international journal of hydrogen energy xxx (xxxx) xxx

Available online at www.sciencedirect.com

ScienceDirect

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

Mathematical modelling of coupled and decoupled

water electrolysis systems based on existing
theoretical and experimental studies

Antonios Antoniou*, Arturo Berastain, Diego Hernandez, Cesar Celis

Mechanical Engineering Section, Pontificia Universidad Cat
olica del Perú, Av. Universitaria 1801, San Miguel, Lima,
Lima 32, Peru

highlights

A new mathematical model for different electrolyzer technologies is proposed.

Mathematical formulae for voltage, Faraday's, and total efficiencies are provided.

Analysis of three different electrolyzer technologies, Alkaline, PEM, and E-TAC.

New model results are in good agreement with other ones available in literature.

article info abstract

Article history: Since it has the potential to significantly reduce gaseous emissions in the near future, elec-
Received 8 December 2021 trolytic hydrogen production using electricity generated from renewable energy sources, such
Received in revised form as solar radiation, is key. Water splitting processes occurring in electrolyzer cells are complex
16 February 2022 phenomena. Therefore, to fully realize such processes, different technologies have been
Accepted 27 March 2022 accounted for. The focus of this work is on the mathematical modeling of three different
electrolyzer cells related technologies, (i) alkaline, (ii) proton exchange membrane (PEM), and
Keywords: (iii) decoupled water splitting. Accordingly, several existing mathematical models for alkaline
Mathematical modeling and PEM electrolyzers are initially revised. Next, a comprehensive mathematical model
Coupled and decoupled water capable of properly predicting the performance of the three electrolyzer technologies
electrolysis accounted for here is proposed. The developed mathematical models are then used to predict
Alkaline and PEM electrolyzers the behavior of electrolyzer cells under different operation conditions. The obtained results are
Hydrogen production finally compared in terms of cell voltages, cell efficiencies, and hydrogen production rates.
Renewable energy When compared to other results available in the literature, the cell voltage ones obtained using
the new proposed model are in relatively good agreement. Specifically, for a current density
range of 0e200 mA/cm2, cell pressures between 10 and 40 bar, and a cell temperature of 60  C,
cell voltage requirements are between 1.25 and 1.75 V, with the E-TAC technology performing
better than the other two ones accounted for. In addition, for current densities of more than
100 mA/cm2 and cell pressures below 5 bar, Faraday's efficiencies are almost the same for all
three technologies, i.e., about 95%. However, for higher cell pressures, significant differences in
Faraday's efficiency appear. Based on the work carried out, it is concluded that developing a
sound mathematical model is crucial both for the comprehension of coupled and decoupled
water electrolysis-related processes and for their use in the simplest and most reliable way.
© 2022 Hydrogen Energy Publications LLC. Published by Elsevier Ltd. All rights reserved.

* Corresponding author.
E-mail address: [email protected] (A. Antoniou).
https://doi.org/10.1016/j.ijhydene.2022.03.264
0360-3199/© 2022 Hydrogen Energy Publications LLC. Published by Elsevier Ltd. All rights reserved.

Please cite this article as: Antoniou A et al., Mathematical modelling of coupled and decoupled water electrolysis systems based on
existing theoretical and experimental studies, International Journal of Hydrogen Energy, https://doi.org/10.1016/j.ijhydene.2022.03.264
2 international journal of hydrogen energy xxx (xxxx) xxx

Nomenclature nCh
H2 O Number of moles of water in chamber
PCh Pressure at specified chamber (bar)
A Area (cm2) comp
PCh Pressure of component at specified chamber (bar)
aH2O Activity of water
Psv Pressure of saturated vapor at a specified
ax Empirical parameter number x for Faraday's
temperature (bar)
efficiency calculation
PH2 o Water pressure (bar)
bx Empirical parameter number x for Faraday's
Pout Output energy flow of the cell (W)
efficiency calculation
Pin Input energy flow to the cell (W)
CN Constant in Nerst equation (V/C)
Po Ambient pressure (bar)
Ccat
H2 O:mem Water concentration at the cathode side of the
pa Pressure at the anodic gas outlet (bar)
membrane (mol/m3)
PaO2 Partial oxygen pressure at the anode (bar)
CanH2 O:mem Water concentration at the anode side of the
pc Pressure at the cathodic gas outlet (bar)
membrane (mol/m3)
PcH2 Partial hydrogen pressure at the cathode (bar)
C Energy contained in a unit of Hydrogen
psv Saturated vapor pressure (bar)
Cx Constant number x for cell voltage calculation
pcomp Partial pressure of a specific component (bar)
CU Unitless constant independent from cell
Q Electrolyte flow rate (l/min)
technology
q Ohmic resistance from electrode-membrane
CT Unitless constant dependent on cell technology
distance (U m2)
Dx Concentration at a specific working condition
R Universal gas constant (8.314 J/mol.K)
(mol/m3)
Rel Electrode resistance (U cm2)
Dw Diffusion coefficient (cm2/s)
Rpl Resistance of bipolar plates (U cm2)
d Membrane thickness under wet conditions (cm)
Ri Losses related to resistance overpotential (V)
E Energy (kJ)
Rsep Separator resistance (U cm2)
E0 Constant (V)
Rmem Membrane resistance (U cm2)
F Faraday constant (96,485 C/mol)
r Resistance related to ohmic overpotential (U m2)
f ð *Þ Function that depends on electrode-membrane
rx Empirical parameter number x for cell voltage
separation
calculation
fx Empirical parameter number x for Faraday's
T Cell temperature (K)
efficiency calculation
T0 Reference temperature (298 K)
G Gibbs energy (kJ)
t Coefficient related to activation overpotential (V)
gð *Þ Function that depends on membrane thickness
tx Empirical parameter number x for cell voltage
h Specific enthalpy (kJ/kg)
calculation
hð *Þ Function that depends on electrolyte flowrate
S Specific entropy (kJ/kg.K)
I Current across the electrolysis cell (A)
Sð *Þ Function that depends on cell temperature. Used
Iin Input voltage to the electrolyzer cell (A)
for cell voltage calculation
i Current density (A/cm2)
s Coefficient related to activation overpotential (V)
i0 Exchange current density (A/cm2)
Uanode Overpotential of oxygen (V)
iloss Current density losses (A/cm2)
Ucathode Overpotential of hydrogen (V)
k Membrane conductivity at cell temperature (S/cm)
Uuniversal Overpotential independent on technology (V)
KDarcy Membrane permeability (cm2)
Utech Overpotential technology dependant (V)
MH2 O Molar weight of water (kg/mol)
_ production Flow of produced hydrogen (mol/s) Uc Voltage across the electrolysis cell (V)
MOLE H2
UN Nerst voltage (V)
M_
production
H2 Flow of produced hydrogen (kg/s) UNC Constant part of Nerst Voltage (V)
Ncell Number of electrolyzer cells V Overpotential of a specific element (V)
N_
gen
O2 Flow of generated oxygen (mol/s) Vhydrogen Hydrogen overpotential (V)
N_ H2
gen
Flow of generated hydrogen (mol/s) Voxygen Oxygen overpotential (V)
N_
mem
Flow of water through the electrolyzer membrane VOC Open Circuit electrolyzer voltage (V)
H2 O
(mol/s) Vact Activation overpotential (V)
N_ H2 O
diff
Flow of water caused by diffusion (mol/s) Vohm Ohmic overpotential (V)
Vcon Concentration overpotential (V)
N_
eod
H2 O Flow of water caused by electro-osmotic drag
Vin Input voltage to the electrolyzer cell (V)
(mol/s)
Wm Membrane thickness (m m)
N_ H2 O
pe
Flow of water caused by pressure difference (mol/s)
Ws Electrode-membrane separation (m m)
n Valence number of the substance as an ion in the comp
XCh Molar fraction of a component in the chamber
solution
Z Number of electrons
nd Fitting parameter (unitless)
Zð *Þ Function that depends on cell temperature. Used
n_H2per Hydrogen permeation flux (mol/s)
for cell voltage calculation
n_O2per Oxygen permeation flux (mol/s)
z Ohmic resistance from electrolyte flow rate (U m2)
ncomp Number of moles of a component

Please cite this article as: Antoniou A et al., Mathematical modelling of coupled and decoupled water electrolysis systems based on
existing theoretical and experimental studies, International Journal of Hydrogen Energy, https://doi.org/10.1016/j.ijhydene.2022.03.264
 international journal of hydrogen energy xxx (xxxx) xxx 3

a Tafel slope (V) gO2 Oxygen partial pressure increase factor at the
ahyd Tafel slope for hydrogen(V) anode (bar cm2/A)
aoxy Tafel slope for oxygen (V) gcomp Partial pressure increase factor of specific
hf Faraday's efficiency component (bar cm2/A)
hcell Cell efficiency εFick
H2 Hydrogen diffusivity (mol/cm s bar)
~cell
h Cell efficiency (including heat balance) εFick
O2 Oxygen diffusivity (mol/cm s bar)
Darcy
~U
n Voltage efficiency (including heat balance) εH2 Hydrogen permeability (mol/cm s bar)
gH2 Hydrogen partial pressure increase factor at the dmem Membrane thickness (um)
cathode (bar cm2/A) rH2 O Water density (kg/m3)
mH2 O Water viscosity (gm/cms)

Introduction maximize hydrogen production, a good understanding of the

essentials of water electrolysis is crucial to properly operate
Fossil fuels have been extensively used as an energy source, electrolysis-based hydrogen production systems.
but since they are not evenly distributed across the globe, over For coupled water splitting, where both hydrogen and ox-
the years control of them has led to geopolitical conflicts, ygen are simultaneously produced inside an electrolytic cell,
energy wars, and tensions [1e3]. Renewable energies sources, two main technologies are available, (i) alkaline and (ii) proton
such as solar, wind, hydro, and geothermal, can help hu- exchange membrane (PEM). Alkaline-based electrolysis has
manity to both stop energy conflicts around the world and been available for a long time [12], and its main limitations are
move towards a more stable and safer world to live in. As it is related to its relatively low hydrogen production rate and cell
reported in Ref. [4], the European Union (EU) aims to be a efficiency, which can be in the range of 55e65% [11]. Ulleberg
pioneer in green hydrogen use by 2050. The so-called green [13] was one of the first researchers to carry out experimental
deal consists of three phases, beginning with 1 million tons of studies using this technology, and to develop, based on
green hydrogen use by 2024 to 10 million by 2030, and then operational variables such as cell current density and tem-
increasing this hydrogen usage systematically until 2050. In perature, a mathematical non-linear equation to predict the
July 2021, the U.S. Department of Energy (DOE) also associated cell voltage requirement and Faraday's efficiency.
announced a $52.5 million budget to fund 31 projects to After this early work, and as highlighted in the review works
advance next-generation clean hydrogen technologies, which by David et al. [1] and Olivier et al. [14], other efforts have been
can have a major role in supporting President Biden's performed over the years. In Ref. [12] for instance, hydrogen
commitment to tackling the climate crisis [5]. In addition, production using seawater and alkaline electrolysis was
there are many publications around the world focusing on the examined and conclusions were made for hydrogen and
hydrogen economy, along with several discussions about chlorine production rates.
different technologies for hydrogen production, storage, and An electrolyzer-related technology with the potential to
use [6e8]. For the hydrogen economy to be sustainable over replace alkaline electrolyzers, and which can be operated at
time, governments, research institutions, and private com- relatively high current densities (500e9000 mA/cm2), is the
panies need to align their interests and collaborate to achieve PEM one [11]. Compared to alkaline electrolyzes, PEM cells
this goal [9]. feature some advantages [15]. PEM technology allows pro-
Hydrogen is mainly produced from fossil fuels reforming ducing hydrogen at relatively high pressure, reducing in this
(78%), coal gasification (18%), and renewable sources (4%) [1]. way hydrogen storage energy requirements. Pressure effects
When involving renewable sources, water electrolysis is the on gas purity and PEM electrolyzer efficiency have been pre-
main technology used for producing the so-called ‘green’ viously studied in several works. For instance, to predict gases
hydrogen. As it is reported in the literature [10,11], hydrogen cross-over in a pressurized PEM, Schalenbach et al. [16]
production via water electrolysis is a promising technology. developed a mathematical model incorporating Fick's diffu-
Electrolysis can be carried out indeed at both high and low sion law, which allowed estimating oxygen and hydrogen flux
temperatures, but the latter is the most widely used type of densities across membranes of different thicknesses. Simi-
electrolysis nowadays [11]. The present work focuses on this larly, Yodwong et al. [10] studied pressure effects on Faraday's
type of electrolysis. One key aspect in water electrolysis efficiency of alkaline and PEM electrolyzers. Measurements
focused on green hydrogen production is to reduce the cost of performed there using a PEM electrolyzer, at pressures
the process. The electricity required in this case can come ranging from 1 to 10 bar, indicate that Faraday's efficiency is at
from different renewable sources such as solar, wind, and least 96%.
hydro. The basic idea is to be able to produce more hydrogen There are many other previous works dealing with PEM
using less power, which is directly related to the voltage and electrolyzers. For instance, Falcao and Pinto [17] review
current requirements of the electrolytic cells employed. As in several past works focused on PEM electrolyzer modeling. In
any other physical process, to minimize energy input and the referred review, a special emphasis is put on dynamic

Please cite this article as: Antoniou A et al., Mathematical modelling of coupled and decoupled water electrolysis systems based on
existing theoretical and experimental studies, International Journal of Hydrogen Energy, https://doi.org/10.1016/j.ijhydene.2022.03.264
4 international journal of hydrogen energy xxx (xxxx) xxx

behavior, two-phase flow effects, and thermal effects parameters reported in the literature [36,37,41]. Therefore,
accounted for during model development. Similarly, Fragia- the focus in this work is on the development of a mathe-
como and Genovese [18,19] discuss several aspects related to matical model incorporating all the parameters previously
alkaline and PEM electrolyzers-based hydrogen production accounted for, (i) current density, (ii) temperature, (iii) pres-
systems including their modeling and energy demand (ac- sure, (iv) electrolyte flow rate, (v) membrane thickness, and
counting for auxiliary power consumption) [18], and the in- (vi) electrode-membrane separation. The developed model
clusion in their modeling of compression and storage systems can be applied to alkaline, PEM, and E-TAC technologies by
[19]. Fragiacomo and Genovese's models were also used to simply changing a single parameter in the associated voltage
both study the energy performance of a PEM water equations. The same is true in the case of Faraday's effi-
electrolysis-based high-pressure hydrogen refueling station ciency. Besides current density, existing models incorporate
[20] and carry out a technical-economic analysis of a pres- either cell temperature or pressure, but never both. To
surized PEM electrolyzer featuring a storage compression address this issue, a Faraday's model is developed based on a
system [21]. Fragiacomo and Genovese [18] results were used simple exponential expression incorporating the two referred
as well in Ref. [22] to verify the accuracy of the voltage pre- cell operational parameters.
diction of a stack of electrolytic cells performed there. There Summarizing the work that follows, in Section Electrolysis-
are also several past works [23e25] involving polygeneration based water splitting, the basics of electrolysis-based water
systems dealing with the simultaneous production of, for splitting are briefly discussed. The mathematical modeling of
instance, power, hydrogen, hot water, and freshwater. These the three electrolyzer technologies studied here is described
works highlight the need to develop an electrolyzer model in Section Mathematical modeling of electrolyzer
capable of properly describing the behavior of different elec- technologies. In Section Results and discussion, in turn, the
trolyzer technologies that can be used in such polygeneration new mathematical model proposed in this work is high-
systems. For further details about PEM water electrolysis, the lighted. Some conclusions drawn for the work carried out here
interested reader may follow references [26,27]. Notice that are finally provided in Section Conclusions. The main contri-
recent electrolysis-related studies are also seeking new con- butions of this work to electrolysis research include the
cepts to make the associated processes greener, using for following two. (i) The development of a new mathematical
instance magnets to reduce voltage requirements in alkaline model, simple but detailed, capable of predicting cell voltage
electrolyzers [28]. requirements for the three electrolyzer technologies studied
Decoupled water electrolysis, in which water splitting here, and especially for E-TAC, which has not been done
hydrogen (HER) and oxygen evolution reactions (OER) occur at before. And (ii) the development of a mathematical model
different places and/or times, is advancing quickly. The first predicting Faraday's efficiency as a function of, besides cur-
ones who explored this relatively new technology were Symes rent density, both operational cell temperature and pressure.
and Cronin [29]. Later, to decouple the associated electro- This mathematical model based on an Arrhenius type equa-
chemical hydrogen and oxygen reactions, Dotan et al. [30] tion is also simple and requires only two model parameters to
developed an innovative new method called E-TAC (electro- be determined from data fitting.
chemical-thermally activated chemical). Unlike conventional
water electrolysis, the E-TAC water-splitting process de-
couples the HER and OER reactions, which occur at different Electrolysis-based water splitting
times and potentially at different places. Rather than being
continuous, therefore, E-TAC can be described as a ‘batch Understanding the basics of water electrolysis is crucial to
process’ featuring two cycles. The first cycle produces developing a mathematical model properly describing the
hydrogen electrochemically, and the second one oxygen via a associated processes occurring inside water-splitting devices.
spontaneous chemical reaction. Electrolytic cell temperature and pressure, electrodes mate-
The present work focuses on the static mathematical rial, and membrane thicknesses are some of the factors
modeling of the two most mature and widely used coupled directly affecting cell voltage requirement, hydrogen produc-
water electrolysis technologies, alkaline and PEM. E-TAC, tion, and efficiency. Some of these aspects are briefly dis-
which is a promising decoupled water electrolysis technol- cussed in this section.
ogy, is also included in the modeling effort carried out here.
Accordingly, some mathematical models available in previ- Basics of water electrolysis
ous works are firstly discussed. After doing so, a new math-
ematical model is proposed for predicting electrolytic cell Electrolysis-based water splitting requires electricity (elec-
voltage, hydrogen production rate, and cell total efficiency. trons) to split water molecules into their constituent compo-
The main goal is to develop a simple but comprehensive nents of hydrogen and oxygen molecules. The processes
mathematical model properly characterizing the three tech- include an electrochemical one that is highlighted in Eq. (1).
nologies accounted for here. Once developed, to evaluate the The basic parts of a coupled water splitting conventional
effect on cell performance of various cell operating parame- electrolyzer consist of a pair of electrodes submerged into an
ters, several parametric studies are performed using the electrolyte tank and the power supply providing the electric
proposed model. Notice that, currently, there is no cell charge, Fig. 1. Depending on the technology utilized, the
voltage model describing all three technologies studied here. electrolyte tank can have an aqueous solution containing ions
Furthermore, none of the previous models simultaneously and a separation membrane made of Nafion material, or an
combines and uses all physical, operational, and cell design oxygen ion exchange membrane made of ceramic material.

Please cite this article as: Antoniou A et al., Mathematical modelling of coupled and decoupled water electrolysis systems based on
existing theoretical and experimental studies, International Journal of Hydrogen Energy, https://doi.org/10.1016/j.ijhydene.2022.03.264
 international journal of hydrogen energy xxx (xxxx) xxx 5

Fig. 1 e Electrolysis basic principle.

Alternative technologies, such as the E-TAC [30] used for The anode and cathode overpotentials, which mean that
decoupled water splitting, use no membrane. extra energy is needed for oxygen and hydrogen to be formed,
are dependent on the electrode's material. For instance, elec-
1
H2 O þ electrons ¼ H2 þ O2 (1) trodes for alkaline electrolysis are usually made from Nickel,
2
which is a relatively cheap material but with high over-
After being produced, hydrogen and oxygen gases must be potential [34]. Finally, the ohmic voltage is due to several
prevented from mixing, so a safety mechanism must be used factors including the resistance to electron flow through
to detect and shut down the electrolysis system when wires, electrolyte, and separation membrane, as well as the
required [31]. formation of bubbles in the electrolyte.

Voltage requirement Thermodynamics

As shown in Eq. (2), the electrolytic cell voltage is the sum of (i) Theoretically, the amount of energy needed at standard at-
the Nerst voltage, (ii) the overpotential at the electrodes, mospheric conditions to split using electrolysis one mole of
anode and cathode, and (iii) the ohmic voltage drop that is a liquid water into oxygen and hydrogen is 285.9 kJmol-1.
function of current density [32], Applying the second law of thermodynamics to this process
leads to Eqs. (9) and (10) [35],
Uc ¼ UN þ Ucathode þ Uanode þ Ri (2)

The Nernst voltage (UN ) depends on the electrode's partial Dh ¼ DG þ TDS (9)
hydrogen and oxygen pressures and the cell temperature [32],
! DG ¼ nFDE (10)
pffiffiffiffiffiffiffi
RT Pc PaO2
UN ¼ UNC þ ln H23=2
 CN T (3) In these two equations, DG is the Gibbs free reaction energy
2F Po aH2O (237 kJmol-1 at standard conditions) to be supplied using
The overpotential due to both hydrogen and oxygen re- electricity. DE is, in turn, a function of temperature and
actions can be in turn approximated by the Tafel equation [33], pressure, which at standard conditions is equal to 1.23 V.
When water in liquid form is supplied to electrolyzers, for the
i ¼ i0 eV=a (4) electrolysis reaction, the higher value of the thermoneutral
voltage (Eq. (9)) is utilized [36,37].
2:3RT
ahyd ¼ (5)
aF Hydrogen generation

2:3RT Faraday's law highlighted in Eq. (11) is usually employed to

aoxy ¼ (6)
ð1  aÞF calculate the hydrogen production rate in moles per second.
where, As noticed from this expression, the amount of hydrogen
  produced is a function of Faraday's efficiency ðhf Þ, which is
2:3RT i usually estimated based on experiments [10,37], the number
Uhyd ¼ ln (7)
aF i0 of electrolytic cells in series Ncell , the number of electrons
 
2:3RT i involved during the reaction Z, the Faraday's constant F, and
Uoxy ¼ ln (8)
ð1  aÞF i0 the current supplied to the electrolyzer I [35],

Please cite this article as: Antoniou A et al., Mathematical modelling of coupled and decoupled water electrolysis systems based on
existing theoretical and experimental studies, International Journal of Hydrogen Energy, https://doi.org/10.1016/j.ijhydene.2022.03.264
6 international journal of hydrogen energy xxx (xxxx) xxx

_ production ¼ hf Ncell I
MOLE (11)
H2
ZF

Cell efficiency

From the very first principle, the electrolyzer cell efficiency

can be defined as the ratio between the electrolyzer power
output and the input power to the electrolyzer, as shown in
Eq. (12) [38],

Pout
hcell ¼ (12)
Pin
The input power in this case is a function of the current
and voltage supplied to the electrolyzer, and the electrolyzer
power output is a function of the hydrogen production rate
and the hydrogen energy density. Therefore, for the same
hydrogen production rate, to maximize cell efficiency, the Fig. 2 e Concept of Alkaline electrolyzer.
voltage requirement must be minimized. After this brief re-
view of water electrolysis basics, mathematical models of
different electrolyzer technologies are discussed in the next Ulleberg model
section. Ulleberg [35] developed a cell voltage and Faraday's efficiency
model as a function of both temperature (in  C) and current
density (in A/m2). The corresponding model equations include
Mathematical modeling of electrolyzer several empirical parameters that were estimated using
technologies regression analysis based on experimental data. Comparisons
between the predicted results, obtained using the mathe-
Experimental studies are important both to verify the accu- matical expressions given in Eqs. (13), (14a), and (14b), and the
racy of theoretical models and to understand the similarities/ experimental ones are in very good agreement. The associated
differences among different electrolyzer technologies. In the model parameters are given in the original works by Ulleberg
past, several research works carried out water electrolysis- [13,36].
related experiments and developed complex mathematical 0 1
Bt1 þ T þ T2
t2 t3
models to match their experimental data. Some of these r1 þ r2 T C
Uc ¼ UN þ I þ s log@ I þ 1A (13)
studies and their associated models are reviewed here. A A

Alkaline electrolyzers !
a2 þ a3 T þ a4 T2 a2 þ a3 T þ a4 T2
hf ¼ a1 exp þ (14a)
Alkaline electrolyzers (AEL) are one of the first technologies I=A ðI=AÞ2
developed for water electrolysis, so this technology is
considered mature today. Since noble materials are expen-  i 2
sive, one of the main advantages of AEL is the use of non- hf ¼ A
 i 2 f2 (14b)
f1 þ A
noble ones. One of their main disadvantages is, in turn, the
material corrosion processes occurring inside them, coming
Amores model
from the use of a caustic water medium, usually potassium
Amores et al. [41] extended the basic Ulleberg voltage model
hydroxide (KOH) (25e30%), sodium hydroxide (NaOH), or so-
[37] to account for other important operational parameters
dium chloride (NaCL) [39,40]. The concept of AEL is shown in
reported before in electrolysis literature [42e45]. Accordingly,
Fig. 2.
Amores et al. [41] model includes, besides temperature and
In the past, many authors have developed mathematical
current density, membrane thickness, distance between
models for alkaline electrolyzers. The referred models include
electrodes, and electrolyte volumetric flow rate. This mathe-
several expressions and algorithms for computing cell
matical model is described in Eq. (15),
voltage, Faraday's efficiency (or current efficiency as it is
sometimes referred to), and cell total (overall) efficiency. To Uc ¼ UN þ ðr þ q þ zÞi þ s logðt i þ 1Þ (15)
have the required basis for developing a more comprehensive
From experiments, Amores et al. [41] concluded that the
mathematical model for the three technologies accounted for
most important operational factor is the cell temperature,
here, an extensive literature review was initially carried out.
which varies inversely with the required cell voltage. They
After doing so, it was decided to analyze three previous AEL
also observed that the separation distance between electrodes
models, which are available in the research works by Ulleberg
and membrane affects the required cell voltage. The electro-
(2003) [37], Amores et al. (2017) [41], and Schalenbach et al.
lyte volumetric flow rate was another important factor
(2016) [36].
considered in their experiments. Regarding this factor, it was

Please cite this article as: Antoniou A et al., Mathematical modelling of coupled and decoupled water electrolysis systems based on
existing theoretical and experimental studies, International Journal of Hydrogen Energy, https://doi.org/10.1016/j.ijhydene.2022.03.264
 international journal of hydrogen energy xxx (xxxx) xxx 7

concluded that at separation distances greater than 4 mm, the

volumetric flow rate has no effect on cell voltage. Neverthe-
less, for small gaps, by reducing bubble concentrations on the
electrodes, the flow rate helps to reduce the required voltage.

Schalenbach model
Schalenbach et al. [36] used the mathematical model given by
Eq. (16) to estimate electrolyzer voltages for both alkaline and
acidic cells. As noticed from this expression, the cell voltage is
a function of current density, temperature, hydrogen and
oxygen partial pressures at both anode and cathode, the dis-
tance between the separation membrane and the electrodes,
and the separation membrane thickness. For the case of
alkaline cells, the hydrogen and oxygen partial pressures (gH2
and gO2) were set equal to zero.

Ucell ¼ 1:229  0:000846ðT  298:15Þ

0 qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi1
RT @ p þ gH2 i  psv
c
pa þ gO2 i  psv   Fig. 3 e Concept of proton exchange membrane (PEM)
þ ln A þ Rel þ Rsep i
2F 3 = electrolyzer.
Po 2

 
þ a ln i=i (16) hydrogen and oxygen produced are expressed as a function of
o
both current and Faraday's constant,
The mathematical expressions used to compute the Fara-
day's or current efficiency is given by either Eq. (17) or Eq. (18), I
N_ O2 ¼
gen
(20)
depending on the dominant factor, accounting for current 4F
density, membrane thickness, and hydrogen and oxygen
I
N_ H2 ¼
gen
partial pressures. (21)
2F
2F  Fick  c   a 
In this model, partial pressures are also calculated ac-
hf ¼ 1  ε p þ gH2 i  psv þ εFick
O2 p þ gO2 i  psv (17)
id H2
counting for species molar fractions as follows,

2F  c Darcy

comp ncomp
hf ¼ 1  ðp  pa ÞεH2 (18) XCh ¼ (22)
id ncomp þ nCh
H2 O

The total (overall) cell efficiency is, in turn, defined as the

comp
product of both voltage and current efficiencies, Eq. (19), pcomp ¼ XCh PCh (23)

~U hf
~cell ¼ n
h (19) The third part of the model deals with the membrane flow
dynamics. In particular, the net flow of water through the
PEM electrolyzers membrane is computed considering three different mecha-
nisms, diffusion, electro-osmotic drag, and hydraulic pressure
In recent years, the number of studies involving proton ex- effect,
change membrane (PEM) electrolytic cells has significantly
N_ H2 O ¼ N_ H2 O þ N_ H2 O  N_ H2 O
mem diff eod pe
increased. Indeed, in the work by Schmidt et al. [40], several (24)
experts believe that by 2030 PEM technology will be the
The mathematical expressions for each RHS (right-hand
dominant one. In the referred work, due to production scale-
side) term appearing in Eq. (24) are the following ones,
up, future capital costs of electrolysis technologies,
including PEM, are expected to reduce up to 30%. So, PEM ADw  cat 
N_ H2 O ¼
diff
CH2 O:mem  Can
H2 O:mem (25)
electrolyzers, whose concept is highlighted in Fig. 3, have also dmem
been considered here. More specifically, to propose a new
mathematical model, three past models have been analyzed. nd I
N_ H2 O ¼
eod
(26)
These three models are fully described in the works by Abdin F
et al. (2015) [46], Espinosa-Lopez et al. (2018) [47], and Koponen
KDarcy ArH2 O DP
N_ H2 O ¼
pe
et al. (2020) [48]. (27)
dmem mH2 O MH2 O
Abdin model Based on the parameters computed in the first three parts
Abdin et al. [46] describe a PEM electrolyzer model featuring of the model, and using Eq. (28), the PEM electrolyzer polari-
four parts. The first two parts deal with balances of mass in zation curve is calculated in the fourth one. The RHS terms of
the cell anode and cathode, where Faraday's efficiency is the polarization curve equation, corresponding to the open-
assumed to be equal to 100%. Accordingly, the amounts of circuit voltage, and the activation, ohmic, and concentration

Please cite this article as: Antoniou A et al., Mathematical modelling of coupled and decoupled water electrolysis systems based on
existing theoretical and experimental studies, International Journal of Hydrogen Energy, https://doi.org/10.1016/j.ijhydene.2022.03.264
8 international journal of hydrogen energy xxx (xxxx) xxx

overpotentials, respectively, are computed in turn using Eqs. empirical model featuring two sub-models. The first sub-
(29)e(32). model computes the required cell voltage for a specific cur-
rent density, whereas the second one mainly focuses on Far-
V ¼ VOC þ Vact þ Vohm þ Vcon (28)
aday's efficiency. Hydrogen and oxygen partial pressures
" required for computing the polarization curve are determined
pffiffiffiffiffiffiffi!#
RT pH2 pO2 in this model using the expression given in Eq. (38). This
VOC ¼ Eo þ ln (29)
2F aH2 O expression includes a temperature-dependent term related to
the saturated water pressure (Psv ), and a pressure-dependent
 
RT i one containing the pressure-increasing factor gcomp , which
Vact ¼ arcsinh (30)
aF 2io depends on membrane thickness, hydrogen permeability, and
cell hydrogen content.
 
Vohm ¼ Rel þ Rpl þ Rmem AI (31) comp
PCh ¼ PCh þ gcomp i  Psv (38)
 
RT D1 Koponen et al. [48] model computes the operating cell
Vcon ¼ ln (32)
zF Do voltage using Eq. (28) and neglecting voltage losses due to
concentration overpotentials. The specific equations utilized
Espinosa-Lopez model to estimate the remaining Eq. (28) RHS terms are the following,
Espinosa-Lopez et al. [47] developed a PEM electrolyzer model
 rffiffiffiffiffiffiffi

that can be divided into 2 sub-models. The first sub-model RT pH2 pO2
VOC ¼ Eo þ ln (39)
deals with the system electrochemical modeling, whereas 2F Po Po
the second one focuses on the electrolytic cell temperature.
 
Particle swarm optimization (PSO) was used in this case to i
Vact ¼ a ln (40)
compute the associated model empirical parameters. Finally, io
the developed model was validated using a 46 kW PEM elec-
 
trolyzer's actual operating data. d
Vohm ¼ þ Rmem AI (41)
As in the case of the Abdin et al. [46] model, Faraday's ef- k
ficiency is considered in this case as being equal to 100%. This
According to this model, when computing Faraday's effi-
means that Eqs. (20) and (21) are also utilized in this model.
ciency, hydrogen losses are quantified in terms of the electric
Hydrogen and oxygen partial pressures are computed in this
current used for hydrogen production. Following [32], the
case using Dalton's law, Eq. (33), where the water partial
referred losses are attributed in this case to two factors, (i)
pressure comes from an empirical expression that depends on
crossover of hydrogen back to the anode side, and (ii) recom-
cell temperature, Eq. (34).
bination of permeated oxygen with hydrogen to form water on
comp
PCh ¼ PCh  PH2 o (33) the cathode side. Accordingly, the expressions used to calcu-
late Faraday's efficiency are the following ones,
 

T  273:15 i  iloss
PH2 o ¼ 6:1078  103 exp 17:2694 (34) hF ¼ (42)
T  34:85 i
In the electrochemical model, to compute the operating
2F  
cell voltage (Eq. (28)), the voltage losses mechanisms accoun- iloss ¼ n_H2per þ 2n_O2per (43)
A
ted for are the same as those considered in the Abdin et al. [46]
model. Nevertheless, the specific values of the voltage losses εH2
are computed according to, n_H2per ¼ pH (44)
d 2
" pffiffiffiffiffiffiffi!#
RT pH2 pO2 εO2
VOC ¼ Eo þ ln (35) n_O2per ¼ pO (45)
2F PH2 O d 2

  Decoupled water splitting electrolyzers

RT i
Vact ¼ arcsinh (36)
2aF 2io
Since the development of the decoupled water electrolysis
concept by Symes and Cronin in 2013 [29], several research
Vohm ¼ Rmem AI (37) works [14,30,49] further explored this potential avenue. In the
Notice that the voltage loss due to concentration over- referred works, different ways of carrying out decoupled
potentials, which becomes relevant in very high current den- water electrolysis were discussed. Some of the decoupled
sities only, was not considered in this model. This was done water electrolysis-related approaches tried in the past involve
because no concentration overvoltages were observed in the producing oxygen first and then hydrogen at different times,
associated experiments, so they were neglected in the modeling. or vice versa, with hydrogen being produced first and then
oxygen. One promising decoupled water electrolysis tech-
Koponen model nology, developed by Dotan et al. [30], is E-TAC. In this case,
To study the effect of power quality on the design of PEM via electrochemical reactions, water molecules are split by
electrolyzers, Koponen et al. [48] proposed a PEM semi- producing hydrogen first and then oxygen. This technology

Please cite this article as: Antoniou A et al., Mathematical modelling of coupled and decoupled water electrolysis systems based on
existing theoretical and experimental studies, International Journal of Hydrogen Energy, https://doi.org/10.1016/j.ijhydene.2022.03.264
 international journal of hydrogen energy xxx (xxxx) xxx 9

allows producing hydrogen at relatively high Faraday's and

Table 1 e E-TAC experimental results [30].
overall electrolyzer efficiencies, which are close to 96%. Based
on this high efficiency, and since there is no model for it yet, Current density Voltage
(mA/cm2) (V)
this technology has been also chosen for modeling purposes
here. 10 1.44
50 1.50
100 1.56
E-TAC
200 1.60
E-TAC water-splitting technology [30] is still in the research
and development stage, but it seems to be a very promising
one to replace or complement existing electrolyzer technolo- New mathematical model
gies. This technology uses the same principles as those
employed in alkaline-based electrolysis, but it is done in two By combining previous electrolyzer modeling approaches,
steps. The anode is charged first and this leads to the occur- some of which have been briefly described in Sections
rence of an electrochemical hydrogen evolution reaction, and Alkaline electrolyzers and PEM electrolyzers, a new, compre-
therefore to the production of hydrogen, Fig. 4 (left). This hensive, and simple mathematical model has been developed.
process usually occurs at ambient temperatures, whereas In the model development process, when necessary, curve
alkaline electrolysis does so in a temperature range of fitting through experimental data was used to estimate some
50e80  C. What is important here is that the E-TAC cathode model parameters for each electrolyzer technology accounted
can be made with the same material as in the case of AEL. In for. The new proposed model can predict the performance of
the original experimental work [30], a platinized nickel-coated the three electrolyzer technologies studied in this work. As
stainless-steel mesh cathode was utilized. Since it exhibits such, it includes mathematical expressions for electrolytic cell
high energy density and cyclability, a Ni(OH)2/NiOOH cathode, voltage, Faraday's efficiency directly affecting hydrogen pro-
widely employed in alkaline batteries, was specifically uti- duction, and total cell efficiency. Notice that the main
lized. Notice that here Ni(OH)2 electrochemically changes to requirement for its development was that the model is fast
NiOOH according to the following reaction, Ni(OH)2 þ OHe 4 enough to be utilized in sizing and optimization processes of
NiOOH þ H2O þ e. As observed from Fig. 4 (left), the reaction hydrogen plants, involving hundreds or even thousands of
products are NiOOH, water, and electrons. model evaluations. The specifics of the new electrolyzer
Currently, there is no mathematical model describing the model developed in this work are provided in this section.
behavior of E-TAC electrolyzers. In the literature [30], how-
ever, for a range of current density of 0e200 mA/cm2, there are Cell voltage
experimental data associated with the operation of this type The development of the new mathematical model proposed
of electrolyzers, Table 1. So, this data has been utilized here as here involved several steps. Accordingly, the first step
a reference to develop a new mathematical model for this included the modifications of the exponential term in Ulle-
decoupled water electrolysis technology. The details of the berg's voltage [37] with the first term appearing in the corre-
referred model are provided in the following section. sponding cell voltage expression given by Schalenbach et al.

Fig. 4 e Two-step decoupled water splitting process.

Please cite this article as: Antoniou A et al., Mathematical modelling of coupled and decoupled water electrolysis systems based on
existing theoretical and experimental studies, International Journal of Hydrogen Energy, https://doi.org/10.1016/j.ijhydene.2022.03.264
10 international journal of hydrogen energy xxx (xxxx) xxx

[36], which is a function of temperature. The second step pffiffiffiffiffiffiffi!

RT Pc PaO2 RT  0:5 
involved, in turn, accounting for changes in the reversible ln H2 ¼ ln P (52)
2F P3=2
o aH2O
2F
voltage as a function of pressure, which should lead to in-
creases in this voltage as pressure increases. This pressure- Accordingly, the complete cell voltage mathematical
related modification was carried out based on the work by model, Eqs. (53)e(58), is a function of current density (i), pres-
Seger [33]. In the third step, both the electrolyte flow rate and sure (P), temperature (T), membrane thickness (Wm ), electrode-
the electrodes to membrane distance parameters from membrane separation distance (Ws ), and electrolyte volu-
Amores et al. [41], and the membrane thickness one from metric flow rate (Q). Additional details of some of the terms
Schalenbach et al. [36] were included. In the proposed model, appearing in Eq. (53) are highlighted in Eqs. (54)e(58), and the
all these parameters were added as part of the ohmic re- corresponding model parameters are provided in Table 3.
sistances. Finally, with some further modifications and sim-
plifications, the complete mathematical model was derived. RðT þ 273:15Þ  0:5 
Uc ¼ E0  C1 ðT  T0 Þ þ ln P þ ½f ðWs Þ
In the newly developed model, thus, the cell voltage is 2F
þ gðWm Þ þ hðQ; Ws Þi þ SðTÞ½lnðZðTÞi þ 1ÞðCU þ CTÞ (53)
given by Eq. (46). This means that the cell voltage involves, as
customary, the sum of the reversible voltage, the ohmic
f ðWs Þ ¼ C2 Ws (54)
voltage drop that is a function of current density, and the
overpotential at the electrodes, anode and cathode. In elec-
trolysis systems, these resistances can be classified into three gðWm Þ ¼ C3 Wm (55)
categories. (i) The first category includes all the electrical re- .
sistances, (ii) the second one does so the reaction resistances, hðQ; Ws Þ ¼ C4 Q ð1mm þ Ws Þ2 (56)
and (iii) the third one, in turn, the transport resistances. As
indicated in Eq. (47), to predict the voltage requirement for the SðTÞ ¼ C5 þ C6 T (57)
three studied technologies, the Ucathode þ Uanode term is split
into two new terms, Uuniversal þ Utech . The mathematical C8 C9
expression for these two new terms, Eq. (48), which represent ZðTÞ ¼ C7 þ þ (58)
T T2
the anode and cathode overvoltages, includes, in turn, two
Finally, it is worth noticing that the parameters appearing
additional parameters, CU and CT. As highlighted in Eq. (50),
in the voltage related equations, Eqs. (53)e(58), indicate or are
CU takes a constant value for the three studied technologies.
related to the following. The U0 term is the thermoneutral
However, as shown in Table 2, the CT values are technology-
voltage [36] (constant term) and C1 is a multiplication factor
dependent. Notice that the CU and CT values were obtained
related to the term involving the difference between the
from calibration using experimental results [30,37,46] char-
electrolyzer and ambient temperatures, the first and second
acterizing the three electrolyzer technologies accounted for
terms in Nerst voltage, respectively. C2 , C3 and C4 are constant
here.
parameters multiplying, respectively, the physical distance
Uc ¼ UN þ Ri þ Ucathode þ Uanode (46) between the electrodes and membrane, the membrane
thickness, and the electrolyzer volumetric flow rate [41]. The
Ucathode þ Uanode ¼ Uuniversal þ Utech (47) above-mentioned parameters are part of the ohmic voltage
drop, due to electron and anion transfer, which is a linear
Uuniversal þ Utech ¼ SðTÞ½lnðZðTÞi þ 1ÞðCU þ CTÞ (48) function of current density. Parameters C5 to C9 are constants
parameters appearing in the last nonlinear term representing
Ri ¼ ½f ðWs Þ þ gðWm Þ þ hðQÞi (49) the overvoltage on both cell cathode and anode [37]. CU and CT
are extra parameters appearing due to the mathematical
CU ¼ 0:25 (50) splitting (into two parts) of the overvoltage related term car-
ried out to represent different electrolyzer technologies. As
In addition, the Nerst voltage UN is computed according to
Eq. (51), where the last RHS term comes from Eq. (52).
pffiffiffiffiffiffiffi!
RT Pc PaO2
UN ¼ U0  C1 ðT  T0 Þ þ ln H23=2
(51) Table 3 e Parameters of the proposed model.
2F Po aH2O
Variable Value
Uo 1:229V
To 25 C
C1  8:46e4 V C1
C2 4:5e6 Vm2 ðmmAÞ1
Table 2 e CT parameter values for different electrolyzer C3 0:45Vm2 ðmmAÞ1
technologies. C4 9e6 minm2 VðlAÞ1
Technology Value of C5 0:243V
CT (unitless) C6  0:001V C1
C7 0:0160A1 m2
Alkaline 0.21 
PEM 0.02
C8  1:3020A1 m2 C
C9 1 2 2
E-TAC 0.00 421:3A m C

Please cite this article as: Antoniou A et al., Mathematical modelling of coupled and decoupled water electrolysis systems based on
existing theoretical and experimental studies, International Journal of Hydrogen Energy, https://doi.org/10.1016/j.ijhydene.2022.03.264
 international journal of hydrogen energy xxx (xxxx) xxx 11

mentioned before, these two parameters were obtained study using the new electrolyzer model proposed here is also
through model calibration using experimental data [30,37,46]. described.

Faraday's efficiency and hydrogen production rate Model verification

As shown in Eq. (11), the hydrogen production rate (in moles
per second) is a function of the current input, Faraday's effi- To verify the prediction accuracy of the new electrolyzer
ciency, and the number of electrolyzer cells. Therefore, to model developed here, a cell voltage-based comparison of the
properly predict the amount of hydrogen produced, an accu- results obtained using this model and others obtained
rate prediction of Faraday's efficiency is a must. Many re- experimentally in the past has been carried out. The results of
searchers in the past have developed mathematical models to this comparison for the three technologies accounted for in
estimate this efficiency using different mathematical ex- this work, alkaline, PEM, and E-TAC, are shown in Fig. 5. From
pressions. The Faraday's efficiency model suggested in this Fig. 5 left plots, it is possible to see that, for current densities
work is an extension of a previous one [38] including, besides between 0 and 200 mA/cm2, the new model results are in
temperature, pressure as a variable. More specifically, an relatively good agreement with those obtained experimentally
Arrhenius type expression, Eq. (59), is suggested here, which is by Ulleberg [37], Abdin et al. [46], and Dotan et al. [30]. More
a function of T (in  C), P (in bars), and current density, i (in A/ specifically, for current densities up to 100 mA/cm2, alkaline
m2). The constant model parameter b1 is equal to 0.1 for PEM and E-TAC technologies operate at almost the same cell
and 0.38 for alkaline electrolyzers. b2 is in turn equal to 0.3. voltage, whereas PEM technology operates at a relatively
This Faraday's efficiency model is simple and involves only lower voltage. For current densities in the range of
two parameters, b1 and b2 , which need to be determined from 100e150 mA/cm2, in turn, PEM and E-TAC operate at the same
experimental data, using curve fitting for instance. To the best voltage and alkaline at a higher one. In addition, Fig. 5 right
of the author's knowledge, this model is the only one in plots show, for each technology studied here and different cell
literature incorporating both temperature and pressure. To temperatures, the cell voltage predictions made using the
compute Faraday's efficiency, other existing models use either newly developed model. Cell temperature was chosen here to
temperature or pressure, but none of them both. produce these results because this parameter is key in low-
  temperature electrolysis. From Fig. 5 right plots it can be
b1 Tð1  ðP  1Þb2 Þ observed that the alkaline technology is more sensitive to
hf ¼ 100 exp (59)
i
temperature changes, whereas PEM and E-TAC technologies
show a similar response to variations in cell temperature.
Total cell efficiency
As highlighted by Eq. (12), from the very first principle, the
Comparison of electrolyzer models
electrolyzer cell efficiency can be defined as the ratio between
the electrolyzer power output (Pout ) and the input power (Pin ) to
Results obtained using the previous alkaline and PEM models
the electrolyzer [38]. These power-related parameters exiting
studied here and the new model proposed in this work have
and entering electrolyzer cells are computed in this work
also been compared. The outcomes from such comparisons
using Eqs. (60) and (61), respectively. In these two equations,
are illustrated in Fig. 6 (alkaline) and Fig. 7 (PEM). In the case of
the hydrogen production (in kg/s) is computed from Eq. (62),
the alkaline cells, as observed from Fig. 6 (left), the new model
whereas the constant parameter C is equal to C ¼ 39:40KWh=
behaves similarly to the Ulleberg's one [37]. This occurs
kgy141:70MJ=kg ¼ 0:283MJ=mol.
because Ulleberg's experimental data have been used to fine-
tune the constant parameters of the electrolyzer model pro-
Pout ¼ CM_ H2
production
(60)
posed here. Furthermore, because of the values of the co-
efficients s and t appearing in Eq. (15), which represent
Pin ¼ Iin Vin (61)
overpotential losses at the anode and cathode, Amores et al.
[41] model predicts cell voltages lower than the other models.
M_ H2
production
_ production ð0:002kg = molÞ
¼ MOLE (62) Schalenbach et al. [36] model predicts, in turn, very high
H2
voltage values compared to the other alkaline models. To
The above presented mathematical models for electrolyzer
illustrate the dependence of the studied alkaline models on
cell voltage, Faraday's efficiency, hydrogen production, and
cell temperature, Fig. 6 (right) shows results obtained with
total cell efficiency are complete and ready to be verified using
these models accounting for two different cell temperatures,
data available in the literature. The results from such verifi-
40 and 80  C. It is observed from this figure that both Ulleberg
cations are discussed in the next section, along with a para-
[37] and Schalenbach et al. [36] models significantly depend on
metric analysis carried out using the newly developed model.
temperature changes, whereas the Amores et al. [41] one does
not. This occurs because in the Amores et al. [41] model, i.e., in
the logarithmic term representing the overpotentials in the
Results and discussion
anode and cathode, specifically, there is no temperature-
dependent factor. As expected as well, after being fine-tuned
Results obtained using the mathematical models described in
using Ulleberg's experimental data, for all cell temperatures
Section Mathematical modeling of electrolyzer technologies,
analyzed here, both the proposed model and the Ulliberg's one
and their comparison with others available in the literature,
behave similarly.
are discussed in this section. A voltage-related parametric

Please cite this article as: Antoniou A et al., Mathematical modelling of coupled and decoupled water electrolysis systems based on
existing theoretical and experimental studies, International Journal of Hydrogen Energy, https://doi.org/10.1016/j.ijhydene.2022.03.264
12 international journal of hydrogen energy xxx (xxxx) xxx

Fig. 5 e New model results versus experimental ones (left) and new model results for different cell temperatures (right). Top,
center, and bottom plots correspond to alkaline, PEM, and E-TAC technologies, respectively.

Fig. 6 e Comparison of alkaline electrolyzer models. Results obtained for an electrolytic cell pressure of 1 bar and
temperatures of 80  C (left) and 40 and 80  C (right).

Please cite this article as: Antoniou A et al., Mathematical modelling of coupled and decoupled water electrolysis systems based on
existing theoretical and experimental studies, International Journal of Hydrogen Energy, https://doi.org/10.1016/j.ijhydene.2022.03.264
 international journal of hydrogen energy xxx (xxxx) xxx 13

Fig. 7 e Comparison of PEM electrolyzer models. Results obtained for an electrolytic cell pressure of 1 bar and temperatures
of 40  C (left) and 40 and 80  C (right).

Regarding the PEM models’ results, Fig. 7 shows a com- to changes in cell membrane thickness and electrolyzer
parison of cell voltage versus current density predictions operational conditions such as cell pressure and temperature.
carried out using the new model proposed here and the three Accordingly, the effect of pressure on cell voltage for the three
previous PEM models analyzed in this work, i.e., Abdin et al. studied technologies is shown in Fig. 8. It is noticed from this
(2015) [46], Espinosa-Lopez et al. (2018) [46], and Koponen et al. figure that, for all three technologies, when pressure increases
(2020) [48]. As it is observed from Fig. 7, the new model leads to cell voltage decreases. In addition, alkaline technology seems
results between those obtained using the Abdin et al. (2015) to be slightly more affected by pressure changes than the
[46] and Espinosa-Lopez et al. (2018) [47] models. It is worth other technologies. It is worth noticing here that cell pressure
noticing here that the influence of cell temperature changes does not affect electrolyzer voltages as much as cell temper-
on the new model results is relatively large (Fig. 7, right). This ature does (Fig. 5, right plots).
occurs because the new model proposed here has a term that Model results for cell voltage at two different membrane
varies strongly nonlinearly with cell temperature, Eq. (53) last thicknesses (0.37 and 1.3 mm) for both alkaline and PEM
term. In the case of the Koponen et al. (2020) model [48], since technologies are shown in Fig. 9. Indeed, for both technolo-
both PH2 and PO2 are functions of temperature [46], cell tem- gies, when membrane thickness increases, cell voltage in-
perature implicitly appears in nonlinear form in Eq. (39). In the creases as well. This increase in cell voltage comes from the
Espinosa-Lopez et al. (2018) [47] model, in turn, cell tempera- increase in ohmic resistance. Accordingly, as shown in Fig. 9,
ture appears explicitly in the Eq. (34) exponential term and cell voltage increases with the increase in ohmic resistance.
implicitly in Eq. (36). Finally, in the Abdin et al. (2015) model As such, membrane thickness is an important factor directly
[46], cell temperature appears multiplied by different affecting the voltage requirements of electrolytic cells used for
nonlinear terms in Eqs. (29), (30) and (32). The results shown in coupled water electrolysis. Since it has no membrane, E-TAC
Fig. 7 (right) suggest that the different cell voltages trends technology has not been included in this analysis.
observed when varying cell temperature come from the way The prediction accuracy of the Faraday's efficiency model
in which this cell parameter is included in these PEM models. proposed here was also verified using the experimental results
obtained by Ulleberg [37] (alkaline) and Yodwong et al. [10]
Voltage-related parametric study using the new electrolyzer (PEM). The results from the referred verifications are shown in
model Fig. 10. As it can be observed from this figure, overall, the model
Using the newly developed model, a parametric study based results agree relatively well with the experimental ones. In the
on cell voltage has been carried out to investigate its response case of alkaline cells, however, at current densities lower than

Fig. 8 e Effect of pressure on cell voltage for the three electrolyzer technologies accounted for.

Please cite this article as: Antoniou A et al., Mathematical modelling of coupled and decoupled water electrolysis systems based on
existing theoretical and experimental studies, International Journal of Hydrogen Energy, https://doi.org/10.1016/j.ijhydene.2022.03.264
14 international journal of hydrogen energy xxx (xxxx) xxx

Fig. 9 e Effect of membrane thickness on cell voltage for alkaline (left) and PEM (right) electrolyzers. Green dashed curve:
thickness 0.37 mm, Red dotted curve: thickness 1.3 mm.

Fig. 10 e Faraday's efficiency comparison of new model results (lines) and experimental data (symbols) for alkaline (left) and
PEM (right) electrolyzers.

about 25 mA/cm2, some discrepancies between numerical and Figs. 12 and 13. More specifically, it is noticed from Fig. 12 that,
experimental results appear. It is observed as well that, in for the range of current densities, temperatures, and pres-
general, when compared to alkaline ones, PEM electrolyzers sures analyzed here, hydrogen production is only slightly
feature higher Faraday's efficiencies, which asymptotically influenced by cell temperature and pressure. Even so, this
tend to 100% as current density increases. influence seems to be more pronounced in the case of alkaline
The effect of cell pressure on Faraday's efficiency has also cells, especially when the cell temperature is relatively high.
been analyzed using the new model proposed here. This has Notice that in this figure the curve identified as “ideal” corre-
been done for both alkaline and PEM electrolyzers, whereas sponds to the E-TAC technology, which considers Faraday's
for E-TAC, following what is reported in the literature [30], a efficiency as being equal to 100%.
constant value of 100% has been considered. Thus, as In the case of the total cell efficiency, and since this
observed from Fig. 11, according to the Faraday's efficiency parameter is affected by both Faraday's efficiency and cell
model proposed here, Faraday's efficiency is significantly voltage, Fig. 13 shows that both cell temperature and pressure
affected by cell pressure. This is particularly true in the case of influence this parameter. Specifically, when compared to
alkaline electrolyzers. For instance, for a current density of temperature, cell pressure has a larger influence on the total
100 mA/cm2, at 1 bar Faraday's efficiency is around 97%, cell efficiency. In addition, from Fig. 13 it is noticed that, as
whereas at 80 bar this efficiency is about 45%. These findings current density increases, alkaline and PEM total efficiencies
need to be confirmed in the future through experimental tests first increase and then decrease. In the case of E-TAC, this
or more detailed numerical analyses. efficiency continually decreases with the increase in current
As shown in Fig. 11, cell pressure affects Faraday's effi- density. Since the total cell efficiency represents the voltage
ciency. In turn, Faraday's efficiency affects both hydrogen efficiency times the Faraday's one, Fig. 13 findings reflect the
production and total cell efficiency. Consequently, it is ex- combined effect of these two parameters. It is worth noticing
pected that both cell temperature and pressure also affect the here that, up to a given point, Faraday's efficiency increases
amount of hydrogen produced by electrolytic cells and their with the increase in current density and then it remains
total efficiency. To assess this aspect, therefore, using the relatively constant (Fig. 10), whereas the voltage efficiency
models developed here, several model runs have been per- continuously decreases with the increase in current density
formed and the results from such runs are summarized in [37].

Please cite this article as: Antoniou A et al., Mathematical modelling of coupled and decoupled water electrolysis systems based on
existing theoretical and experimental studies, International Journal of Hydrogen Energy, https://doi.org/10.1016/j.ijhydene.2022.03.264
 international journal of hydrogen energy xxx (xxxx) xxx 15

Fig. 11 e Comparison of Faraday's efficiencies at different cell pressures.

Fig. 12 e Hydrogen production versus current density at different pressures and temperatures for alkaline (top) and PEM
(bottom) electrolyzers.

Please cite this article as: Antoniou A et al., Mathematical modelling of coupled and decoupled water electrolysis systems based on
existing theoretical and experimental studies, International Journal of Hydrogen Energy, https://doi.org/10.1016/j.ijhydene.2022.03.264
16 international journal of hydrogen energy xxx (xxxx) xxx

Fig. 13 e Total cell efficiency versus current density at different pressures (top) and temperatures (bottom) for all three
technologies.

pressure. From the two technologies of coupled water elec-

Conclusions trolysis studied here, alkaline and PEM, the latter features
higher total cell efficiencies. These results agree with previous
Several mathematical models for alkaline, PEM, and E-TAC ones obtained for the same range of current densities
electrolyzer cells have been discussed in this work. In partic- accounted for in this work. This outcome suggests that, in
ular, a new mathematical model capable of properly situations where cell efficiency is the main driving factor, PEM
describing the behavior of these three electrolyzer technolo- technology should be utilized. Finally, the decoupled water-
gies has been proposed here. The main requirement for its splitting technology analyzed in this work, E-TAC, has great
development was that the model is fast enough to be utilized potential to be used as an alternative to the already existing
in hydrogen plant sizing and optimization processes, conventional water electrolysis technologies. So further
involving hundreds or even thousands of model evaluations. research is needed to both understand and improve this
As such, the developed model includes mathematical ex- relatively new technology and to realize all its advantages.
pressions for electrolytic cell voltage, Faraday's efficiency, and Since only experimental results exist for E-TAC, this first
total cell efficiency. More specifically, the voltage prediction attempt to mathematically model cell voltage and Faraday's
expression involves six important parameters affecting cell efficiency in E-TAC based electrolysis systems contributes to
operational voltage, (i) current density, (ii) temperature, (iii) filling the existing gap in the associated theoretical research.
pressure, (iv) electrolyte flow rate, (v) membrane thickness,
and (vi) electrode-membrane separation. The Faraday's
efficiency-related expression incorporates, in turn, three pa- Declaration of competing interest
rameters, (i) current density, (ii) temperature, (iii) and pres-
sure. Finally, the cell total efficiency expression includes both The authors declare that they have no known competing
the cell voltage and Faraday's efficiency, so a good prediction financial interests or personal relationships that could have
of the latter two also implies a good prediction of the former. appeared to influence the work reported in this paper.
When compared to other results available in the literature,
those ones obtained using the new electrolyzer model pro-
posed here are in relatively good agreement.
From the parametric study carried out using the newly Acknowledgments
suggested model, it is observed that the cell voltage decreases
when the separation membrane thickness does so. It is This work was supported by CONCYTEC-FONDECYT (PRO-
noticed as well that Faraday's efficiency is affected by cell CIENCIA) (Peru), contract N 159-2020-FONDECYT, and the

Please cite this article as: Antoniou A et al., Mathematical modelling of coupled and decoupled water electrolysis systems based on
existing theoretical and experimental studies, International Journal of Hydrogen Energy, https://doi.org/10.1016/j.ijhydene.2022.03.264
 international journal of hydrogen energy xxx (xxxx) xxx 17

Pontifical Catholic University of Peru. The authors would like crossover. Int J Hydrogen Energy 2013;38:14921e33. https://
to thank all researchers that contributed to the development doi.org/10.1016/j.ijhydene.2013.09.013.
[17] Falca ~ o DS, Pinto AMFR. A review on PEM electrolyzer
of the previous electrolyzer models analyzed here.
modelling: guidelines for beginners. J Clean Prod
2020;261:121184. https://doi.org/10.1016/
references j.jclepro.2020.121184.
[18] Fragiacomo P, Genovese M. Modeling and energy demand
analysis of a scalable green hydrogen production system. Int
J Hydrogen Energy 2019;44:30237e55. https://doi.org/10.1016/
[1] David M, Ocampo-Martı́nez C, Sa
nchez-Pen~ a R. Advances in j.ijhydene.2019.09.186.
alkaline water electrolyzers: a review. J Energy Storage [19] Fragiacomo P, Genovese M. Developing a mathematical tool
2019;23:392e403. https://doi.org/10.1016/j.est.2019.03.001. for hydrogen production, compression and storage. Int J
[2] Klare MT. Twenty-first century energy wars: how oil and gas Hydrogen Energy 2020;45:17685e701. https://doi.org/10.1016/
are fuelling global conflicts. Energy 2014. https://energypost. j.ijhydene.2020.04.269.
eu/twenty-first-century-energy-wars-oil-gas-fuelling-global- [20] Fragiacomo P, Genovese M. Numerical simulations of the
conflicts/. [Accessed 1 February 2022]. energy performance of a PEM water electrolysis based high-
[3] Saragerova B. Greece and Turkey: energy security tensions in pressure hydrogen refueling station. Int J Hydrogen Energy
the Eastern Mediterranean. Glob Risk Insights 2020. https:// 2020;45:27457e70. https://doi.org/10.1016/
globalriskinsights.com/2020/09/greece-and-turkey-energy- j.ijhydene.2020.07.007.
security-tensions-in-the-eastern-mediterranean/. [Accessed [21] Genovese M, Fragiacomo P. Parametric technical-economic
1 February 2022]. investigation of a pressurized hydrogen electrolyzer unit
[4] Ro€ hm-Malcotti E. The EU focuses on a hydrogen economy coupled with a storage compression system. Renew Energy
and the integration of the energy system. Axpo 2021. https:// 2021;180:502e15. https://doi.org/10.1016/
www.axpo.com/lu/en/about-us/magazine.detail.html/ j.renene.2021.08.110.
magazine/renewable-energy/eu-hydrogen-strategy.html. [22] Mas R, Antoniou A, Celis C, Berastain A. A comprehensive
[Accessed 1 February 2022]. analysis of an electrolytic hydrogen production system based
[5] DOE. Announces $52.5 million to accelerate progress in clean on solar radiation for the generation of clean energy. In:
hydrogen. Dep Energy; 2021. https://www.energy.gov/ ASME 2021 International Mechanical Engineering Congress
articles/doe-announces-525-million-accelerate-progress- and Exposition; 2021. https://doi.org/10.1115/IMECE2021-
clean-hydrogen. [Accessed 1 February 2022]. 69444. Virtual, Online.
[6] Ball M, Wietschel M. Why hydrogen? Hydrog. Econ. Oppor. [23] Murthy SS, Dutta P, Rao BS, Sharma R. Performance analysis of
Challenges. 2009. https://doi.org/10.5860/choice.47-5038. a stand-alone polygeneration microgrid. Therm Sci Eng Prog
[7] Abdalla AM, Hossain S, Nisfindy OB, Azad AT, Dawood M, 2020;19:100623. https://doi.org/10.1016/j.tsep.2020.100623.
Azad AK. Hydrogen production, storage, transportation and [24] Manesh MHK, Rabeti SAM, Nourpour M, Said Z. Energy,
key challenges with applications: a review. Energy Convers exergy, exergoeconomic, and exergoenvironmental analysis
Manag 2018;165:602e27. https://doi.org/10.1016/ of an innovative solar-geothermal-gas driven polygeneration
j.enconman.2018.03.088. system for combined power, hydrogen, hot water, and
[8] Ball M, Weeda M. The hydrogen economy - vision or reality? freshwater production. Sustain Energy Technol Assessments
Int J Hydrogen Energy 2015;40:7903e19. https://doi.org/ 2022;51:101861. https://doi.org/10.1016/j.seta.2021.101861.
10.1016/j.ijhydene.2015.04.032. [25] Xu YP, Lin ZH, Ma TX, She C, Xing SM, Qi LY, Farkoush SG,
[9] Hauch A, Küngas R, Blennow P, Hansen AB, Hansen JB, Pan J. Optimization of a biomass-driven Rankine cycle
Mathiesen BV, et al. Recent advances in solid oxide cell integrated with multi-effect desalination, and solid oxide
technology for electrolysis. Science 2020;370. https://doi.org/ electrolyzer for power, hydrogen, and freshwater
10.1126/science.aba6118. production. Desalination 2022;525:115486. https://doi.org/
[10] Yodwong B, Guilbert D, Phattanasak M, Kaewmanee W, 10.1016/j.desal.2021.115486.
Hinaje M, Vitale G. Faraday's efficiency modeling of a proton [26] Shiva Kumar S, Himabindu V. Hydrogen production by PEM
exchange membrane electrolyzer based on experimental water electrolysis e a review. Mater Sci Energy Technol
data. Energies 2020;13:1e14. https://doi.org/10.3390/ 2019;2:442e54. https://doi.org/10.1016/j.mset.2019.03.002.
en13184792. [27] Falca ~ o DS, Pinto AMFR. A review on PEM electrolyzer
[11] Nikolaidis P, Poullikkas A. A comparative overview of modelling: guidelines for beginners. J Clean Prod 2020;261.
hydrogen production processes. Renew Sustain Energy Rev https://doi.org/10.1016/j.jclepro.2020.121184.
2017;67:597e611. https://doi.org/10.1016/j.rser.2016.09.044. [28] Garce
s-Pineda FA, Blasco-Ahicart M, Nieto-Castro D, Lo

pez N,
[12] Zohdy KM, Kareem MA. Hydrogen production using sea Gala
n-Mascaro
s JR. Direct magnetic enhancement of
water electrolysis. Open Fuel Cell J 2010;3:1e7. https:// electrocatalytic water oxidation in alkaline media. Nat Energy
doi.org/10.2174/1875932701003010001. 2019;4:519e25. https://doi.org/10.1038/s41560-019-0404-4.
[13] Ulleberg Ø. Stand-alone power systems for the future: [29] Symes MD, Cronin L. Decoupling hydrogen and oxygen
optimal design, operation and control of solar-hydrogen evolution during electrolytic water splitting using an
energy systems. Norwegian University of Science and electron-coupled-proton buffer. Nat Chem 2013;5:403e9.
Technology; 1998. https://doi.org/10.1038/nchem.1621.
[14] Olivier P, Bourasseau C, Bouamama PB. Low-temperature [30] Dotan H, Landman A, Sheehan SW, Malviya KD, Shter GE,
electrolysis system modelling: a review. Renew Sustain Grave DA, et al. Decoupled hydrogen and oxygen evolution
Energy Rev 2017;78:280e300. https://doi.org/10.1016/ by a two-step electrochemicalechemical cycle for efficient
j.rser.2017.03.099. overall water splitting. Nat Energy 2019;4:786e95. https://
[15] McHugh PJ, Stergiou AD, Symes MD. Decoupled doi.org/10.1038/s41560-019-0462-7.
electrochemical water splitting: from fundamentals to [31] Results of the FCH JU workshop on safety of electrolysis. Fuel
applications. Adv Energy Mater 2020;10:1e21. https://doi.org/ Cell and Hydrogen 2020. https://www.fch.europa.eu/news/
10.1002/aenm.202002453. results-fch-ju-workshop-safety-electrolysis. [Accessed 1
[16] Schalenbach M, Carmo M, Fritz DL, Mergel J, Stolten D. February 2022].
Pressurized PEM water electrolysis: efficiency and gas

Please cite this article as: Antoniou A et al., Mathematical modelling of coupled and decoupled water electrolysis systems based on
existing theoretical and experimental studies, International Journal of Hydrogen Energy, https://doi.org/10.1016/j.ijhydene.2022.03.264
18 international journal of hydrogen energy xxx (xxxx) xxx

[32] Alkaline electrolyser. EPFL. n.d. https://www.epfl.ch/labs/ aspects. Open Eng 2017;7:141e52. https://doi.org/10.1515/
lepa/grid2mobility-demonstrator-martigny/key- eng-2017-0020.
technologies/alkaline-electrolyser/. [Accessed 1 February [42] Nagai N, Takeuchi M, Nakao M. Influences of bubbles
2022]. between electrodes onto efficiency of alkaline water
[33] Seger B. Electricity to fuel. Seger Res. n.d. http:// electrolysis. In: Proceedings of PSFVIP-4; 2003. p. 3e9.
segerresearch.com/PSE/Electrolyzer.pdf. [Accessed 1 [43] Amores E, Rodrı́guez J, Carreras C. Influence of operation
February 2022]. parameters in the modeling of alkaline water electrolyzers
[34] Zeng K, Zhang D. Recent progress in alkaline water for hydrogen production. Int J Hydrogen Energy
electrolysis for hydrogen production and applications. Prog 2014;39:13063e78. https://doi.org/10.1016/
Energy Combust Sci 2010;36:307e26. https://doi.org/10.1016/ j.ijhydene.2014.07.001.
j.pecs.2009.11.002. [44] Schillings J, Doche O, Deseure J. Modeling of
[35] Tsakyridis G, Xiros NI, Sultan C, Scharringhausen M, electrochemically generated bubbly flow under buoyancy-
Vanzwieten JH. A hydrogen storage system for efficient driven and forced convection. Int J Heat Mass Tran
ocean energy harvesting by hydrokinetic turbines. In: 26th 2015;85:292e9. https://doi.org/10.1016/
Int. Ocean Polar Eng. Conf. Rhodes, Greece: OnePetro; 2016. j.ijheatmasstransfer.2015.01.121.
[36] Schalenbach M, Tjarks G, Carmo M, Lueke W, Mueller M, [45] Nagai N, Takeuchi M, Kimura T, Oka T. Existence of optimum
Stolten D. Acidic or alkaline? Towards a new perspective on space between electrodes on hydrogen production by water
the efficiency of water electrolysis. J Electrochem Soc electrolysis. Int J Hydrogen Energy 2003;28:35e41. https://
2016;163:F3197e208. https://doi.org/10.1149/2.0271611jes. doi.org/10.1016/S0360-3199(02)00027-7.
[37] Ulleberg Ø. Modeling of advanced alkaline electrolyzers a [46] Abdin Z, Webb CJ, Gray EM. Modelling and simulation of a
system. Int J Hydrogen Energy 2003;28:21e33. https://doi.org/ proton exchange membrane (PEM) electrolyser cell. Int J
10.1016/S0360-3199(02)00033-2. Hydrogen Energy 2015;40:13243e57. https://doi.org/10.1016/
[38] Antoniou A, Celis C, Berastain A. A mathematical model to j.ijhydene.2015.07.129.
predict alkaline electrolyzer performance based on physical [47] Espinosa-Lo
pez M, Darras C, Poggi P, Glises R, Baucour P,
principles and previous models reported in literature. In: Rakotondrainibe A, et al. Modelling and experimental
ASME 2021 Int. Mech. Eng. Congr. Expo. IMECE2021, Virtual. validation of a 46 kW PEM high pressure water electrolyzer.
ASME; 2021. https://doi.org/10.1115/IMECE2021-68815. Renew Energy 2018;119:160e73. https://doi.org/10.1016/
[39] Navarro RM, Guil R, Fierro JLG. Introduction to hydrogen. In: j.renene.2017.11.081.
Subramani V, Basile A, Veziroglu TN, editors. Compend. [48] Koponen J, Ruuskanen V, Hehemann M, Rauls E, Kosonen A,
Hydrog. Energy; 2015. p. 21e61. Ahola J, et al. Effect of power quality on the design of proton
[40] Schmidt O, Gambhir A, Staffell I, Hawkes A, Nelson J, Few S. exchange membrane water electrolysis systems. Appl
Future cost and performance of water electrolysis: an expert Energy 2020;279:115791. https://doi.org/10.1016/
elicitation study. Int J Hydrogen Energy 2017;42:30470e92. j.apenergy.2020.115791.
https://doi.org/10.1016/j.ijhydene.2017.10.045. [49] Huang J, Wang Y. Efficient renewable-to-hydrogen
[41] Amores E, Rodrı́guez J, Oviedo J, De Lucas-Consuegra A. conversion via decoupled electrochemical water splitting.
Development of an operation strategy for hydrogen Cell Reports Phys Sci 2020;1:100138. https://doi.org/10.1016/
production using solar PV energy based on fluid dynamic j.xcrp.2020.100138.

Please cite this article as: Antoniou A et al., Mathematical modelling of coupled and decoupled water electrolysis systems based on
existing theoretical and experimental studies, International Journal of Hydrogen Energy, https://doi.org/10.1016/j.ijhydene.2022.03.264