Journal Pre-proof

A review on PEM Electrolyzer Modelling: guidelines for beginners

D.S. Falcão, A.M.F.R. Pinto

PII: S0959-6526(20)31231-2
DOI: https://doi.org/10.1016/j.jclepro.2020.121184
Reference: JCLP 121184

To appear in: Journal of Cleaner Production

Received Date: 12 September 2019

Accepted Date: 16 March 2020

Please cite this article as: D.S. Falcão, A.M.F.R. Pinto, A review on PEM Electrolyzer Modelling:
guidelines for beginners, Journal of Cleaner Production (2020), https://doi.org/10.1016/j.jclepro.
2020.121184

This is a PDF file of an article that has undergone enhancements after acceptance, such as the
addition of a cover page and metadata, and formatting for readability, but it is not yet the definitive
version of record. This version will undergo additional copyediting, typesetting and review before it
is published in its final form, but we are providing this version to give early visibility of the article.
Please note that, during the production process, errors may be discovered which could affect the
content, and all legal disclaimers that apply to the journal pertain.

© 2019 Published by Elsevier.

 Journal Pre-proof

A review on PEM Electrolyzer Modelling:

guidelines for beginners

D.S. Falcão* and A.M.F.R. Pinto*

CEFT - Transport Phenomena Research Center

Chemical Engineering Department, Oporto University, Engineering Faculty

Rua Dr. Roberto Frias, 4200-465 Porto, Portugal

*corresponding authors, E-mail: dfalcao@fe.up.pt, apinto@fe.up.pt

Phone: +351220414857, Fax: +351225081449

Abstract

From a sustainability perspective, a synergy between hydrogen, electricity and

Renewable Energy Sources (RES) is particularly attractive. A combination of Proton

Exchange Membrane (PEM) fuel cells and PEM electrolyzers provides a back-up system

for RES avoiding the intermittency: electrolyzers convert the excess of energy from RES

into hydrogen and PEM fuel cells use this hydrogen to convert it back into electricity

when it is needed. Models are an essential tool in electrolyzer development because it

allows to understand the influence of different parameters on the electrolyzer

performance enabling an efficient simulation, design and optimization of electrolyzer

systems. Modelling studies have been developed with different degrees of complexity.

This review presents a compilation of published models focusing on the main equations

used to predict cell voltage, including reversible voltage, activation losses, ohmic losses

and mass transport losses. Special sections are devoted to dynamic behaviour, two-phase

flow effects and inclusion of thermal effects in model development. Also, a brief

1
 Journal Pre-proof

description of empirical/semi-empirical models is provided. This review aims to provide

the main guidelines for the beginners on PEM electrolyzer modelling and, for this

purpose, includes a section with the technology basic principles.

Keywords – PEM Electrolyzer, Modelling, Review.

1. Introduction

Among various alternatives to produce hydrogen, the electrolysis using renewable energy

is nowadays considered an option with potential to overcome the limitations of the

intermittency occurring on typical renewable energy sources stations such as wind parks.

Electrolyzers could convert the electricity provided from RES into chemical energy and

combined with fuel cells could transform hydrogen back into electricity. This system

should lead to an efficient storage and rapid wider diffusion of the electricity to the grid.

In particular, PEM electrolyzers have advantages when compared to the alkaline devices:

they are less caustic, can be reversible devices and are able to operate at lower cell

voltages, higher current densities, higher temperatures and pressures leading to higher

efficiencies (80-90%). The main disadvantages are: high materials cost, cross permeation

phenomena that increase with pressure and the presence of water vapor together with the

produced hydrogen, requiring dehumidification (Carmo, Fritz et al. 2013). While new

materials are developed to reduce the total cost of the electrolyzer to enable the rapid

commercialization, there is a need for further understanding of the main heat, mass and

electrochemical processes, providing realistic predictions. Mathematical models play, at

this point, an important role facilitating the dynamic connection between the electrolysis

system and an intermittent electrical source.

2
 Journal Pre-proof

In literature, models with different degrees of complexity can be found. Analytical models

are a satisfactory tool to recognize the effect of the main variables on electrolyzer

performance recurring to simplified considerations to simulate a fairly accurate

polarization curve. Empirical/semi-empirical models allow the prediction of the

electrolyzer behavior as a function of operating conditions (such as pressure, temperature)

recurring to simple empirical equations. The main disadvantage is that, generally, the

application of this kind of models is restricted to the range of operating conditions studied

and for that specific electrolyzer design. Mechanistic models use differential and

algebraic equations which are derived considering the electro-chemistry phenomena that

occur in the electrolyzer and are numerically solved by different methods. These models

involving extensive calculations accurately predict the polarization curve and the flux and

concentration of multiple species in the electrolyzer, conducing however to simulation

times considered too high for real time applications.

Regarding PEM electrolyzer modeling state of the art, the number of publications is

considerably lower when compared to the number of papers on fuel cell modeling. The

earliest models for PEM electrolyzers were reported in 2002 (Onda, Murakami et al.

2002). Since then, some works, including those taking into account the cell dynamic

behavior, were reported (Nie, Chen et al. 2009, Nie and Chen 2010, Awasthi, Scott et al.

2011, García-Valverde, Espinosa et al. 2012). Review papers considering models to

simulate eletrolyzers performance and/or specific phenomena occurring in a particular

electrolyzer component are very useful to the research community, not only by providing

an easier state-of-the art perception but also by offering a range of input parameters and

guidelines useful in the development of new models. In the last few years, some reviews

were published: a review focused on PEM water electrolysis including a section on

modeling PEM water electrolysis (Carmo, Fritz et al. 2013) and a recent one (Olivier,

3
 Journal Pre-proof

Bourasseau et al. 2017) presenting a very complete review for low-temperature

electrolysis system modeling (including alkaline and PEM technologies) with a

classification of the models analyzed based on different criteria such as the physical

domains involved or the modeling approaches. These review papers are, obviously, of

extreme importance to the research community but they lack a more comprehensive

description of the theory behind the PEM electrolyzers’ operation principle. The main

goal of the present review is to be a starting point to those interested in the PEM

electrolyzer modeling area, providing a simple compilation of the models published in

the last years, including the equations used to model cell voltage and enlightening

important aspects that should be taken into account on the development of PEM

electrolyzer models, as well as the basic principles of electrolyzer operation. A summary

of the equations used to calculate the electrolyzer voltage is presented, plus the software

used to implement the models. Because of its importance, special sections devoted to

thermal effect, dynamic behavior inclusion and two-phase flow effects, in some models,

are considered. A brief description of empirical/semi-empirical models is also provided.

2. Basic principles

As seen in Figure 1, a PEM electrolyzer is an electrochemical energy converter which

uses electricity to oxidize water, producing oxygen and protons in the anode side. As

oxygen is produced, it leaves the device while the protons pass through the membrane

and electrons circulate via an external circuit. At the cathode side, the electrons reduce

the protons, producing hydrogen. In a PEM FC the reverse happens, with hydrogen and

oxygen producing a DC electric current, water and heat. The basic design of an

electrolyzer consists of two half cells with a thin, proton conducting and electron

4
 Journal Pre-proof

insulating PEM at the center of the cell. Going outwards from each side of the membrane,

there is a porous catalyst layer on each side of the membrane, where the reactions occur.

The PEM and the double catalyst layer form the Membrane Electrode Assembly (MEA).

Enveloping the MEA, there is the current collector that physically and electrically

connects the catalyst layer to the bipolar plate. Even so, the bipolar plate is the structure

that provides physical integrity to the cell, provide pathways for products and reactants

and separate one cell from the other in a stack.

Figure 1 - Schematic representation of PEM EL and PEM FC (Lamy 2016).

The reactions occurring in a PEM electrolyzer (anodic, cathodic and global) are presented

below:

1
Anodic reaction: H 2O  2H   O 2  2e  (1)
2

Cathodic reaction: 2H  2e  H2 (2)

1
Global reaction: H 2O  H 2  O2 (3)
2

Equations (1) and (2) are generally known as Oxygen Evolution Reaction (OER) and

Hydrogen Evolution Reaction (HER), respectively. The equation (3) is the global reaction

resulting from the sum of the two electrochemical half reactions which take place at the

electrodes under acidic media and need an electrical DC power source to occur. An

electrical DC power source is connected to the electrodes and the decomposition of water

starts when a DC voltage higher than thermodynamic reversible potential is applied. The

water can be supplied to the anode or cathode side, the majority of PEM electrolyzers

operate with a water anode feed because water is consumed at this side. Under reversible

conditions (without losses), the potential difference between the anode and cathode

0
electrodes is called the reversible cell potential, Erev , corresponding to the minimal

5
 Journal Pre-proof

electrical work that is needed to split up water if the necessary contribution of thermal

energy is present. At standard state (P=1 atm and T=298.15 K) the Gibbs free energy of

the global reaction (3) is 236.483 kJ/mol (this positive value indicates that the water split

reaction is a nonspontaneous process, requiring energy from an external source).

Knowing the values of Gibbs free energy of reaction ( GR ), the Faraday constant (F)
0

and the number of moles of electrons transferred during the reaction (n) enables the

0
estimation of Erev (Bessarabov, wang et al. 2015):

GR0
0
Erev   1.229 V (4)
nF

Considering that no external heat source is available, the entire energy needed for the

reaction occur ( HR ) must be delivered by means of electrical energy. In the case of the
0

water split reaction (3), the value for HR under standard conditions is -285.83 kJ/mol.
0

0
Therefore, the voltage required to occur the referred reaction is higher than the Erev and

0
known as thermoneutral voltage at standard state ( E th ) (Bessarabov, wang et al. 2015):

HR0
Eth0   1.481V (5)
nF

Once the current passes through the cell, the actual voltage for the water splitting reaction

becomes higher that the reversible cell potential due to irreversible losses occurring in the

cell. These losses can be split into Faradaic (activation losses) and non-Faradaic losses

(ohmic and mass transfer or concentration losses). Activation losses are due to

electrochemical reaction activation: there is a shift from thermodynamic equilibrium

which reduces the velocity of the reactions taking place at the anode and cathode

electrodes surface (Bessarabov, wang et al. 2015). Some of the available cell

thermodynamic potential must be lost because the charge build-up is changed with a

6
 Journal Pre-proof

noticeable impact on the activation barriers. The activation losses (Vact) are the major

cause contributing to lower efficiency when operating a cell at high voltage and low

current density values and are governed by Faraday’s law, that state that the amount of a

gas produced by an electrochemical process can be related to the electrical charge

consumed by the cell:

Q  nzF (6)

where Q is the charge in Coulombs (C) and z is the charge number.

Considering equation (6) and Arrhenius equation, it is possible to deduce Butler-Volmer

equation, which is commonly used to describe activation overpotential, as referred in

section 3.1.2. All the considerations and equation developments are very well explained

and can be found in Bessarabov, Wang et al. (2015).

In the middle of the operating range the predominant losses are due to the electrolyzer

cell resistance (Rcell) generated by the ionic and electronic conduction. The total ohmic

losses are determined by the application of Ohm’s law, as referred in section 3.1.4.

Usually, due to the high conductivity of commonly used materials (as titanium), it is often

considered that electron transport is much faster than protonic transport and, therefore,

only ohmic losses due to proton transport through the PEM are taken into account.

At very high current densities, the major losses (the concentration losses or diffusion

overpotential, Vdiff) are due to mass transport limitations and are not considered by many

authors since electrolyzers usually do not operate at high current densities. However,

some works (Hwang, Lai et al. 2009, Lebbal and Lecœuche 2009, Marangio, Santarelli

et al. 2009, Kim, Park et al. 2013) considered these losses. At high current densities

(higher than 2 A/cm2) the gas bubbles produced can block the active area, damage the

contact between the electrode and the electrolyte and decrease the catalyst utilization.

The mass transport losses are governed by Nernst equation stating that the overpotential

7
 Journal Pre-proof

due to mass transport limitation increases with the increase in the product species

concentration at the reaction interface (Bessarabov, wang et al. 2015), as referred in

section 3.1.3. Figure 2 depicts an example of a PEM electrolyzer polarization curve with

the contribution of the three losses referred above: activation, ohmic and mass transport

losses.

Figure 2 – Typical PEM electrolyzer polarization curve with the contribution of activation

and ohmic and mass transport losses.

3. Models review

3.1 Voltage

The most common representation of the electrolyzer performance is the polarization

curve that represents the relation between the current density and the voltage.

Nafeh (2011) modeled the electrolyzer voltage (V) using the following expression:

𝑉 = 𝑁𝑐(𝐸𝐶𝑒𝑙𝑙 + 𝑉𝐴𝑐𝑡,𝑐 + 𝑉𝐴𝑐𝑡, 𝑎 + 𝑖𝑅𝑐𝑒𝑙𝑙) (7)

where Nc is the number of electrolyzer cells, Ecell is the open circuit voltage VAct,a and

VAct,c are the anode and cathode activation overpotentials, i is the current density and Rcell

is the electrolyzer cell resistance (ohmic losses). Similar expressions are used by almost

all the authors (Harrison, Hernández-Pacheco et al. 2005, Görgün 2006, Brown, Brouwer

et al. 2008, Ni, Leung et al. 2008, Awasthi, Scott et al. 2011, García-Valverde, Espinosa

et al. 2012, Myles, Nelson et al. 2012, Aouali, Becherif et al. 2014, Chandesris, Médeau

et al. 2015). This expression excludes the mass transport losses since, as referred before,

electrolyzers usually do not operate at high current densities.

An empirical equation to determine the electrolyzer cell voltage is presented by Atlam

and Kolhe (2011) and another is used by Sassan, Bindner et al. (2014):

8
 Journal Pre-proof

𝑉 = 0.326𝐼 + 1.476 (8)

and

𝑉 = 33.1 ― 43.0 × 10 ―3𝐼 (9)

Where I is the electrolyzer current.

Considering the expressions used for the electrolyzer voltage calculation, it seems that

almost all the authors used the same expression, varying only the importance given to the

losses occurring in the electrolyzer with most of the authors neglecting the concentration

losses. Some empirical expressions can be found and should be applied carefully because

they are developed for a specific electrolyzer design and a particular range of operating

conditions.

2 3.1.1 Open circuit voltage

4 The open circuit voltage is usually determined using the Nernst equation. This equation,

5 or very similar ones, is used in many works reported in literature (Han, Mo et al. 2016,

6 Yigit and Selamet 2016, Ruuskanen, Koponen et al. 2017, Moradi Nafchi, Afshari et al.

7 2019, Toghyani, Fakhradini et al. 2019):

𝑅𝑇
[(
𝐸 𝑐𝑒𝑙𝑙 = 𝐸 0𝑟𝑒𝑣 + 2𝐹 𝑙𝑛
𝑃 𝐻 2𝑃 𝑂 21/2
𝑃 𝐻 2𝑂 )] ,
(10)

8 where P is the partial pressure of reactants/products, T is the temperature

9 , F is the Faraday constant and 𝐸 0𝑟𝑒𝑣 is the reversible cell potential at standard temperature

10 and pressure. It is calculated by equation (4), as referred in section 2, and is used by

11 several authors (Görgün 2006, Brown, Brouwer et al. 2008, Aouali, Becherif et al. 2014,

12 Abdol Rahim, Tijani et al. 2015, Saeed and Warkozek 2015, Mohamed, Alli et al. 2016,

13 Yigit and Selamet 2016, Aouali, Becherif et al. 2017). Some authors considered empirical

14 temperature dependent expressions (Dale, Mann et al. 2008, Awasthi, Scott et al. 2011,

9
 Journal Pre-proof

1 Abdin, Webb et al. 2015, Han, Mo et al. 2016, Ruuskanen, Koponen et al. 2017, Moradi

2 Nafchi, Afshari et al. 2019):

E0rev = 1.5241 ― 1.2261 × 10 ―3𝑇 + 1.1858 × 10 ―5𝑇𝑙𝑛(𝑇) + 5.6692 × 10 ―7𝑇2 (11)

E0rev = 1.229 ― 0.9 × 10 ―3(𝑇 ―298) (12)

𝑅𝑇
E0rev = 1.229 ― 0.9 × 10 ―3(𝑇 ― 298) + 2.34𝐹𝐿𝑜𝑔(𝑃2𝐻2𝑃𝑂2). (13)

3 In Garcia-Valverde, Espinosa et al (2012) the open circuit potential is determined by an

4 empirical equation for atmospheric pressure condition, depending from temperature,

5 expression also in others work (Agbli, Péra et al. 2011, Kaya and Demir 2017):

𝐸𝑐𝑒𝑙𝑙 = 1.5184 ― 1.5421 × 10 ―3𝑇 + 9.523 × 10 ―5𝑇𝑙𝑛(𝑇) + 9.84 × 10 ―8𝑇2 (14)

6 Kim, Park et al. (2013) introduced a new parameter, k, in Ecell calculation, as follows:

∆𝐺
𝐸𝑐𝑒𝑙𝑙 = ―(1 ― 𝑘) (15)
2𝐹

7 where ∆𝐺 is the Gibbs energy variation and k the fraction of proton back-permeation

8 relative to total proton permeation. The term 1-k is included to take into account the effect

9 of hydrogen permeation and proton back-permeation.

10 Since non-standard pressure and temperature conditions are generally used, other

11 equations must be applied (more details can be found in Marangio, Santarelli et al.

12 (2009)). The equations reported in this work enable the consideration of different

13 temperatures at the anode and cathode sides, leading to more accurate values of 𝐸 0𝑟𝑒𝑣.

14

15 Some equations are available to use for open circuit voltage and reversible potential

16 calculations. If the electrolyzer works at standard temperature and pressure, simpler

17 equations could be used with no significant errors associated. However, for different

18 temperatures and higher pressures the use of more accurate equations is recommended.

19

10
 Journal Pre-proof

1 3.1.2 Activation Overpotential

2 Activation losses occur, as referred, because of the necessity to sacrifice some potential

3 to activate the electrochemical reactions taking place at anode and cathode sides, as

4 described in Section 2.

5 One of the more common expressions for activation overpotential calculations is based

6 in the Butler-Volmer equation and is used by many authors (Marangio, Santarelli et al.

7 2009, Agbli, Péra et al. 2011, Awasthi, Scott et al. 2011, Kim, Park et al. 2013, Abdin,

8 Webb et al. 2015, Yigit and Selamet 2016, Ruuskanen, Koponen et al. 2017, Sartory,

9 Wallnöfer-Ogris et al. 2017, Moradi Nafchi, Afshari et al. 2019) on electrolyzer model

10 development:

𝑅𝑇𝑎 𝑅𝑇𝑐
𝑉𝐴𝑐𝑡 =
𝛼𝑎𝐹
𝑎𝑟𝑐𝑠𝑖𝑛ℎ( ) 𝑖
2𝑖0,𝑎
+
𝛼𝑐𝐹
𝑎𝑟𝑐𝑠𝑖𝑛ℎ( )𝑖
2𝑖0,𝑐
(16)

11 where i0,a and i0,c are the exchange current density at anode and cathode, respectively and

12 𝛼𝑎 and 𝛼𝑐 are the charge transfer coefficients at anode and cathode, respectively.

13 Some authors (Choi, Bessarabov et al. 2004, Harrison, Hernández-Pacheco et al. 2005,

14 Biaku, Dale et al. 2008, Dale, Mann et al. 2008, Ni, Leung et al. 2008, Aouali, Becherif

15 et al. 2014, Aouali, Becherif et al. 2017) used a similar expression but with a factor of 2

16 coupled to the charge transfer coefficient, as follows:

𝑅𝑇𝑎 𝑅𝑇𝑐
𝑉𝐴𝑐𝑡 =
2𝛼𝑎𝐹
𝑎𝑟𝑐𝑠𝑖𝑛ℎ
𝑖
( )
2𝑖0,𝑎
+
2𝛼𝑐𝐹
𝑎𝑟𝑐𝑠𝑖𝑛ℎ
𝑖
2𝑖0,𝑐 ( ) (17)

17 Other authors used a simplified expression (Görgün 2006, Lebbal and Lecœuche 2009,

18 Nafeh 2011, Abdol Rahim, Tijani et al. 2015, Saeed and Warkozek 2015):

𝑉𝐴𝑐𝑡 =
𝑅𝑇
𝑙𝑛
𝑛𝛼𝐹 𝑖0
𝑖
() (18)

19 This equation can be applied for both sides, anode and cathode (García-Valverde,

20 Espinosa et al. 2012). The values reported in literature for the exchange current density

11
 Journal Pre-proof

1 coefficients are very dissimilar. For this reason, some authors decided to choose the

2 values that better adjust their models. These values increase with temperature, therefore,

3 an expression relating it with the temperature is very useful and is presented by Garcia-

4 Valverde, Espinosa et al. (2012) applying an Arrhenius expression, as follows:

[ (
𝑖0 = 𝑖0,𝑟𝑒𝑓𝑒𝑥𝑝 ―
𝐸𝑎𝑐𝑡 1
―
𝑅 𝑇 𝑇𝑟𝑒𝑓
1
)] (19)

5 where Eact is the activation energy of the electrode and i0,ref is the exchange current density

6 measured at a reference temperature Tref.

7 Regarding the charge transfer coefficient (CTC), the most part of the authors consider it

8 constant with temperature, however, some works presented an estimation of the CTC

9 depending on the operating temperature. Biaku, Dale et al. (2008) reported that there is a

10 reasonable variation of the anode CTC from the called symmetry factor (0.5 – the most

11 common value used in the literature). The authors noticed that the average anode CTC

12 changes from 0.18 to 0.42 within a temperature range of 20-60 ºC. Tijani, Binti

13 Kamarudin et al. (2018) simulated the CTC values (but the model was successfully

14 validate with experimental data). The simulation results show that as the electrolyzer

15 operating temperature increases (range 10-90ºC), the values of CTC also increase but this

16 increment is more pronounced at the anode (ranges between 0.807 and 1.035) compared

17 to cathode (ranges beteween 0.202 and 0.259) electrodes. The authors also observed that

18 activation overvoltage decreases when the CTC increases from 0.1 to 2.0 both at anode

19 and cathode electrodes. Curiously the pressure doesn’t seem to affect CCT values,

20 remaining the same even at balanced and unbalanced pressure. Touré, Konaté et al.

21 (2018) estimated CTC first from the current-voltage characteristic of a fuel cell, by using

22 only the activation polarization and found a value of 0.669. Then the Lagrange’s

23 undetermined multiplier technique was used to evaluate the transfer coefficient and was

12
 Journal Pre-proof

1 found a value of 0.222. The authors believe that the Lagrange’s multiplier technique (that

2 takes into account the ohmic polarization, the activation polarization and also the current-

3 power characteristic of the device) can be a better way to estimate of the charge transfer

4 coefficient.

5 Table 1 displays the values reported in the literature for exchange current density for

6 anode and cathode reactions and also for the charge transfer coefficients for both sides.

8 To describe the activation losses, almost all the authors used an expression based on

9 Butler-Volmer equation. However, the formulation of this equation is not consensual;

10 some authors used a factor of 2 coupled to the charge transfer coefficient. The values for

11 the charge transfer coefficient must be carefully chosen taking into account this

12 consideration. The values reported for exchange current density are very different; the

13 values are generally higher for the cathode side, indicating that the reaction kinetics is

14 faster than for the anode side. The most common value chosen for charge transfer

15 coefficients is 0.5. Some authors consider this coefficient temperature dependent.

16

17 3.1.3 Mass transport overpotential

18

19 The mass transport losses occur when the current density is high enough to impede the

20 access of reactants to active sites by the overpopulation of reacting molecules and for this

21 reason slowing down the reaction rate (Biaku, Dale et al. 2008). In PEM electrolyzer,

22 usually, this behavior is not observed until moderate operating current densities (1.6

23 A/cm2). Accordingly, the current should be limited in order to reach high efficiencies

24 (Barbir 2005).

13
 Journal Pre-proof

1 The mass transport overpotential (also usually called as diffusion overpotential – VDiff)

2 can be estimated using the Nernst equation as reported in Marangio, Santarelli et al.

3 (2009):

𝑅𝑇 𝐶
𝑉𝐷𝑖𝑓𝑓 = 𝑙𝑛 (20)
𝑛𝐹 𝐶0

4 where C is the concentration of oxygen or hydrogen at the interface membrane-electrode

5 and C0 is a working concentration taken as the reference concentration. This equation can

6 be applied for both anode and cathode sides (Agbli, Péra et al. 2011, Kim, Park et al.

7 2013, Abdin, Webb et al. 2015, Han, Mo et al. 2016, Aouali, Becherif et al. 2017, Moradi

8 Nafchi, Afshari et al. 2019, Toghyani, Fakhradini et al. 2019).

9 Saeed and Warkozek (2015) used an expression based on ilim, the maximum current

10 density that can be provided to the electrolyzer, so it reveals the maximum production

11 rate permitted by the electrolyzer:

𝑉𝐷𝑖𝑓𝑓 = ―
𝑅𝑇
𝑛𝐹 (
𝑙𝑛 1 ―
𝑖
𝑖𝑙𝑖𝑚 ) (21)

12 Mohamed, Alli et al. (2016) used a very similar expression but with a multiplicative factor

13 on denominator, a, defined as the transfer coefficient of diffusion layer.

14

15 The mass transport losses are not relevant until moderate operating current densities.

16 Under current densities of 1.6 A/cm2, these losses can be neglected with no significant

17 errors associated to voltage prediction.

18

19 3.1.4 Ohmic losses

20

21 The ohmic overpotential is related with the materials resistance to the protons flux. The

22 magnitude of ohmic losses depends on the materials properties. Manufacturing

14
 Journal Pre-proof

1 techniques and processes are a significant aspect to reduce this overpotential (Biaku, Dale

2 et al. 2008).

3 The standard Ohm’s law is applied in all the works analyzed (Choi, Bessarabov et al.

4 2004, Harrison, Hernández-Pacheco et al. 2005, Biaku, Dale et al. 2008, Brown, Brouwer

5 et al. 2008, Dale, Mann et al. 2008, Ni, Leung et al. 2008, Lebbal and Lecœuche 2009,

6 Agbli, Péra et al. 2011, Awasthi, Scott et al. 2011, Nafeh 2011, Kim, Park et al. 2013,

7 Aouali, Becherif et al. 2014, Abdin, Webb et al. 2015, Abdol Rahim, Tijani et al. 2015,

8 Saeed and Warkozek 2015, Yigit and Selamet 2016, Ruuskanen, Koponen et al. 2017):

𝛿
𝑉𝑂ℎ𝑚 = 𝑅𝐼 = 𝐼 (22)
𝜎

9 where 𝛿 is the material thickness and 𝜎 the material conductivity. This equation is usually

10 applied to the membrane but can also be applied to diffusion and catalytic layers. Sartory,

11 Wallnofer-Ogris et al. (2017), in their semi-empirical high pressure PEM water

12 electrolyzer model, added a multiplier term (a(T,P)) to conductivity to enable the

13 membrane conductivity calculation at different pressures and temperatures. More details

14 are given in the referred work.

15 The membrane conductivity, considering Nafion, can be expressed as (Görgün 2006,

16 Lebbal and Lecœuche 2009, Awasthi, Scott et al. 2011, Abdin, Webb et al. 2015, Abdol

17 Rahim, Tijani et al. 2015, Saeed and Warkozek 2015, Yigit and Selamet 2016):

𝜎 = (0.005139 - 0.00326)𝑒𝑥𝑝 1268 [ ( 1

―
303 T
1
)] (23)

18 where T is the electrolyzer temperature and  is the membrane water content.

19 There is a consensus in the use of Ohm’s law to describe the ohmic losses. This equation

20 is commonly used to take into account the membrane resistance (which is the most

21 significant) as well as the diffusion and catalytic layers’ resistances.

22

15
 Journal Pre-proof

1 3.2 Dynamic Behaviour

2

3 The connection of the PEM electrolyzer to renewable intermittent and variable power

4 sources is of crucial importance and, therefore, the consideration of dynamic behavior in

5 the models has special interest. This inclusion involves a significant increase in the model

6 complexity with the need of resolution of systems of Ordinary Differential Equations

7 (ODE) or even Partial Differential Equations (PDE) for system control analysis in the

8 first case and for sizing, chemical and thermal process design and analysis in the latter

9 (Olivier, Bourasseau et al. 2017).

10 Kim, Park et al. (2013) introduced the dynamic effects by considering unsteady mass and

11 energy balance equations for the two electrolyzer channels and the MEA separately. A

12 simplification was proposed by assuming that the water and gas flow simultaneously at

13 the same speed. The non-ideal gas behavior of high-pressure hydrogen was represented

14 by a compressibility factor. The dynamic responses were numerically studied for changes

15 in the average current density and in inlet water flow rates. However the model was not

16 validated with experimental data.

17 Lebbal and Lecoeuche (2009) developed an electrolyzer model split into two parts:

18 electrical and thermal. The electrical model was developed for steady state conditions.

19 The thermal model includes the dynamic temperature behavior for current and voltage.

20 The electrical and thermal model parameters were predicted using different techniques.

21 For the thermal model, the parameters were identified using the properties of a first order

22 linear model. The model was used to develop monitoring algorithms. In particular, the

23 monitoring of the temperature sensor allowed avoiding membrane hot point and tear

24 providing an important tool to guarantee the electrolyzer safety control.

25 Garcia-Valverde, Espinosa et al. (2012) developed a semi-empirical model considering

26 dynamic behavior in the H2 production submodel and in the thermal submodel. The

16
 Journal Pre-proof

1 authors obtained the theoretical H2 flowrate from the Faraday’s law. The real H2 flowrate

2 at steady state was then computed incorporating the several losses and the dynamic

3 response was estimated using a first order response system. The submodel parameters

4 could be identified from a current-step response, measuring the time to reach the steady

5 state in the hydrogen output. The thermal submodel also included the dynamic response

6 for changes in temperature. Both the dynamic submodels were validated with

7 experimental data and strengths and limitations of the modelling approach were

8 presented.

9 The dynamic model of Görgün (2006) considered four ancillaries: anode, cathode,

10 membrane and voltage. Each ancillary dynamics and interaction between them was taken

11 into account. Additionally, a storage ancillary was considered. The model was developed

12 in Simulink. The results confirmed that the model can simulate correctly the transient

13 behavior of the PEM electrolyzer. However, the model was not validated against

14 experimental data. Awasthi, Scott et al. (2011) developed a similar model, with the same

15 four blocks in Simulink. This model was validated with experimental results and reaching

16 a good agreement. A similar approach was also followed by Yigit and Selamet (2016).

17 The model was tested against dynamic changes and it responded with fast outputs. The

18 loss of each system component at different current densities was considered in the

19 numerical simulations. The results showed that the loss of the stack dominated the losses

20 of other components in the region of high current densities due to the effect of the

21 increasing pressure. The model was experimentally validated.

22 Agbli, Péra et al. (2011) used a model based in Energetic Macroscopic Representation

23 (EMR) and described the dynamic evolution of the temperature in the stack and in the

24 water tank. The thermal model parameters were obtained using experimental fitted

25 values. Some of these parameters are specific of the experimental apparatus and

17
 Journal Pre-proof

1 experimental conditions reported by the authors. The model was also implemented in

2 Simulink environment and experimental and simulated curves considering voltage

3 responses and temperature evolutions showed good agreement against experimental data.

4 Oliveira, Jallut et al. (2013) developed a physical multiscale and transient model for PEM

5 electrolyzer anodes. This model presented results for a large set of operating parameters

6 and material parameters in the anode modelling. The model development was based on

7 an elementary kinetics approach alternative to the Butler-Volmer equation. The model

8 results adjusted well both the in-house obtained experimental data and the published

9 available experimental results.

10

11 Some examples of PEM electrolyzer models considering dynamic behavior are available;

12 however, only a few are validated against experimental data. The model validation is an

13 important issue to confer credibility to the developed models for the prediction of

14 electrolyzer performance. Unlike what has been done for fuel cell systems, very few PEM

15 electrolysis modelling works include input-output models which are suitable for diagnosis

16 studies such as for example, degradation mechanisms.

17
18
19 3.3 Thermal effects
20

21 The consideration of thermal effects in the development of electrolyzer models is

22 essential since it enables to describe the temperature distribution in the cell and its impact

23 on efficiency and durability. Most of the available models are based on the analytical

24 characterization of heat accumulation, production, exchange and transport terms together

25 with the resolution of thermal balances. Some examples are here reported.

18
 Journal Pre-proof

1 Lebbal and Lecoeuche (2009) considered, in their model development, four major heat

2 contributions: chemical reaction (entropy), chemical component thermodynamics (gases

3 and water), external ambient temperature and Joule effect due to the current circulation.

4 The thermal model produced the dynamic temperature impact for both current and

5 voltage. The thermal model parameters were predicted using the properties of a first order

6 linear model based on experimental measurements. As referred in Section 2.2, the model

7 was used to develop monitoring algorithms.

8 The thermal submodel of Garcia-Valverde, Espinosa et al. (2012) based on the lumped

9 thermal capacitance which is the overall thermal capacity of the electrolyzer. This

10 important parameter was estimated from an experimental set of heating curves, thermally

11 insulating the stack and turning off the auxiliary cooling system. A statistical comparison

12 between experimental and simulated data concerning dynamic operating temperature

13 calculation was done and good agreement was obtained. However, for other experiments

14 some limitations were founded, namely, large errors were obtained for high current

15 densities values

16 Kim, Park et al. (2013) included in their model the development the unsteady energy

17 balance equations and obtained the temperature profiles after implementation and

18 resolution. A linearly increasing distribution along the anode flow direction was found:

19 the MEA was acting as heat source and the anode flow as the major heat carrier. A

20 weakness of this model is the necessity of validation with experimental data.

21 Agbli, Péra et al. (2011) developed a thermal model with two degrees of freedom: the

22 temperature along the electrolyzer and the water tank temperature. The thermal balance

23 was established considering all heat sources and sinks (electrical energy released by

24 overpotentials, heat flow evacuated from the environment and received by the water tank

25 and the heat exchange between the device and the surrounding atmosphere). The model

19
 Journal Pre-proof

1 parameters were obtained using experimental values, literature values and water tank

2 measures. The stack and tank predicted temperature evolutions showed notable

3 agreement with the experimental data.

4 A similar approach was followed by Aouali, Becherif et al. (2017), incorporating the

5 thermal effect in the development of a model to simulate a PEM electrolyzer for hydrogen

6 production. The thermal model included the heat exchange between the device and the

7 surrounding atmosphere, the water reservoir thermal energy accumulation, the entropy

8 flow of both heat sources and sinks and the heat losses by conduction. The model was

9 successfully validated against experimental data from both the PEM electrolyzer stack

10 and the water reservoir temperature profiles.

11 Toghyani, Afshari et al. (2018) conducted a non-isothermal three-dimensional

12 Computational Fluid Dynamics (CFD) numerical study to investigate the distribution of

13 current density, temperature and pressure drop on an innovative spiral pattern anode and

14 cathode flow fields. The model was able to provide precise data for mass and heat transfer,

15 electrodes kinetics, and potential fields within the PEM electrolyzer. The model was

16 validated against experimental data with a satisfactory agreement at low to medium

17 currents densities. The model results showed that the spiral flow field conducted to a

18 uniform distribution of produced hydrogen and current density and to a uniform

19 distribution of temperature. However, non-uniformities were found at 90º bends due to

20 stagnation of water and decrease in its velocity.

21

22 In spite of its importance, thermal effects are not included in the majority of the models

23 available in literature, in particular those including the thermal architecture of the

24 electrolysis system, fundamental to gain insight on the stack thermal behavior.

25

20
 Journal Pre-proof

1 3.4 Empirical and Semi-Empirical Models

2
3 Among the simpler approaches in electrolysis modelling the most common is the

4 development of empirical and semi-empirical models. The first depends on parametric

5 equations obtained by the mathematical handling of experimental data and generate

6 parameters which do not have a physical meaning. The latter models are based on basic

7 physical laws with physically meaningful parameters even though some or most of them

8 are obtained from the use of empirical correlations. Some selected examples are next

9 reported.

10 Atlam and Kolhe (2011) developed an electrical equivalent model for a PEM electrolyzer

11 using the experimental results under steady-state conditions following the Faraday’s Law.

12 The model adjusts well the experimental results in the active region and produces an

13 estimation of the hydrogen production rate, the power-hydrogen production rate and

14 practical electrolyzer efficiencies. This model is very useful to analyze the temperature

15 and pressure effects on the electrolysis efficiency. The conductivity of the PEM

16 electrolyzer cell is superior using higher temperatures and lower pressures. The pressure

17 effects on the voltage are logarithmic. Although the model is empirical, it can be applied

18 to different sizes of active area in addition to different parallel/series combinations of

19 cells.

20 Santarelli, Medina et al. (2009) presented a useful method for the design of experiments

21 that produces first-order regression models for the investigated dependent variables in

22 terms of significant parameters. The results of the Yate’s technique experimental design

23 have been reported. Two analyses have been done at extreme stack operating conditions:

24 0.1 and 1 A/cm2. This approach enables the establishment of analytical relations between

25 the dependent variables (the stack voltage) and the explored independent variables

26 (pressure, temperature, water flow and current). The authors concluded that pressure and

21
 Journal Pre-proof

1 temperature have a predominant effect on the electrolyzer performance while the voltage

2 remains almost invariable with the water flow.

3 Biaku, Dale et al. (2008) developed a semi-empirical study regarding the temperature

4 dependence of the anode charge transfer coefficient of a PEM electrolyzer stack. The

5 charge transfer coefficient is a very significant model parameter to predict the current-

6 voltage characteristics of electrolyzers providing insight into the properties of the

7 electrode. The model results indicate that the anode charge transfer coefficient changes

8 from 0.18 to 0.42 in the temperature range of 20 – 60 ºC which is noticeably lower than

9 the standard symmetry factor of 0.5.

10 Dale, Mann et al. (2008) used a semi-empirical equation to describe the performance of

11 a 6 kW PEM electrolyzer stack. The authors considered the temperature dependence in

12 reversible potential calculations. The experimental I-V curve obtained was predicted and

13 nonlinear curve fitting algorithms were used to extract model critical parameters like

14 anode and cathode exchange current densities and membrane conductivity.

15 As referred in Section 3.3, Lebbal and Lecoeuche (2009) developed a thermal model

16 producing the model parameters using the properties of a first order linear model based

17 on experimental measurements.

18 Harrison, Hernández-Pacheco et al. (2005) developed a semi-empirical equation to model

19 the behavior of a 20-cell PEM electrolyzer stack. The model parameters, such as exchange

20 current density and membrane conductivity were obtained from experimental data using

21 a nonlinear curve fitting method.

22 Sartory, Wallnöfer-Ogris et al. (2017) presented a semi-empirical zero-dimensional

23 steady state model of an asymmetric high pressure PEM water electrolyzer where

24 empirical parameters were established taking into account the experimental investigations

25 on a 9.6 kW electrolyzer. The referred model is able to predict the stack voltage, gas

22
 Journal Pre-proof

1 production flow rates, water consumption and stack efficiency as function of input

2 current, process temperature and production pressure. The results showed a satisfactory

3 uniformity of measurements and simulations.

4 Becker and Karri (2010) developed predictive models for PEM electrolyzer system

5 efficiency and hydrogen flow rate using neural network based Adaptive Neuro-Fuzzy

6 Inference Systems. These models showed a very good agreement (accuracy of +/- 3%)

7 when compared with experimental data. The utilization of these models could be of

8 special interest as generic virtual sensors for wide safety and monitoring applications.

9 Medina and Santarelli (2010) studied the water transport trough the membrane under

10 different operating conditions. Based on experimental data obtained for the PEM

11 electrolyzer, the Design of Experiments (DoE) methodology is used to obtain a regression

12 model for the net electro-osmotic drag coefficient. The goal was to find the optimal

13 operating conditions to reduce the amount of water driven from the anode side to the

14 cathode side and, therefore, reduce the amount of water that has to be separated from the

15 produced hydrogen.

16

17 There are several empirical/semi-empirical models available to simulate PEM

18 electrolyzer performance. These models can be easily implemented to obtain polarization

19 curves in real time calculations. However, a careful selection must be performed to pick-

20 up those corresponding to the users´ desired operating conditions and electrolyzer

21 design.

22

23 3.5 Two-phase flow effects

24

23
 Journal Pre-proof

1 In PEM electrolyzers, oxygen evolution at the anode and flooding due to water crossover

2 at the cathode yield different two-phase transport conditions, which strongly affect

3 performance. Unfortunately, there are only a few models which deal with these issues.

4 Arbabi, Montazeri et al. (2016) developed a 3D, two-phase flow model to investigate the

5 oxygen bubble transportation through the GDLs of a PEM electrolyzer. The model used

6 the Volume-of-Fluid (VOF) technique to simulate the gas-liquid interface through liquid-

7 saturated porous media. The authors successfully validated their model with previous

8 experimental microfluidic investigations. The model was used to calculate pressure

9 variations in bubbles during propagation and the maximum threshold capillary pressure

10 was suggested as a pointer of oxygen bubble removal effectiveness. The model can be

11 used to help on designing and evaluating GDL microstructures for next generation

12 electrolyzer materials.

13 Nie, Chen et al. (2008) presented a 3D CFD model considering two-phase flow

14 (oxygen/water) effects in a basic anode bipolar flowfield of a PEM electrolysis cell

15 aiming to examine flow characteristics in a bipolar plate and inside electrolyte cells. The

16 authors used the mixture model to simulate the two phases solving the momentum and

17 continuity equations for the mixture, the volume fraction equations for the secondary

18 phase and algebraic expressions for the relative velocities. The model was only validated

19 with the measurements of the pressure drop for single phase fluid dynamics, which

20 represents a limitation of the referred work. The authors performed simulations with an

21 oxygen bubble rate of 0 to 0.014 g/s and the results revealed that the average volume

22 fraction of oxygen at the exit is higher when the mass flow rate of the oxygen bubbles

23 increases.

24 Han, Mo et al. (2016) developed a two-phase transport model to study the liquid water

25 distribution into the liquid/gas diffusion layer (LGDL) and an electrochemical

24
 Journal Pre-proof

1 performance model coupled with an equivalent resistance model. The model was

2 successfully validated by comparison of predicted polarization curves with experimental

3 results. However, the authors used the exchange current densities parameters values and

4 membrane humidification degrees that better fit their experimental data. The effects of

5 LGDL properties (contact angle, porosity and pore size) on the two-phase transport

6 dynamics were studied and the liquid saturation distribution inside the LGDL was

7 determined. Additionally, the influence of two-phase transport on the performance and

8 efficiency of a PEM electrolyzer was examined. As main conclusions the authors

9 reported that the liquid water saturation along the LGDL thickness direction is enhanced

10 (with a consequent improving on cell performance) using higher porosities and/or pore

11 size or using lower contact angles.

12 Lee, Lee et al. (2019) used pore network modelling to predict the saturations of oxygen

13 gas and liquid water and the resultant permeability in sintered titanium powder-based

14 porous transport layers (PTLs). The authors studied the influence of PTL-catalyst coated

15 membrane (CCM) contact angle, pore and throat sizes and porosity on two-phase flow

16 transport under different operating conditions. As main conclusion, the authors strongly

17 recommend the use of a backing layer, or a micro-porous layer, in order to enhance mass

18 transport as well as interfacial contact.

19 Briguglio, Brunaccini et al. (2013) developed a hydrodynamic study in an electrolyzer

20 stack (with 10 cells). The study focuses on diffusional aspects concerning water mass

21 distribution along the channels and was investigated using a CFD program. The model

22 showed a reasonably uniform distribution of water along the channels and a homogeneous

23 pressure inside the cells. A uniform water distribution in each cell allows avoiding gas

24 accumulation inside the stack and hot spots formation that could damage the MEA. This

25
 Journal Pre-proof

1 type of analysis is very useful to identify operating conditions that conducts to a good

2 water distribution in cells channels.

3 Myles, Nelson et al. (2012) developed a species transport model within a high pressure

4 oxygen-generating PEM water electrolyzer using the dusty-fluid model (DFM) when

5 liquid water was fed to the cathode. Simulation results were validated with performance

6 data obtained for Hamilton Sundstrand’s high pressure oxygen generating assembly

7 (HPOGA) at varying differential oxygen pressures. The dehydration of the cell was found

8 to occur at a Damkohler number of 0.196. It was concluded that the electro-osmotic drag

9 has a notable influence because of its direct and substantial effect on the necessary water

10 flux for a given operating current density to avoid membrane dehydration.

11

12 3.6 Software used

13

14 The model equations usually need to be solved by using software stronger than Excel.

15 Some authors do not provide any information on the software used. A compilation of the

16 software’s referred by some works to implement the developed models is presented in

17 Table 2.

18 The software more extensively used to implement empirical models is Mathematica, while

19 the majority of authors used Matlab to execute analytical/mechanistic models.

20

21

22

23

24

25

26 Conclusions

26
 Journal Pre-proof

2 This review compiles the PEM electrolyzer models developed during the last years and

3 presents the technology basic principles. Particular emphasis on the equations used to

4 predict cell voltage, including open circuit voltage, activation losses, ohmic losses and

5 mass transport losses is provided. Special sections are devoted to dynamic behavior,

6 thermal effects and two-phase flow inclusion in model development. A brief description

7 of empirical/semi-empirical models is also presented. Although the total number of

8 articles published with PEM electrolyzer models is much lower than the total amount of

9 papers devoted to PEM fuel cells models, in the last five years, interest in PEM

10 electrolysis has increased and is expected to increase significantly in the next decade.

11 The models published are mainly empirical/semi-empirical and analytical models that can

12 predict easily the PEM electrolyzer performance. Some of these models were not

13 validated with experimental data, which is a handicap for its use to simulate PEM

14 electrolyzer performance. Taking into account the main application of PEM electrolyzers

15 (as a back-up system together with fuel cells to avoid RES intermittency) the dynamic

16 behavior is a very important aspect to take into account on PEM electrolyzer model

17 development but a relatively low number of papers include dynamic responses to voltage

18 perturbations. Also, the description of two-phase flow effects (water and gas dynamics in

19 the channels) is rare in CFD models used to simulate the PEM electrolyzer performance,

20 mainly due to the corresponding extremely high simulation times. However, the inclusion

21 of two-phase flow effects on PEM electrolyzer’s model simulation is very important to

22 simulate accurately its performance as well as to be possible to suggest new bipolar plates

23 configurations as well as the most suitable materials to PTLs. Other issue that should be

24 included in PEM electrolyzer models development is the performance degradation due to

25 components ageing. So far there are no studies on this subject, crucial to develop PEM

27
 Journal Pre-proof

1 electrolyzers with a higher durability and, consequently, with a lower cost considering its

2 useful lifetime, which will certainly increase the TRL.

3 This review aims to be useful for the beginners on PEM electrolyzer modeling area,

4 providing a compilation of the models published in the last years, including the equations

5 used to model cell voltage and enlightening important aspects that should be taken into

6 account on PEM electrolyzer models further development.

8 Acknowledgements

9
10 This work is a result of the project “UniRCell”, with the reference POCI-01-0145-

11 FEDER- 016422 supported by Norte Portugal Regional Operational Programme

12 (NORTE 2020), under the Portugal 2020 Partnership Agreement, through the European

13 Regional Development Fund (FEDER) and also supported by the Project “HYLANTIC”–

14 EAPA_204/2016, which is co-financed by the European Regional Development Found

15 in the framework of the Interreg Atlantic programme.

16 POCI (FEDER) also supported this work via CEFT, project UID/EMS/00532/2019.

17

18
19
20

21 Nomenclature

Symbol (Unit) Meaning

C(mol/cm3) Concentration of reactants/products

C0 (mol/cm3) Reference concentration

Activation energy
Eact (J/mol)
Open circuit voltage
Ecell (V)

28
 Journal Pre-proof

E th0 (V) Thermoneutral voltage at standard state

0
Erev (V) Reversible cell potential

F (C/mol) Faraday constant

Enthalpy
H(J/mol)
Current
I (A)
Current density
i (A/cm2)
Limit current density
ilim (A/cm2)
Anode exchange current density
i0,a(A/cm2)
Cathode exchange current density
i0,c (A/cm2)
Exchange current density at reference temperature
i0,ref (A/cm2)
n Number of electrons transferred in the reaction

Number of electrolyzer cells

Nc
Partial pressure of reactants/products
𝑃(Pa)
Standard pressure
P0 (Pa)
Q (C) Charge

R (J/molK) Ideal gas constant

Electrolyzer cell resistance

Rcell ()
Electrolyzer temperature
𝑇 (K)
Anode temperature
Ta (K)
Cathode temperature
Tc (K)
Tref (K) Reference temperature

V (V) Electrolyzer voltage

Anode activation overpotential

VAct,a (V)
Cathode activation overpotential
VAct,c (V)

29
 Journal Pre-proof

Diffusion or concentration overpotential

VDiff (V)
z Charge number

Greek symbols

Charge transfer coefficients at anode

a
Charge transfer coefficients at cathode
c
Variation of Gibbs energy
∆𝐺 (J/mol)

GR0 (J/mol) Gibbs free energy of reaction

HR0 (J/mol) Enthaphy of reaction

Fraction of proton back-permeation to total proton permeation

k
Membrane water content

Material thickness
𝛿(cm)
Material conductivity
𝜎 (S/cm)
1

2 Bibliography

3 Abdin, Z., C. J. Webb and E. M. Gray, 2015. Modelling and simulation of a proton exchange
4 membrane (PEM) electrolyser cell. Int J Hydrogen Energy 40, 13243-13257.
5 http://dx.doi.org/10.1016/j.ijhydene.2015.07.129.
6 Abdol Rahim, A. H., A. Tijani, T. Farah and H. Shukri, 2015. Simulation Analysis of the Effect of
7 Temperature on Overpotentials in PEM Electrolyzer System,
8 Agbli, K. S., M. C. Péra, D. Hissel, O. Rallières, C. Turpin and I. Doumbia, 2011. Multiphysics
9 simulation of a PEM electrolyser: Energetic Macroscopic Representation approach. Int J
10 Hydrogen Energy 36, 1382-1398. http://dx.doi.org/10.1016/j.ijhydene.2010.10.069.
11 Aouali, F. Z., M. Becherif, H. S. Ramadan, M. Emziane, A. Khellaf and K. Mohammedi, 2017.
12 Analytical modelling and experimental validation of proton exchange membrane electrolyser for
13 hydrogen production. Int J Hydrogen Energy 42, 1366-1374.
14 http://dx.doi.org/10.1016/j.ijhydene.2016.03.101.
15 Aouali, F. Z., M. Becherif, A. Tabanjat, M. Emziane, K. Mohammedi, S. Krehi and A. Khellaf, 2014.
16 Modelling and Experimental Analysis of a PEM Electrolyser Powered by a Solar Photovoltaic
17 Panel. Energy Procedia 62, 714-722. http://dx.doi.org/10.1016/j.egypro.2014.12.435.
18 Arbabi, F., H. Montazeri, R. Abouatallah, R. Wang and A. Bazylak, 2016. Three-Dimensional
19 Computational Fluid Dynamics Modelling of Oxygen Bubble Transport in Polymer Electrolyte
20 Membrane Electrolyzer Porous Transport Layers. J Electrochem Soc 163, F3062-F3069.
21 10.1149/2.0091611jes.
22 Atlam, O. and M. Kolhe, 2011. Equivalent electrical model for a proton exchange membrane
23 (PEM) electrolyser. Energy Conversion and Management 52, 2952-2957.
24 http://dx.doi.org/10.1016/j.enconman.2011.04.007.

30
 Journal Pre-proof

1 Awasthi, A., K. Scott and S. Basu, 2011. Dynamic modeling and simulation of a proton exchange
2 membrane electrolyzer for hydrogen production. Int J Hydrogen Energy 36, 14779-14786.
3 http://dx.doi.org/10.1016/j.ijhydene.2011.03.045.
4 Barbir, F., 2005. PEM electrolysis for production of hydrogen from renewable energy sources.
5 Solar Energy 78, 661-669. http://dx.doi.org/10.1016/j.solener.2004.09.003.
6 Becker, S. and V. Karri, 2010. Predictive models for PEM-electrolyzer performance using
7 adaptive neuro-fuzzy inference systems. Int J Hydrogen Energy 35, 9963-9972.
8 http://dx.doi.org/10.1016/j.ijhydene.2009.11.060.
9 Bessarabov, D., H. wang, H. Li and N. Zhao, 2015. PEM Electrolysis for Hydrogen Production:
10 Principles and Applications,
11 Biaku, C. Y., N. V. Dale, M. D. Mann, H. Salehfar, A. J. Peters and T. Han, 2008. A semiempirical
12 study of the temperature dependence of the anode charge transfer coefficient of a 6 kW PEM
13 electrolyzer. Int J Hydrogen Energy 33, 4247-4254.
14 http://dx.doi.org/10.1016/j.ijhydene.2008.06.006.
15 Briguglio, N., G. Brunaccini, S. Siracusano, N. Randazzo, G. Dispenza, M. Ferraro, R. Ornelas, A.
16 S. Aricò and V. Antonucci, 2013. Design and testing of a compact PEM electrolyzer system. Int J
17 Hydrogen Energy 38, 11519-11529. http://dx.doi.org/10.1016/j.ijhydene.2013.04.091.
18 Brown, T. M., J. Brouwer, G. S. Samuelsen, F. H. Holcomb and J. King, 2008. Dynamic first
19 principles model of a complete reversible fuel cell system. J Power Sources 182, 240-253.
20 http://dx.doi.org/10.1016/j.jpowsour.2008.03.077.
21 Carmo, M., D. L. Fritz, J. Mergel and D. Stolten, 2013. A comprehensive review on PEM water
22 electrolysis. Int J Hydrogen Energy 38, 4901-4934.
23 http://dx.doi.org/10.1016/j.ijhydene.2013.01.151.
24 Chandesris, M., V. Médeau, N. Guillet, S. Chelghoum, D. Thoby and F. Fouda-Onana, 2015.
25 Membrane degradation in PEM water electrolyzer: Numerical modeling and experimental
26 evidence of the influence of temperature and current density. Int J Hydrogen Energy 40, 1353-
27 1366. http://dx.doi.org/10.1016/j.ijhydene.2014.11.111.
28 Choi, P., D. G. Bessarabov and R. Datta, 2004. A simple model for solid polymer electrolyte (SPE)
29 water electrolysis. Solid State Ionics 175, 535-539. http://dx.doi.org/10.1016/j.ssi.2004.01.076.
30 Dale, N. V., M. D. Mann and H. Salehfar, 2008. Semiempirical model based on thermodynamic
31 principles for determining 6 kW proton exchange membrane electrolyzer stack characteristics. J
32 Power Sources 185, 1348-1353. http://dx.doi.org/10.1016/j.jpowsour.2008.08.054.
33 García-Valverde, R., N. Espinosa and A. Urbina, 2012. Simple PEM water electrolyser model and
34 experimental validation. Int J Hydrogen Energy 37, 1927-1938.
35 http://dx.doi.org/10.1016/j.ijhydene.2011.09.027.
36 Görgün, H., 2006. Dynamic modelling of a proton exchange membrane (PEM) electrolyzer. Int J
37 Hydrogen Energy 31, 29-38. http://dx.doi.org/10.1016/j.ijhydene.2005.04.001.
38 Han, B., J. Mo, Z. Kang and F.-Y. Zhang, 2016. Effects of membrane electrode assembly properties
39 on two-phase transport and performance in proton exchange membrane electrolyzer cells.
40 Electrochim Acta 188, 317-326. https://doi.org/10.1016/j.electacta.2015.11.139.
41 Harrison, K. W., E. Hernández-Pacheco, M. Mann and H. Salehfar, 2005. Semiempirical Model
42 for Determining PEM Electrolyzer Stack Characteristics. J Fuel Cell Sci Technol. 3, 220-223.
43 10.1115/1.2174072.
44 Hwang, J. J., L. K. Lai, W. Wu and W. R. Chang, 2009. Dynamic modeling of a photovoltaic
45 hydrogen fuel cell hybrid system. International Journal of Hydrogen Energy 34, 9531-9542.
46 http://dx.doi.org/10.1016/j.ijhydene.2009.09.100.
47 Kaya, M. F. and N. Demir, 2017. Numerical Investigation of PEM Water Electrolysis Performance
48 for Different Oxygen Evolution Electrocatalysts. Fuel Cells 17, 37-47.
49 doi:10.1002/fuce.201600216.
50 Kim, H., M. Park and K. S. Lee, 2013. One-dimensional dynamic modeling of a high-pressure
51 water electrolysis system for hydrogen production. Int J Hydrogen Energy 38, 2596-2609.
52 http://dx.doi.org/10.1016/j.ijhydene.2012.12.006.

31
 Journal Pre-proof

1 Lamy, C., 2016. From hydrogen production by water electrolysis to its utilization in a PEM fuel
2 cell or in a SO fuel cell: Some considerations on the energy efficiencies. International Journal of
3 Hydrogen Energy 41, 15415-15425. https://doi.org/10.1016/j.ijhydene.2016.04.173.
4 Lebbal, M. E. and S. Lecœuche, 2009. Identification and monitoring of a PEM electrolyser based
5 on dynamical modelling. Int J Hydrogen Energy 34, 5992-5999.
6 http://dx.doi.org/10.1016/j.ijhydene.2009.02.003.
7 Lee, J. K., C. H. Lee and A. Bazylak, 2019. Pore network modelling to enhance liquid water
8 transport through porous transport layers for polymer electrolyte membrane electrolyzers. J
9 Power Sources 10.1016/j.jpowsour.2019.226910.
10 Marangio, F., M. Santarelli and M. Calì, 2009. Theoretical model and experimental analysis of a
11 high pressure PEM water electrolyser for hydrogen production. Int J Hydrogen Energy 34, 1143-
12 1158. http://dx.doi.org/10.1016/j.ijhydene.2008.11.083.
13 Medina, P. and M. Santarelli, 2010. Analysis of water transport in a high pressure PEM
14 electrolyzer. Int J Hydrogen Energy 35, 5173-5186.
15 http://dx.doi.org/10.1016/j.ijhydene.2010.02.130.
16 Mohamed, B., B. Alli and B. Ahmed, 2016. Using the hydrogen for sustainable energy storage:
17 Designs, modeling, identification and simulation membrane behavior in PEM system
18 electrolyser. J Ene Storage 7, 270-285. http://dx.doi.org/10.1016/j.est.2016.06.006.
19 Moradi Nafchi, F., E. Afshari, E. Baniasadi and N. Javani, 2019. A parametric study of polymer
20 membrane electrolyser performance, energy and exergy analyses. Int J Hydrogen Energy 44,
21 18662-18670. 10.1016/j.ijhydene.2018.11.081.
22 Myles, T. D., G. J. Nelson, A. A. Peracchio, R. J. Roy, B. L. Murach, G. A. Adamson and W. K. S.
23 Chiu, 2012. Species transport in a high-pressure oxygen-generating proton-exchange membrane
24 electrolyzer. Int J Hydrogen Energy 37, 12451-12463.
25 http://dx.doi.org/10.1016/j.ijhydene.2012.05.123.
26 Nafeh, A. E.-S. A., 2011. Hydrogen production from a PV/PEM electrolyzer system using a neural-
27 network-based MPPT algorithm. International Journal of Numerical Modelling: Electronic
28 Networks, Devices and Fields 24, 282-297. 10.1002/jnm.778.
29 Ni, M., M. K. H. Leung and D. Y. C. Leung, 2008. Energy and exergy analysis of hydrogen
30 production by a proton exchange membrane (PEM) electrolyzer plant. Energy Conversion and
31 Management 49, 2748-2756. http://dx.doi.org/10.1016/j.enconman.2008.03.018.
32 Nie, J. and Y. Chen, 2010. Numerical modeling of three-dimensional two-phase gas–liquid flow
33 in the flow field plate of a PEM electrolysis cell. Int J Hydrogen Energy 35, 3183-3197.
34 http://dx.doi.org/10.1016/j.ijhydene.2010.01.050.
35 Nie, J., Y. Chen and R. F. Boehm, 2008. Numerical Modeling of Two-Phase Flow in a Bipolar Plate
36 of a PEM Electrolyzer Cell. 783-788. 10.1115/IMECE2008-68913.
37 Nie, J., Y. Chen, S. Cohen, B. D. Carter and R. F. Boehm, 2009. Numerical and experimental study
38 of three-dimensional fluid flow in the bipolar plate of a PEM electrolysis cell. International
39 Journal of Thermal Sciences 48, 1914-1922.
40 http://dx.doi.org/10.1016/j.ijthermalsci.2009.02.017.
41 Oliveira, L. F. L., C. Jallut and A. A. Franco, 2013. A multiscale physical model of a polymer
42 electrolyte membrane water electrolyzer. Electrochim Acta 110, 363-374.
43 http://dx.doi.org/10.1016/j.electacta.2013.07.214.
44 Olivier, P., C. Bourasseau and P. B. Bouamama, 2017. Low-temperature electrolysis system
45 modelling: A review. Renew Sust Energ Rev 78, 280-300.
46 https://doi.org/10.1016/j.rser.2017.03.099.
47 Onda, K., T. Murakami, T. Hikosaka, M. Kobayashi, R. Notu and K. Ito 2002. Performance Analysis
48 of Polymer-Electrolyte Water Electrolysis Cell at a Small-Unit Test Cell and Performance
49 Prediction of Large Stacked Cell. J Electrochem Soc 149, A1069-A1078. 10.1149/1.1492287.
50 Ruuskanen, V., J. Koponen, K. Huoman, A. Kosonen, M. Niemelä and J. Ahola, 2017. PEM water
51 electrolyzer model for a power-hardware-in-loop simulator. Int J Hydrogen Energy 42, 10775-
52 10784. https://doi.org/10.1016/j.ijhydene.2017.03.046.

32
 Journal Pre-proof

1 Saeed, E. W. and E. G. Warkozek, 2015. Modeling and Analysis of Renewable PEM Fuel Cell
2 System. Energy Procedia 74, 87-101. http://dx.doi.org/10.1016/j.egypro.2015.07.527.
3 Santarelli, M., P. Medina and M. Calì, 2009. Fitting regression model and experimental validation
4 for a high-pressure PEM electrolyzer. Int J Hydrogen Energy 34, 2519-2530.
5 http://dx.doi.org/10.1016/j.ijhydene.2008.11.036.
6 Sartory, M., E. Wallnöfer-Ogris, P. Salman, T. Fellinger, M. Justl, A. Trattner and M. Klell, 2017.
7 Theoretical and experimental analysis of an asymmetric high pressure PEM water electrolyser
8 up to 155 bar. Int J Hydrogen Energy 42, 30493-30508.
9 https://doi.org/10.1016/j.ijhydene.2017.10.112.
10 Sossan, F., H. Bindner, H. Madsen, D. Torregrossa, L. R. Chamorro and M. Paolone, 2014. A model
11 predictive control strategy for the space heating of a smart building including cogeneration of a
12 fuel cell-electrolyzer system. International Journal of Electrical Power & Energy Systems 62, 879-
13 889. http://dx.doi.org/10.1016/j.ijepes.2014.05.040.
14 Tijani, A. S., N. A. Binti Kamarudin and F. A. Binti Mazlan, 2018. Investigation of the effect of
15 charge transfer coefficient (CTC) on the operating voltage of polymer electrolyte membrane
16 (PEM) electrolyzer. Int J Hydrogen Energy 43, 9119-9132.
17 https://doi.org/10.1016/j.ijhydene.2018.03.111.
18 Toghyani, S., E. Afshari and E. Baniasadi, 2018. Three-dimensional computational fluid dynamics
19 modeling of proton exchange membrane electrolyzer with new flow field pattern. Journal of
20 Thermal Analysis and Calorimetry 10.1007/s10973-018-7236-5.
21 Toghyani, S., S. Fakhradini, E. Afshari, E. Baniasadi, M. Y. Abdollahzadeh Jamalabadi and M.
22 Safdari Shadloo, 2019. Optimization of operating parameters of a polymer exchange membrane
23 electrolyzer. Int J Hydrogen Energy 44, 6403-6414.
24 https://doi.org/10.1016/j.ijhydene.2019.01.186.
25 Touré, S., A. Konaté, D. Traoré and D. Fofana, 2018. Novel determination method of charge
26 transfer coefficient of PEM fuel cell using the Lagrange’s multiplier method. IOP Conference
27 Series: Earth and Environmental Science 188, 012041. 10.1088/1755-1315/188/1/012041.
28 Yigit, T. and O. F. Selamet, 2016. Mathematical modeling and dynamic Simulink simulation of
29 high-pressure PEM electrolyzer system. Int J Hydrogen Energy 41, 13901-13914.
30 https://doi.org/10.1016/j.ijhydene.2016.06.022.
31

32

33
 Journal Pre-proof

1 Table 1 - Values reported in the literature for exchange current density and charge transfer

2 coefficients for anode and cathode reactions.

Reference 𝒊𝟎,𝒂(A/cm2) 𝒊𝟎,𝒄(A/cm2) 𝜶𝒂 𝜶𝒄

Hwang, Lai et al. 0.5 × 10 ―3 0.4 × 10 ―6 0.5 0.25

(2009)

Dale, Mann et al. 2.93 × 10 ―5a) 0.21a) n.r. 0.5

(2008)

Kim, Park et al. 10-2 10-4 n.r. n.r.

(2013)

Marangio, 10-13 to 10-6b) 10-13 to 10-6b) 2 0.5

Santarelli et al.

(2009)

Ni, Leung et al. 1.0 × 10 ―9 1.0 × 10 ―3 0.5 0.5

(2008)

Harrison, 1.0 × 10 ―8 9.0 × 10 ―2 0.5 0.5

Hernández-

Pacheco et al.

(2005)

Choi, Bessarabov 1.0 × 10 ―7 1.0 × 10 ―3 0.5 0.5

et al. (2004)

Yigit and Selamet 2.0 × 10 ―7 2.0 × 10 ―3 2 0.5

(2016)

Abdin, Webb et al. 1.0 × 10 ―7 1.0 × 10 ―1 0.8 0.25

(2015)

34
 Journal Pre-proof

Han, Mo et al. 0.5 × 10 ―6 0.1 --- ---

(2016)

Biaku, Dale et al. 1.9 × 10 ―3a 0.303a 0.257 a) 0.5

(2008)

Agbli, Péra et al. 1.548 × 10 ―3 3.539 × 10 ―2 0.7178 0.6935

(2011)

Tijani, Binti 1.0 × 10 ―9 1.0 × 10 ―3 0.864c) 0.216c)

Kamarudin et al.

(2018)

Rahim, Tijani et al. 1.0 × 10 ―9 1.0 × 10 ―3 --- ---

(2015)

Toghyani,

Fakhradini et al. 10-6 to 10-1 0.01-10 --- ---

(2019)

Lebbal and
0.13 × 10 ―3d) 0.452d)
Lecoeuche (2009)

Brown, Brouwer et
0.01 × 10 ―4e) 0.5e)
al. (2008)

2 a) Values for T=25 ºC extracted from adjust experimental data.

3 b) Values chosen between this range that better fits the experimental results.

4 c) Values for T=30ºC.

5 d) The authors determined a global value for I0 and  by applying the non-linear

6 least square method on real data given by the PEM electrolyser cell. The authors

7 didn’t report the units for I0.

35
 Journal Pre-proof

1 e) The authors used a global value for I0 and 

36
 Journal Pre-proof

1 Table 2 – Software used to implement the developed models found in literature.

Reference Software/ algorithms used

Harrison, Hernández et Mathematica

al. (2005)

Görgün (2006) MATLAB/Simulink

Biaku, Dale et al. (2008) Mathematica

Dale, Mann et al. (2008) Mathematica

Nie, Chen et al. (2008) ANSYS Fluent

Santarelli, Medina et al. Yate’s method and ANOVA

(2009)

Awasthi, Scott et al. MATLAB/Simulink

(2011)

Agbli, Péra et al. (2011) MATLAB/Simulink

Kim, Park et al. (2013) MATLAB

Rahim, Tijani et al. MATLAB

(2015)

Becker and Karri (2010) MATLAB/Simulink

Oliveira, Jallut et al. MATLAB/Simulink and MEMEPhys

(2013)

Briguglio, Brunaccini et COMSOL Multiphysics

al. (2013)

Abdin, Webb et al. MATLAB/Simulink

(2015)

Saeed and Warkozek MATLAB

(2015)

37
 Journal Pre-proof

Mohamed, Alli et al. MATLAB

(2016)

Yigit and Selamet (2016) MATLAB/Simulink

Arbabi, Montazeri et al. OpenFOAM

(2016)

Han, Mo et al. (2016) Fortran 90

Aouali, Becherif et al. MATLAB/Simulink

(2017)

Kaya and Demir (2017) COMSOL Multiphysics

Sartory, Wallnöfer- MATLAB/Simulink

Ogris et al. (2017)

Lee, Lee et al. (2019) OpenPNM

38
 Journal Pre-proof

Figure 1

Falcão and Pinto, 2019

 Journal Pre-proof

Figure 2
Falcão and Pinto, 2019
 Journal Pre-proof

Highlights

 This review aims to be useful for the beginners on PEM electrolyzer modeling area;
 Provides a compilation of the models published in the last years;
 Includes a section with the technology basic principles;
 Enlighten important aspects for PEM electrolyzer models further development.