, nd is the electro-osmotic drag


NH2 Oca $Ru $Tel coefficient which can be calculated by;
PH2 Oca ¼ (8)
yca
nd ¼ 0:016 TðKÞ  2:89556 (13)
Pca ¼ PH2 Oca þ PH2

Fig. 4 e Simulink model for the anode module.

Please cite this article in press as: Yigit T, Selamet OF, Mathematical modeling and dynamic Simulink simulation of high-pressure PEM
electrolyzer system, International Journal of Hydrogen Energy (2016), http://dx.doi.org/10.1016/j.ijhydene.2016.06.022
6 i n t e r n a t i o n a l j o u r n a l o f h y d r o g e n e n e r g y x x x ( 2 0 1 6 ) 1 e1 4

Fig. 5 e Simulink model the cathode module.

Fig. 6 e Simulink model for the membrane module.

Another issue with the membrane is the water uptake, where, Cwc and Cwa are the water concentration of the cathode
which acts a vital role for the ion mobility. The water uptake of and the anode respectively, tme is the membrane thickness
the membrane (l) is defined by Li et al. as [26]; and equals to N117 thickness (0.215 mm) in this study. The
diffusion coefficient of water can be calculated as;
l ¼ 0:043 þ 17:81a  39:85a2 þ 36a3
  
1 1
Dw ¼ Dl exp 2416  (17)
lm ¼ ðlca þ lan Þ=2 (14) 303 T

where l is calculated for both anode and cathode. lm is the

Dl ¼ 1010 ; lm < 2;
mean of lan and lca and used to calculate the diffusion coef- Dl ¼ 1010 ½1 þ 2 ðlm  2Þ ; 2  lm  3;
(18)
ficient of the water. a is the water vapor activity and calculated Dl ¼ 1010 ½3  1:67ðlm  3Þ ; 3  lm  4:5;
as; Dl ¼ 1:25  1010 ; lm  4:5;

P The water concentration (Cw) is defined by as [27];

a¼ (15)
Psat rme
Cw ¼ $lm (19)
Assuming that the gas mixture behaves as an ideal gas, the EWme
water vapor activity can be replaced by the relative humidity. where, rme is the density of the membrane and EWme is the
Another significant mechanism in the membrane is the equivalent weight of the membrane. Thus, the total amount of
transport of the water by diffusion. N_ H2 Odiff is the diffusion of the water transported from the anode to the cathode is the
water in moles can be expressed as; summation of transport by electro-osmotic drag and
  diffusion;
Cwc  Cwa
N_ H2 Odiff ¼ Dw $MH2 O $A (16)
tme
N_ H2 Ototal ¼ N_ H2 Odiff þ N_ H2 Oeod (20)

Please cite this article in press as: Yigit T, Selamet OF, Mathematical modeling and dynamic Simulink simulation of high-pressure PEM
electrolyzer system, International Journal of Hydrogen Energy (2016), http://dx.doi.org/10.1016/j.ijhydene.2016.06.022
 i n t e r n a t i o n a l j o u r n a l o f h y d r o g e n e n e r g y x x x ( 2 0 1 6 ) 1 e1 4 7

Voltage calculations module

 
Ru $Tan i
Vact;an ¼ $arcsinh (27)
The operating cell voltage is the sum of thermo-neutral fan F 2i0;an
voltage, activation overpotentials, ohmic overpotentials and
mass transfer overpotentials;  
Ru $Tca i
Vact;ca ¼ $arcsinh
fca F 2i0;ca
Vcell ¼ Vth þ Vact; an þ Vact; ca þ Vohm;e þ Vohm;i þ Vmt; an þ Vmt; ca
(21) where, f is charge transfer coefficient and taken as fca ¼ 0.5
and fan ¼ 2, i is the current density and i0 is the exchange
The mass transfer losses are not significant for low and
current density and accepted as i0,an ¼ 2  107 and
moderate current densities if the flow field is appropriate for
i0,ca ¼ 2  103 in this study [21].
gas removal. Thus, the mass transfer overpotentials can be
The ohmic overpotential can be divided into two categories
neglected for up to 3 A/cm2, which is kept out of the scoop of
due to the electronic and the ionic resistances. Ionic resis-
this study.
tance can be calculated by;
The enthalpy change (DH) is described as the energy
required for building/breaking up a bond for a chemical re- Vohm ¼ i$Rohm; i
action, while the entropy change (DS) is a measure of the ir-
reversibilities for a thermodynamic system. The third term is tme
Rohm; i ¼ (28)
the change in Gibbs free energy of reaction (DG). Thus, the sme
enthalpy can be described as;
where, sme is the membrane conductivity which can be ob-
DH ¼ DG þ T$DS (22) tained from Ref. [31];
  
DH and DS can be calculated according to enthalpies, en- 1 1
sme ¼ ð0:00514 $lm  0:00326Þ$exp 1268  (29)
tropies and stoichiometric factors of the reactants and the 303 Tel
products of the reaction. At the standard state (P ¼ 1 atm, The ionic over potential is dependent of the membrane
T ¼ 298.15 K), the Gibbs free energy of reaction DG is 236.483 kj/ water content, the temperature and the operating current.
mol [28]. For this amount of energy the potential difference On the other hand the electronic resistance Rohm, e does not
between the electrodes is called the reversible cell voltage change significantly depending on the operating conditions. A
(Vrev). However, minimum voltage required for the electrolysis previous study results are used to calculate the contact and
reaction is higher than Vrev and is called the thermoneutral the electronic resistances for ohmic losses. The contact re-
voltage (Vth) because of the irreversibilities. Thermoneutral sistances can be included in the electronic part [32]. Thus, a
and reversible voltage can be calculated as; constant value can be taken for the contact and the bulk re-
sistances. Therefore, the ohmic overpotential can be obtained
DG by;
Vrev ¼ ¼ 1:229 V
n$F


Vohm ¼ i$ Rohm;i þ Rohm;e (30)
DH
Vth ¼ ¼ 1:481 V (23) The Simulink model of the voltage calculations for a PEM
n$F
electrolyzer is designed as in Fig. 7.
where F is the Faraday constant Hence, efficiency of an elec-
trolyzer (εHHV) according to higher heating value can be
calculated by,
Balance of plant
Vth
εHHV ¼ (24)
Vcell
The most complicated and dynamic component is the stack
The thermo-neutral potential, Eth, can be obtained by using itself, in a PEM electrolyzer based hydrogen generator. The
Nernst equation for the calculating Vcell [29]; response to the instant changes of operating conditions of a
PEM electrolyzer stack can be drastic. However, all the com-
" 1=2
!# ponents should be considered to optimize the best system
RU $Tel PH2 $PO2
Vth ¼ V0 þ $ ln (25) performance. Thus, the model considered in this study con-
2F PH2 O
tains not only the electrolyzer stack but also the water pump,
where, E0 is the standard potential which is calculated from the cooling fan, the power supply, the storage tank, the water
Gibbs free energy; tank and the electronic control unit [33]. The system works in
a cycle and all the parts work together, shares information
DG
V0 ¼ (26) with each other. The losses from each component and the
2F
overall losses can be obtained separately.
PH2 O , PH2 , PO2 are the partial pressures of the water, the
hydrogen and the oxygen, Tel is the absolute temperature, R is
Water pump and tank
the universal gas constant.
The activation overpotentials of the anode and the cathode
The water pump model is a type of a differential model of a DC
can be calculated as [6], [30];
motor. Motor rotation speed which is directly related to

Please cite this article in press as: Yigit T, Selamet OF, Mathematical modeling and dynamic Simulink simulation of high-pressure PEM
electrolyzer system, International Journal of Hydrogen Energy (2016), http://dx.doi.org/10.1016/j.ijhydene.2016.06.022
8 i n t e r n a t i o n a l j o u r n a l o f h y d r o g e n e n e r g y x x x ( 2 0 1 6 ) 1 e1 4

Fig. 7 e Simulink model for the voltage calculations.

intended water flow rate. An equation is obtained to calculate calculate angular velocity values, according to voltage
the motor speed versus intended flow rate and it is embedded values. Voltage value is calculated by PID (proportional in-
to DC motor model. DC motor model can be described as [34]; tegral derivative) controller according to the difference be-
tween reference and present velocity values. Reference
dia ðtÞ 1
¼ ½Va ðtÞ  Ra ia ðtÞ  Vb ðtÞ velocity values calculated by Equation (32). Water flow rate of
dt La
the water inlet (N_ H2 Oai ) is a constant value which is deter-
mined at the beginning. Electrical power consumption of the
tm ðtÞ ¼ Ki ia ðtÞ
DC motor is calculated by multiplying the voltage values
calculated by PID and current values calculated by DC motor
dqm ðtÞ
Vb ðtÞ ¼ Kb ¼ Kb um ðtÞ (31) model.
dt
The water tank gives information about the change in
  water level instantly. The instantaneous level of the water
d2 qm ðtÞ 1 dqm ðtÞ
¼ t m ðtÞ  tL ðtÞ  Bm inside the tank is calculated from subtraction of the inlet flow
dt2 Jm dt
and the outlet flow;
where, i current, R resistance, V voltage, t torque, K torque  
constant, B viscous-friction coefficient, q rotor displacement, J
 MH2 O
LevelH2 O ¼ Levelinitial þ N_ H2 Oi  N_ H2 Oo $ (33)
motor inertia and um rotor angular velocity. The DC motor A$rH2 O
model can be builded according to Equation (31). Angular ve-
locity of the motor can be calculated as; Cooling fan

um ðtÞ ¼ N_ H2 Oai $kpump (32)

All the components in the electrolysis system, mainly the
where N_ H2 O is the water flow rate, um is the revolution per stack, have some kind of losses, which emerge as excessive
minute of motor and kpump is constant calculated from data- heat, and it should be removed from the system. Because, the
sheets of commercial pumps. membrane is sensitive to the temperature, it loses its stability
Inductance, resistance and other parameters of dc motor at high temperatures. On the other hand, the electrolyzer
was obtained from the datasheets. Simulink model is built to performance is directly proportional to the temperature [1].

Please cite this article in press as: Yigit T, Selamet OF, Mathematical modeling and dynamic Simulink simulation of high-pressure PEM
electrolyzer system, International Journal of Hydrogen Energy (2016), http://dx.doi.org/10.1016/j.ijhydene.2016.06.022
 i n t e r n a t i o n a l j o u r n a l o f h y d r o g e n e n e r g y x x x ( 2 0 1 6 ) 1 e1 4 9

Thus, the temperature of the stack should be maintained at a

desired temperature range.
The model of the cooling fan collects all the losses from
other modules and calculates the amount of excessive heat in
the system. The controller of the fan model tries to balance
intended temperature with the present temperature and cal-
culates a speed for the fan motor. The same DC motor model
which is used for the water pump is also employed for the fan
motor. Thus, the fan motor model calculates the power con-
sumption and amount of heat removed, processing the speed
of the motor.

Power supply, controller and sensors

All the power supplies have some losses. The amount of loss
varies with the operating current and voltage values. Gener-
Fig. 8 e The effect of current density on the overpotentials
ally commercial power supply companies publish datasheets
and the performance of the PEM electrolyzer stack at
presenting the efficiency values at different voltages and
atmospheric pressure and 300 K.
corresponding operating currents. An efficiency equation is
fitted from a commercial power supply and the loss is calcu-
lated while system running according to the equation. high, the cell potential keeps below 1.8 V, which is equal to,
Each Simulink module contains own controller in the higher than 80% efficiency (Fig. 8).
simulation even though, all the control processes are gener- The performance of the stack strongly depends on the
ally executed from one center in reality. A constant value operating conditions such as the temperature and the pres-
added to the model for the losses by the controller and the sure. The pressure also affects the thermo-neutral voltage.
sensors such as the pressure, the temperature and the The model executed for three different pressures; atmo-
current. spheric, 20 and 350 bars. Fig. 9 shows that the electrolysis
starts above the 1.48 V. Since the other losses are almost the
Hydrogen storage tank same, the graph paths undergo similar course as the operating
current density raises.
The generated hydrogen is stored in a high pressure tank. The Another parameter affecting on the performance is the
hydrogen tank model collects the generated hydrogen and operating temperature of the cell/stack. Fig. 10 shows the
calculates the amount of stored hydrogen and the internal performance curves depending on the temperature. As the
pressure. Initially, the tank is assumed to be empty. The temperature increases, the energy needed for cracking the
generated hydrogen at desired pressure fills to the storage water decreases. The model is run up to 360 K, since the
tank. The flow rate of the hydrogen is calculated at operating literature presents experimental results generally up to 90  C.
pressure [35]. The pressure in the hydrogen tank is calculated Fig. 11 shows model results for the efficiency curves at
as; different temperatures from 280 K to 360 K. The efficiency
becomes quite close to % 100 as the temperature raises to-
NH2 $Ru $Ttube
Pstorage ¼ z$ þ Pinitial (34) wards 360 K, since the efficiency of a PEM electrolyzer calcu-
ytube
lations generally are based on thermoneutral voltage, 1.48 V.
where z is compressibility factor and it is approximately 1 for However, some of the losses especially the activation and the
hydrogen. ohmic over potentials decreases with increasing temperature
due to higher reaction kinetics and enhanced ionic
conductivity.
Results

The performance of the PEM electrolysis stack is generally the

most important contribution to the system overall efficiency.
Therefore, the losses from the stack should be investigated
and presented in the first place. The cell potential, ohmic and
activation overpotentials is shown in Fig. 8. The electronic
part of the ohmic losses and the cathodic activation losses are
relatively low in all current densities. However, the influence
of anodic activation overpotential is significant even at very
low current densities. The ohmic loss due to ionic resistance
increases linearly with the operating current. The dominating
part of the losses beyond the 1.5 A/cm2 becomes ionic over-
potential, since the ionic resistance is multiplied by operating Fig. 9 e The effect of the pressure on the performance of the
current. Even though, the operating current density is quite PEM electrolyzer at 300 K.

Please cite this article in press as: Yigit T, Selamet OF, Mathematical modeling and dynamic Simulink simulation of high-pressure PEM
electrolyzer system, International Journal of Hydrogen Energy (2016), http://dx.doi.org/10.1016/j.ijhydene.2016.06.022
10 i n t e r n a t i o n a l j o u r n a l o f h y d r o g e n e n e r g y x x x ( 2 0 1 6 ) 1 e1 4

performance while an increase in the pressure reduces the

cell performance. The cell voltage decreases from 2.06 V to
1.74 V when the operating temperature is increased from
273 K to 373 K at atmospheric pressure. On the other hand, the
cell voltage increases only 0.06 V (1.96 Ve2.02 V) when the
operating pressure is increased from atmospheric pressure to
200 bars at 298 K.
The effect of operating current density on the losses on the
stack is given in Fig. 13. The percentage of the losses are
calculated, so that the thermo-neutral voltage and the
contribution of each loss such as the activation and the ohmic
Fig. 10 e The effect of temperature on the performance the overpotentials are calculated depending on the current den-
PEM electrolyzer stack at atmospheric pressure. sity. At the relatively low current density (0.1 A/cm2), the loss
by the stack is similar to the power supply and pump. As the
current density increases the power loss rate by the stack
becomes much higher than other components. Even though,
the power consumption by the pump and the controller
remain almost the same, the contribution to the loss per-
centage becomes smaller, since the increase in the power loss
by the stack and power supply grows larger. The result of the
model related to the power loss from the stack follows similar
course to Santarelli and Medina's study [36]. At 0.3 A/cm2
current density, approximately 8.08 Watt power is lost while
86 Watt dissipates as heat at 1 A/cm2.
The model is acquired for different scenarios on hydrogen
production. Fig. 14 shows a constant power consumption
Fig. 11 e The effect of current density on the efficiency of
scenario for the stack. The stack assumed to have 5 cells with
the PEM electrolyzer at different temperatures at
100 cm2 active area each at 300 K. Since the power given to the
atmospheric pressure.
stack is constant, the stack voltage increases as the pressure
increases and the current drawn by the stack eventually de-
The effect of the temperature and the pressure on the cell creases. The stack voltage is below 8.8 V at the beginning, and
performance is presented at the same time in Fig. 12. An in- it increases to 9.05 V approximately (Fig. 14a). The current
crease in the operating temperature enhances the cell starts above 82 A and drops to 80 A approximately. According

Fig. 12 e The effect of the operating temperature and the pressure on the cell voltage. The change in temperature is more
dominant than the change in pressure on the cell voltage.

Please cite this article in press as: Yigit T, Selamet OF, Mathematical modeling and dynamic Simulink simulation of high-pressure PEM
electrolyzer system, International Journal of Hydrogen Energy (2016), http://dx.doi.org/10.1016/j.ijhydene.2016.06.022
 i n t e r n a t i o n a l j o u r n a l o f h y d r o g e n e n e r g y x x x ( 2 0 1 6 ) 1 e1 4 11

Fig. 13 e The loss percentages of the system components. The loss from the stack is much higher at high current densities.

Fig. 14 e The outputs of the model for a fixed power consumption scenario for pressurizing the generated gas to 350 bars.
The power consumed by the stack is limited to 720 Watts.

Fig. 15 e The response of the voltage and the hydrogen production rate of the PEM electrolyzer stack model to the dynamic
changes in operating current.

Please cite this article in press as: Yigit T, Selamet OF, Mathematical modeling and dynamic Simulink simulation of high-pressure PEM
electrolyzer system, International Journal of Hydrogen Energy (2016), http://dx.doi.org/10.1016/j.ijhydene.2016.06.022
12 i n t e r n a t i o n a l j o u r n a l o f h y d r o g e n e n e r g y x x x ( 2 0 1 6 ) 1 e1 4

to scenario, the stack starts at ambient pressure and the In this study, mathematical models of a PEM electrolyzer
pressure of the cathode compartment increases as it produces cell/stack and other required system components are devel-
hydrogen. The hydrogen is not allowed to pass to the high oped for a hydrogen generation system. The current versus
pressure tank until the internal pressure of the stack reaches the voltage graphs, the efficiency curves and the losses are
350 bar (Fig. 14b). The hydrogen flow rate of the stack outlet is plotted. A scenario is planned for fixed power consumption
null for 10 h roughly. Once it reaches to 350 bars the hydrogen and the stack performance investigated at high pressure
is allowed to pass to the tank with a flow rate of 8.7 ml/min at condition up to 350 bar. The system model works at steady
high pressure. If the storage tank is assumed to have 50-L state condition, once it reaches 350 bar pressure. As the stack
volume, it will take around 4 days to fill up with this flow rate. voltage increases due to pressurizing process, the current
The dynamic response of the model is presented in Fig. 15. decrease to 80 A and it produces 8.7-mL hydrogen per minute.
When the current drawn by the stack is changed the gener- The Simulink model is compared with experimental data
ated hydrogen flow rate is also changing accordingly. Simi- for a single cell performance curve and the model found to be
larly, when the current alters, the operating cell voltage also consistent with the experimental results.
changes consistently. The stack model can be improved by embedding more
Fig. 16 shows comparison of a single cell simulation results physics and mathematical expressions as the literature de-
with experiment's which has already performed earlier in velops on the field such as mass transfer inside the cell.
another study [3]. Both experimental and simulation results
are obtained at 60  C and atmospheric pressure. The simula- Nomenclature
tion results are found to be very well consistent to experi-
mental data. A active area
a water vapor activity
B viscous-friction coefficient
EW equivalent weight
F Faraday constant
Conclusion I current
i current density
A PEM electrolyzer based hydrogen generator system model i0 exchange current density
was developed using Simulink to estimate the stack and sys- J rotor inertia
tem behavior and losses at different operating conditions. The Ki torque constant
system has a stack having 5 cells with 100 cm2 active area and L inductance
balance of plant components. The model can be modified to M molecular weight
simulate similar systems with higher or lower hydrogen n number of cells
generation capacity. Therefore, it can be possible to estimate nd electro-osmotic drag coefficient
the power consumption of each component and have decision N_ flow rate
on the operating parameters. P pressure
The system losses are calculated by the model. The stack R resistance
losses are found relatively small at the low current densities, Ru universal gas constant
as the current density increases, the stack losses are found to T temperature
be quite large compared to the others' due to increasing t thickness
overpotentials especially the ohmic losses by the ionic re- y volume
sistivity. Therefore, the efficiency of the system becomes V voltage
significantly small at high working current density above the 1 y molar rate
A/cm2. On the other hand, low working current density below z compressibility factor
0.7 A/cm2 is not reasonable, since hydrogen production rate
per cell would be relatively low. Greeks
r density
s conductivity
f charge transfer coefficient
l water uptake
t torque
q rotor displacement
u rotor angular velocity

Subscripts and superscripts

a armature
act activation
ai anode in
an anode
Fig. 16 e The voltage output of the PEM electrolyzer ao anode out
Simulink model is compared to experimental data and the ca cathode
results are found the consistent to each other. ci cathode in

Please cite this article in press as: Yigit T, Selamet OF, Mathematical modeling and dynamic Simulink simulation of high-pressure PEM
electrolyzer system, International Journal of Hydrogen Energy (2016), http://dx.doi.org/10.1016/j.ijhydene.2016.06.022
 i n t e r n a t i o n a l j o u r n a l o f h y d r o g e n e n e r g y x x x ( 2 0 1 6 ) 1 e1 4 13

co cathode out [13] Abdin Z, Webb CJ, Gray EM. Modelling and simulation of a
diff Diffusion proton exchange membrane (PEM) electrolyser cell. Int J
el electrolyzer Hydrogen Energy Oct. 2015;40(39):13243e57.
[14] Haraldsson K, Wipke K. Evaluating PEM fuel cell system
eod electro-osmotic drag
models. J Power Sources Feb. 2004;126(1e2):88e97.
g generated [15] Pathapati PR, Xue X, Tang J. A new dynamic model for
H2 hydrogen predicting transient phenomena in a PEM fuel cell system.
H2O water Renew Energy Jan. 2005;30(1):1e22.
L load [16] del Real AJ, Arce A, Bordons C. Development and
m motor experimental validation of a PEM fuel cell dynamic model. J
me membrane Power Sources Nov. 2007;173(1):310e24.
[17] Wang C, Nehrir MH, Shaw SR. Dynamic models and model
mt mass transfer
validation for PEM fuel cells using electrical circuits. Energy
O2 oxygen Convers IEEE Trans On 2005;20(2):442e51.
ohm Ohmic [18] Arsov GL. Improved parametric PSpice model of a PEM fuel
sat saturation cell. In: Optimization of electrical and electronic equipment,
2008. OPTIM 2008. 11th International Conference on; 2008.
p. 203e8.
[19] Spiegel C. PEM fuel cell modeling and simulation using
Matlab. Amsterdam: Boston: Academic Press/Elsevier;
references
2008.
[20] Olesen AC, Rømer C, Kær SK. A numerical study of the gas-
liquid, two-phase flow maldistribution in the anode of a high
€ Becerikli F, Mat MD, Kaplan Y. Development and
[1] Selamet OF, pressure PEM water electrolysis cell. Int J Hydrogen Energy
testing of a highly efficient proton exchange membrane Jan. 2016;41(1):52e68.
(PEM) electrolyzer stack. Int J Hydrogen Energy Aug. [21] Dale NV, Mann MD, Salehfar H. Semiempirical model based
2011;36(17):11480e7. on thermodynamic principles for determining 6kW proton
[2] Selamet OF, Pasaogullari U, Spernjak D, Hussey DS, exchange membrane electrolyzer stack characteristics. J
Jacobson DL, Mat MD. Two-phase flow in a proton exchange Power Sources Dec. 2008;185(2):1348e53.
membrane electrolyzer visualized in situ by simultaneous [22] Shen M, Bennett N, Ding Y, Scott K. A concise model for
neutron radiography and optical imaging. Int J Hydrogen evaluating water electrolysis. Int J Hydrogen Energy Nov.
Energy May 2013;38(14):5823e35. 2011;36(22):14335e41.
€ Acar MC, Mat MD, Kaplan Y. Effects of operating
[3] Selamet OF, [23] Aouali FZ, Becherif M, Tabanjat A, Emziane M,
parameters on the performance of a high-pressure proton Mohammedi K, Krehi S, et al. Modelling and experimental
exchange membrane electrolyzer: effects of operating analysis of a PEM electrolyser powered by a solar
parameters on the performance of a high-pressure PEM photovoltaic panel. Energy Proced 2014;62:714e22.
electrolyzer. Int J Energy Res Apr. 2013;37(5):457e67. [24] Mench MM. Fuel cell engines. Hoboken, NJ: Wiley; 2008.
[4] Abdol Rahim AH, Tijani AS, Kamarudin SK, Hanapi S. An [25] LaConti AB, Swette L. Special application using PEM-
overview of polymer electrolyte membrane electrolyzer for technology. In: Vielstich W, Lamm A, Gasteiger HA, editors.
hydrogen production: modeling and mass transport. J Power Handbook of fuel cells, vol. 4. Chichester: John Wiley & Sons,
Sources Mar. 2016;309:56e65. Ltd.; 2003. p. 745e61. Chapter 55.
[5] Grigoriev SA, Kalinnikov AA, Millet P, Porembsky VI, [26] Li X, Qu S, Yu H, Hou M, Shao Z, Yi B. Membrane water-flow
Fateev VN. Mathematical modeling of high-pressure PEM rate in electrolyzer cells with a solid polymer electrolyte
water electrolysis. J Appl Electrochem May 2010;40(5):921e32. (SPE). J Power Sources May 2009;190(2):534e7.
[6] Marangio F, Santarelli M, Cali M. Theoretical model and [27] Medina P, Santarelli M. Analysis of water transport in a high
experimental analysis of a high pressure PEM water pressure PEM electrolyzer. Int J Hydrogen Energy Jun.
electrolyser for hydrogen production. Int J Hydrogen Energy 2010;35(11):5173e86.
Feb. 2009;34(3):1143e58. [28] Bessarabov D, Wang H, Li H, Zhao N. PEM electrolysis for
[7] Nie J, Chen Y, Cohen S, Carter BD, Boehm RF. Numerical and hydrogen production: principles and applications. CRC Press;
experimental study of three-dimensional fluid flow in the 2016.
bipolar plate of a PEM electrolysis cell. Int J Therm Sci Oct. [29] Santarelli MG, Torchio MF, Cochis P. Parameters estimation
2009;48(10):1914e22. of a PEM fuel cell polarization curve and analysis of their
[8] Nie J, Chen Y, Boehm RF, Katukota S. A photoelectrochemical behavior with temperature. J Power Sources Sep.
model of proton exchange water electrolysis for hydrogen 2006;159(2):824e35.
production. J Heat Transf 2008;130(4):042409. [30] Schalenbach M, Carmo M, Fritz DL, Mergel J, Stolten D.
[9] Gorgun H. Dynamic modelling of a proton exchange Pressurized PEM water electrolysis: efficiency and gas
membrane (PEM) electrolyzer. Int J Hydrogen Energy Jan. crossover. Int J Hydrogen Energy Nov. 2013;38(35):14921e33.
2006;31(1):29e38. [31] Springer TE, Zawodzinski TA, Gottesfeld S. Polymer
[10] Awasthi A, Scott K, Basu S. Dynamic modeling and electrolyte fuel cell model. J Electrochem Soc
simulation of a proton exchange membrane electrolyzer for 1991;138(8):2334e42.
hydrogen production. Int J Hydrogen Energy Nov. [32] Selamet OF, Ergoktas MS. Effects of bolt torque and contact
2011;36(22):14779e86. resistance on the performance of the polymer electrolyte
[11] Lee B, Park K, Kim H. Dynamic simulation of PEM water membrane electrolyzers. J Power Sources May
electrolysis and comparison with experiments. Int J 2015;281:103e13.
Electrochem Sci 2013;8:235e48. [33] Musio F, Tacchi F, Omati L, Gallo Stampino P, Dotelli G,
[12] Han B, Steen SM, Mo J, Zhang F-Y. Electrochemical Limonta S, et al. PEMFC system simulation in MATLAB-
performance modeling of a proton exchange membrane Simulink® environment. Int J Hydrogen Energy Jul.
electrolyzer cell for hydrogen energy. Int J Hydrogen Energy 2011;36(13):8045e52.
Jun. 2015;40(22):7006e16.

Please cite this article in press as: Yigit T, Selamet OF, Mathematical modeling and dynamic Simulink simulation of high-pressure PEM
electrolyzer system, International Journal of Hydrogen Energy (2016), http://dx.doi.org/10.1016/j.ijhydene.2016.06.022
14 i n t e r n a t i o n a l j o u r n a l o f h y d r o g e n e n e r g y x x x ( 2 0 1 6 ) 1 e1 4

[34] Golnaraghi MF, Kuo BC, Golnaraghi MF. Automatic control [36] Santarelli M, Medina P, Calı̀ M. Fitting regression model
systems. 9th ed. Hoboken, NJ: Wiley; 2010. and experimental validation for a high-pressure PEM
nard P, Chahine R. Simulation of hydrogen
[35] Xiao J, Hu M, Be electrolyzer. Int J Hydrog Energy Mar.
storage tank packed with metal-organic framework. Int J 2009;34(6):2519e30.
Hydrogen Energy Sep. 2013;38(29):13000e10.

Please cite this article in press as: Yigit T, Selamet OF, Mathematical modeling and dynamic Simulink simulation of high-pressure PEM
electrolyzer system, International Journal of Hydrogen Energy (2016), http://dx.doi.org/10.1016/j.ijhydene.2016.06.022