Self-Discharge of a Proton Exchange Membrane

Electrolyzer: Investigation for Modeling Purposes

Ángel Hernández-Gómez, Victor Ramirez, Damien Guilbert, Belem Saldivar

To cite this version:

Ángel Hernández-Gómez, Victor Ramirez, Damien Guilbert, Belem Saldivar. Self-Discharge of a
Proton Exchange Membrane Electrolyzer: Investigation for Modeling Purposes. Membranes, 2021, 11
(6), pp.379. �10.3390/membranes11060379�. �hal-03232981�

HAL Id: hal-03232981

https://hal.science/hal-03232981v1
Submitted on 23 May 2021

HAL is a multi-disciplinary open access L’archive ouverte pluridisciplinaire HAL, est

archive for the deposit and dissemination of sci- destinée au dépôt et à la diffusion de documents
entific research documents, whether they are pub- scientifiques de niveau recherche, publiés ou non,
lished or not. The documents may come from émanant des établissements d’enseignement et de
teaching and research institutions in France or recherche français ou étrangers, des laboratoires
abroad, or from public or private research centers. publics ou privés.
 membranes

Article
Self-Discharge of a Proton Exchange Membrane Electrolyzer:
Investigation for Modeling Purposes
Ángel Hernández-Gómez 1,† , Victor Ramirez 1,2,† , Damien Guilbert 3, *,† and Belem Saldivar 2,4

1 Department of Renewable Energy, Centro de Investigación Científica de Yucatán (CICY), Mérida,

Yucatán P.C. 97205, Mexico; [email protected] (Á.H.-G.); [email protected] (V.R.)
2 Cátedras CONACYT, Ciudad de México P.C. 03940, Mexico; [email protected]
3 Group of Research in Electrical Engineering of Nancy (GREEN), Université de Lorraine, GREEN,
F-54000 Nancy, France
4 Facultad de Ingeniería, Universidad Autónoma del Estado de México (UAEM), Ciudad de México,
Toluca P.C. 50000, Mexico
* Correspondence: [email protected]
† These authors contributed equally to this work.

Abstract: The self-discharge phenomenon results in a decrease of the open-circuit voltage (OCV),
which occurs when an electrochemical device is disconnected from the power source. Although the
self-discharge phenomenon has widely been investigated for energy storage devices such as batteries
and supercapacitors, no previous works have been reported in the literature about this phenomenon
for electrolyzers. For this reason, this work is mainly focused on investigating the self-discharge
voltage that occurs in a proton exchange membrane (PEM) electrolyzer. To investigate this voltage
drop for modeling purposes, experiments have been performed on a commercial PEM electrolyzer



to analyze the decrease in the OCV. One model was developed based on different tests carried out


on a commercial-400 W PEM electrolyzer for the self-discharge voltage. The proposed model has
Citation: Hernández-Gómez, Á.;
been compared with the experimental data to assess its effectiveness in modeling the self-discharge
Ramirez, V.; Guilbert, D.; Saldivar, B.
phenomenon. Thus, by taking into account this voltage drop in the modeling, simulations with a
Self-Discharge of a Proton Exchange
higher degree of reliability were obtained when predicting the behavior of PEM electrolyzers.
Membrane Electrolyzer: Investigation
for Modeling Purposes. Membranes
Keywords: PEM electrolyzer; modeling; dynamic model; self-discharge voltage
2021, 11, 379. https://doi.org/
10.3390/membranes11060379

Academic Editor: Jin-Soo Park

1. Introduction
Received: 28 April 2021 Electrolyzers are reliable devices that have been employed for various applications
Accepted: 20 May 2021 in residential, commercial, and industrial areas thanks to their well-established technol-
Published: 22 May 2021 ogy [1,2]. Additionally, electrolyzers do not require continuous maintenance since they
hardly include mobile elements [3]. The main function of electrolyzers is to produce
Publisher’s Note: MDPI stays neutral high-purity hydrogen (up to 99.999 vol %) by using the water electrolysis process, so the
with regard to jurisdictional claims in produced hydrogen can be directly used in low-temperature fuel cells, which are sensitive
published maps and institutional affil-
to impurities of the hydrogen stream [4]. Currently, according to the type of electrolyte,
iations.
there are three main technologies: proton exchange membrane (PEM) electrolyzer, alkaline
electrolyzer, and solid oxide (SO) electrolyzer.
Although each type of electrolyzer has its advantages and disadvantages, PEM elec-
trolyzers have attracted a lot of attention in recent years from researchers due to their
Copyright: © 2021 by the authors. compact system design, specific production capacity, high energy efficiency, and simplic-
Licensee MDPI, Basel, Switzerland. ity [5,6]. Additionally, compared to alkaline technology, PEM technology provides superior
This article is an open access article performance when it is coupled with renewable energy sources (RES) due to its high
distributed under the terms and flexibility and quick responses to dynamics. This important feature enables capturing
conditions of the Creative Commons
energy during dynamic operations, which are consistent during the operation of RES.
Attribution (CC BY) license (https://
On the other hand, since electrolyzers are electrochemical devices such as batteries and
creativecommons.org/licenses/by/
supercapacitors, they feature a self-discharge phenomenon. This phenomenon is caused
4.0/).

Membranes 2021, 11, 379. https://doi.org/10.3390/membranes11060379 https://www.mdpi.com/journal/membranes

Membranes 2021, 11, 379 2 of 17

by internal chemical reactions leading to a decrease of the open-circuit voltage (OCV)

when the electrolyzer is disconnected from the power supply. In comparison, for batteries
and supercapacitors, the self-discharge phenomenon reduces their stored charge and their
lifespan.
To disseminate PEM electrolyzer technology on a large-scale market, different research
works have been carried out. In particular, mathematical modeling is a valuable tool for
predicting static and dynamic behavior when coupling PEM electrolyzers with RES [7,8].
Research using modeling tools has been of vital importance to studying the electrical
domain in PEM electrolyzers, which is the key to understanding and enhancing the per-
formance of the electrolyzer [9,10]. Besides, different models have already been proposed
and can be divided into three categories. The first category concerns empirical models
reported in [6,9]. These models allow taking into consideration the electrical domain of
PEM electrolyzers, including the OCV, activation, ohmic, and concentration overvolt-
age according to the temperature and pressure. These models are useful for developing
electrolyzer emulators [6]. In comparison, the second category includes semi-empirical
models proposed in [11,12]. The voltage–current curve describing the performance of the
electrolyzer is modeled by using semi-empirical equations depending on the temperature
and pressure. Both developed semi-empirical and empirical models neglect the dynamic
operations of electrolyzer, and they have to be taken into consideration for coupling the
electrolyzers with renewable energy sources. As a result, the last category is related to
dynamic models firstly developed in [13,14]. The dynamic behavior and the losses of
the electrolyzer have been modeled as an equivalent electrical circuit whose parameters
strongly depend on the operating conditions (supply current, temperature, and pressure).
Additionally, investigations on how voltage affects other electrolyzer parameters have
been reported in the literature. For example, in [15], the authors have investigated the gas
crossover phenomenon, which is related to the ohmic loss of the membrane. Other impor-
tant parameters are the efficiencies in the PEM electrolyzer, which indicate the performance
of the device (i.e., if the specific energy consumption regarding the hydrogen production
is efficient) [16]. The main efficiencies reported in the literature are Faraday’s efficiency,
voltage efficiency, and energy efficiency [16,17].
However, despite these extensive research on PEM electrolyzer modeling, no work in
the literature has reported the phenomenon of self-discharge voltage in PEM electrolyzers.
The term "self-discharge" is sometimes associated with the chemical reactions discharging
the surface and excluding any physical processes which cause the voltage drop [18,19]. The
phenomenon of self-discharge voltage has proven to be an important subject of study for
any electrochemical device, particularly for batteries and supercapacitors. For batteries,
the self-discharge voltage is the main limitation to storing energy for a long time, so the
efficiency and autonomy of systems are poor [20], which is a challenging issue when
using batteries in electric vehicles, industries, and residences [21,22]. To reduce battery
self-discharge voltage, many strategies have been developed, such as optimizing materials
for battery construction [22,23] and developing different models with a high degree of
reliability (i.e., by taking into account several variables or assimilating the battery like
an electronic circuit) [24,25]. On the other hand, for supercapacitors, besides being deter-
minant for the duration of energy storage (i.e., rest phases), self-discharge voltage is an
important indicator to quantify performance [26,27]. As a result, by decreasing the effects
of supercapacitors’ self-discharge voltage, the lifespans of devices that depend on the
power supply of supercapacitors are improved [28,29]. Like batteries, different materials
for the construction of supercapacitors have been used, and the development of models
has been implemented as a strategy to decrease the self-discharge voltage [30,31].
In comparison, the self-discharge phenomenon for fuel cells is usually neglected in
the literature, and the first investigation about this important issue was reported in [32].
It has been highlighted that decreases of the OCV of fuel cells are mainly caused by the
leakage currents due to gas crossover through the membrane. On the one hand, currently
manufactured membranes feature thin membranes to reduce the ohmic losses associated
Membranes 2021, 11, 379 3 of 17

with the conduction of protons. On the other hand, the thinner the membrane, the higher
the gas crossover and leakage currents [33,34]. Over a long period of operation, gas
crossover through the membrane may reduce the lifespan of the fuel cell and degrade the
membrane (e.g., pinholes) [34]. Therefore, this phenomenon remains a major concern for
electrolyzers and further investigations are required.
From this current state-of-the-art, it is important to start studying the phenomenon
of self-discharge voltage that occurs in PEM electrolyzers, since it is significant when the
electrolyzer is coupled with RES and should be useful for the development of future works
as PEM electrolyzer emulators (i.e., by taking this phenomenon into account, more realistic
behavior of the PEM electrolyzer voltage can be obtained). For this reason, this we aimed
at describing the self-discharge voltage met in a PEM electrolyzer through different experi-
mental tests and developing a model enabling reproducing the self-discharge phenomenon
according to the operating conditions, such as the current. This model has been developed
based on models proposed in the literature for supercapacitors and batteries to describe the
behavior of self-discharge voltage. The comparison between the developed model and the
experimental data for different operating conditions has demonstrated the high reliability
of the model in reproducing the self-discharge phenomenon.
This paper is organized as follows. After the introduction, the sighting of the self-
discharge voltage in a PEM electrolyzer (description of the experimental test setup and
self-discharge voltage issues) is described in Section 2. The development of the model for
self-discharge voltage is introduced in Section 3. Validation and discussion are provided in
Section 4. Finally, the conclusion of this work is presented in Section 5.

2. Sighting of the Self-Discharge Voltage in a PEM Electrolyzer

2.1. Description of the Experimental Test Setup
To study the self-discharge voltage of the PEM electrolyzer stack, an experimental test
rig has been designed and built at the GREEN laboratory, IUT de Longwy, as shown in
Figure 1. The experimental test rig is composed of the following devices and components:
(1) a laptop; (2) dSPACE control desk software; (3) a DC power supply; (4) a DS1104
controller board; (5) a commercial-400W PEM electrolyzer; (6) a voltage probe MTX 1032-B
from the Metrix Company to acquire the stack voltage. The DC power supply is controlled
through a virtual control panel installed on the laptop. The PEM electrolyzer is supplied
from pure water featuring low water conductivity (less than 2 µS · m−1 ). Higher water
conductivity can contaminate and damage the electrolyzer. The specifications of the studied
commercial PEM electrolyzer system NMH2 1000 from HELIOCENTRIS® Company are
provided in Table 1. Since the PEM electrolyzer NMH2 1000 system has been developed
for educational purposes, it includes power electronics connected to the stack. In this
work, only the stack of the PEM electrolyzer is considered and connected to an external DC
power supply as depicted in Figure 1. The studied PEM electrolyzer has a solid polymer
electrolyte based on fluoropolymer Nafion material from DuPont® company. The thickness
of the membrane is very thin (around 25 µm), resulting in lower ohmic losses and may
influence the self-discharge phenomenon based on previous works reported for PEM fuel
cells [33,34] and electrolyzers [15,35]. The acquired stack voltage from the voltage probe
is then transferred into the DS1104 controller board. Finally, the experimental data are
monitored and saved through dSPACE control desk software. The sampling time was
chosen as 1 s to collect the experiment data. The obtained experimental data were plotted
by using Matlab-Simulink® software.
Membranes 2021, 11, 379 4 of 17

Figure 1. Experimental test setup.

Table 1. Specifications of the HELIOCENTRIS® NMH2 1000 PEM electrolyzer.

Parameter Value Unit

Rated electrical power 400 W
Stack voltage operating range 4.2–8 V
Stack current range 0–50 A
Operating temperature range 288.15–313.15 K
Hydrogen outlet pressure 10.5 bar
Cells number 3 -
Active area section 50 cm2
Hydrogen flow rate range at STP SLPM (Standard
0–1
(Standard Temperature and Liter Per Minute)
Pressure, 20 ◦ C and 1 bar)

2.2. Self-Discharge Voltage Issues

From the experimental test rig presented in Figure 1, several experimental tests based
on different initial conditions (10, 20, 30, and 35 A) have been performed to emphasize the
self-discharge phenomenon in a PEM electrolyzer. The obtained results are depicted in a
single figure (Figure 2) to make easier their analysis for comparison purposes. The duration
of the test was 5500 s. These tests consisted of supplying first the PEM electrolyzer with a
constant current (10, 20, 30, and 35 A) for one hundred seconds, and then disconnecting
the PEM electrolyzer from the DC power supply to observe the decrease of OCV. Based
on Figure 2 it is important to point out that the duration in which a constant current is
applied to the electrolyzer, and the initial operating conditions, influence the self-discharge
phenomenon. As a result of the disconnection from the DC power supply, an immediate
voltage drop can be observed in the stack voltage. This immediate decrease in the stack
voltage can be explained based on a previous study where an equivalent electrical circuit
was developed to model the static and dynamic behavior of the PEM electrolyzer [13].
The immediate voltage drop was due to the sum of the equivalent resistances modeling
Membranes 2021, 11, 379 5 of 17

the anode, cathode, and electrolyte. This first voltage drop ended when the stack voltage
reached 4.2 V related to the OCV of the electrolyzer stack.

Figure 2. Self-discharge tests according to different initial conditions (10, 20, 30, and 35 A).

After that, the voltage drop related to the self-discharge phenomenon can be em-
phasized. This self-discharge can be divided into two phenomena: one is related to an
accelerated self-discharge and the second to a slow, continuous self-discharge. These
two phenomena are well known when characterizing batteries, as reported in the liter-
ature [36,37]. The accelerated self-discharge is commonly called OCV “depolarization,”
and the slow continuous self-discharge is defined as OCV “relaxation”. In Figure 2, it
can be seen that the OCV for the test (35–0 A) decreased more quickly and continuously
compared to the other tests at lower currents. This observation demonstrates that the initial
operating conditions affect the self-discharging. To emphasize those two phenomena (OCV
depolarization and relaxation), the results of each test have been plotted in Figures 3–6.
The OCV relaxation occurred when the OCV reached around 0.5 V. For each test, the time
separating these two OCV phenomena can be apprehended. Depending on the initial
operating conditions, the duration of the OCV depolarization was longer or shorter. For
instance, in Figure 6, the duration of the OCV depolarization is shorter (around 2850 s) com-
pared to the other tests. A summary of the measured duration of the OCV depolarization
for each test is provided in Table 2.

Table 2. Summary of the measured duration of the OCV depolarization.

Test Length of the OCV Depolarization

10–0 A (Figure 3) 3950 s
20–0 A (Figure 4) 3350 s
30–0 A (Figure 5) 3150 s
35–0 A (Figure 6) 2850 s
Membranes 2021, 11, 379 6 of 17

Figure 3. OCV depolarization and relaxation (10–0 A).

Figure 4. OCV depolarization and relaxation (20–0 A).

Membranes 2021, 11, 379 7 of 17

Figure 5. OCV depolarization and relaxation (30–0 A).

Figure 6. OCV depolarization and relaxation (35–0 A).

Membranes 2021, 11, 379 8 of 17

Based on the previous works reported for self-discharge in batteries [18,20] and
supercapacitors [27,28], the studied PEM electrolyzer features a higher self-discharge rate.
Indeed, lead-acid and lithium-ion batteries feature low self-discharge rates (noticeable in
terms of days, weeks, or even months), and supercapacitors higher self-discharge rates
(noticeable in terms of hours or days). In supercapacitors, the high self-discharge is due
to their organic electrolyte. For batteries, the self-discharge rate strongly depends on the
temperature and the wear (decrease in the number of charge/discharge cycles) due to
deep discharges [18]. The higher the temperature, the higher the self-discharge rate. The
same phenomenon has been reported when the batteries are subject to deep discharge.
In comparison, for supercapacitors, it has been highlighted that their self-discharge rates
are dependent on their double-layer capacitance values, and on the temperature [19].
The higher the temperature, the higher the self-discharge rate; whereas the smaller the
double-layer capacitance value, the higher the self-discharge rate.
As it has been emphasized in previous works [13,14], when supplying the PEM
electrolyzer with dynamic current profiles, the electrolyzer behaves as a capacitor because
of the charge double-layer effect. Indeed, between the electrode and the electrolyte, there is
a layer of charge, which can store electrical charge, and therefore, energy. The accumulation
of charges generates an electrical voltage, which corresponds to the activation overvoltage
both at the anode and at the cathode. Thus, when the current immediately changes, the
activation overvoltage at both the anode and the cathode takes some time before following
the change in current due to the reaction kinetics. Besides, the PEM electrolyzer features
two dynamics, the first one slower regarding the anode, and the second one faster regarding
the cathode. These two different dynamics can be modeled by two RC branches where their
constant times depend on the values of the two resistors considering the values of both
capacitances equal [38]. Given that the dynamics both at the anode and cathode change
according to the operating conditions, the values of both RC branches cannot be considered
constant. For this reason, in [14], the parameters of both RC branches have been assessed
through mathematical modeling whatever the operating conditions. The reported values
for the double-layer capacitance are similar to the ones for supercapacitors (between 3
and 69 F) [14]. By starting from this analysis and the previous observations reported for
supercapacitors, the high-self discharge rate depicted in Figure 2 can be explained. Indeed,
the low double-layer capacitance value of the PEM electrolyzer leads to an accelerated and
slow continuous self-discharge.
Finally, based on previous works reported for PEM electrolyzers [15,35] and the
knowledge of the authors on this topic, the self-discharge phenomenon in PEM electrolyzers
is mainly induced by the leakage currents due to gas crossover through the membrane.
Indeed, a part of the current at the cathode side has not been combined with the protons
to generate hydrogen. Hence, this current goes back to the anode through the membrane
that, at low current densities, is more permeable to gas crossover, as demonstrated in
the literature [15,35]. Besides, it is important to point out that gas crossover is strongly
dependent on temperature, pressure, and membrane thickness, as emphasized in the
literature [15,39]. In the case under study, the PEM electrolyzer features a very thin
membrane (i.e., around 25 µm). As highlighted in [39], the thinner the membrane, the
higher the gas crossover. Therefore, the thin membrane of the PEM electrolyzer may
explain the high self-discharge rate observed in Figure 2.
To conclude, despite these first observations, further investigation is required to
understand the self-discharge rates of PEM electrolyzers and the parameters affecting the
self-discharge rate, such as temperature and accelerated wear linked to their operating
conditions (static and dynamic).

3. Mathematical Model
As mentioned in the introduction, there are no reported models for the self-discharge
voltage of the PEM electrolyzer. For this reason, this work is also focused on developing a
model to describe the self-discharge voltage behavior.
Membranes 2021, 11, 379 9 of 17

As can be seen, there is a lag time in the experimental data before self-discharge voltage
occurs; see Figure 2. For this reason, it is assumed that the voltage in the electrolyzer has
normal behavior before the self-discharge voltage occurs. Therefore, taking into account
the model for voltage developed in [13,14], which describes the static-dynamic behavior of
the voltage drop in small periods (i.e., time lapses around 50 s) into a PEM electrolyzer,
the static-dynamic model for the voltage before the self-discharge voltage occurs can be
expressed as:
(
V0 if t < tc ,
V (t) = (1)
Vrev + i (t) · Rmem + ηact (t) if t ≥ tc ,

where V0 is the initial voltage of the PEM electrolyzer (i.e., it is the value obtained before
disconnecting the PEM electrolyzer from the power source). Vrev is the OCV, [V ]. i is
the current in the cell [ A]; Rmem is the membrane resistance [Ω]. ηact is the activation
overvoltage, [V ]. t is the time in [s]. Finally, tc is the time that the electrolyzer remains
connected to the energy source, [s].
Furthermore, due to the experiment carried out, the current function can be expressed
as the step function:
(
A1 if t < tc ,
i (t) = (2)
A2 if t ≥ tc ,

where both A1 and A2 take constant values, [ A]. Besides, the activation overvoltage is
expressed as ηact = ηact,c + ηact,a , where ηact,c is:

tc − t
      
τc
ηact,c (t) = · ( A2 − A1 ) · 1 − exp + f 1 ( t ) + g1 ( t ) (3)
Cc τc

and ηact,a :

tc − t
      
τa
ηact,a (t) = · ( A2 − A1 ) · 1 − exp + f 2 ( t ) + g2 ( t ) , (4)
Ca τa

where Cc and Ca are capacitance for cathode and anode; as for [ F ], both capacitances are
considered equal based on a previous work reported in [38]. τc and τa are time constants,
[s]. The functions f 1 (t) and f 2 (t) are given by:
 h i  h i
f 1 (t) = A1 · 1 − exp − τc
t
, f 2 (t) = A1 · 1 − exp −
τa
t
, (5)

and the functions g1 (t) and g2 (t) as:

h i h i
g1 (t) = ηact,c (0) · exp − t
τc , g2 (t) = ηact,a (0) · exp − t
τa . (6)

To model the self-discharge voltage, the models of supercapacitors and batteries

reported in [18,19] have been taken into consideration. In addition, the behavior of the
self-discharge voltage curve has been taken into account, which has a voltage drop rate
and subsequently stable behavior (see Figure 2). Thus, the proposed model for the self-
discharge voltage is Verhulst’s equation given by:
 
dV V
= r·V· 1− (7)
dt K

where r is the rate of voltage drop (self-discharge voltage). K is the voltage that the
electrolyzer maintains after the self-discharge phenomenon, [V ].
Membranes 2021, 11, 379 10 of 17

Equation (7) has the analytical solution:

K · V (0) · exp[r · t]
V (t) = (8)
K + V (0) · exp[r · t]

Therefore, by considering Equations (1) and (8), the self-discharge voltage model for
PEM electrolyzer is expressed as:



 V0 if t < tc ,






V (t) = Vrev + i (t) · Rmem + ηact (t) if tc ≤ t < tτ , (9)





 K · V (tτ ) · exp[r · t]
if tτ ≤ t.


K + V (tτ ) · exp[r · t]


where tτ indicates the time in which the self-discharge phenomenon begins in the PEM
electrolyzer, [s]. The following section presents the validation and a discussion on the
model developed in this section.

4. Validation and Discussion

To validate the model, the parameters of Equation (9) have been estimated using an
m-file in Matlab® together with the command lsqcurvefit, which is based on the least-square
regression algorithm. The values of the parameters are shown in Table 3.

Table 3. Calculated parameters for self-discharge voltage model.

Parameter Value Unit

Vrev 0.21 V
Cc 800 F
Ca 800 F
τc 180 s
τa 2558.9
h  i s
ηact,c (0) 1 V
5.5719· log 9.1933 · i
ηact,a (0) −8.2919 × 10−6 · i4 + 9.8489 × 10−4 · i3
− 0.0314 · i2 − 0.0027 · i + 3.6606 V
r −4.1226 × 10−4 (-)
K 0.3941 V

As can be seen in Table 3, the values of ηact,c (0) and ηact,a (0) depend on the current
input i, which agrees with the work reported in [14,40].
After calculating the parameters, the effectiveness of the model at reproducing the
real behavior of the self-discharge voltage in the PEM electrolyzer has been evaluated. To
carry out this evaluation, a comparison between the experimental data and the model has
been created using the mean absolute percentage error Er and the mean absolute error Em ,
as follows:
!
N V
exp,k − Vsim,k
 
100
Er = · ∑ (10)
N k =1
Vexp,k

!
  N
1
Em =
N
· ∑ Vexp,k − Vsim,k (11)
k =1

where N is the number of experimental data. Vexp,k and Vsim,k are the k experimental data
and k simulation data, respectively.
Membranes 2021, 11, 379 11 of 17

To observe the comparison between the experimental data and the model simulations,
an m-file was developed in Matlab® . The results are shown in Figures 7–10. In these figures,
the OCV depolarization and relaxation are emphasized. In Figure 7, one can appreciate
the comparison between the curve of the experimental data from 10 to 0 A and the curve
generated by the model. To generate this simulation, the values A1 = 10 A, A2 = 0 A,
V0 = 6.78 V, tc = 155 s, and tτ = 3200 s have been used.

7
Experimental Data
6 Model
Stack voltage (V)

3 Depolarization Relaxation

0
0 1000 2000 3000 4000 5000
Time (s)
Figure 7. Comparison between the experimental data from 10 to 0 A and the simulation of the model.

In Figure 8, the comparison between the curve of the experimental data from 20 to
0 A and the curve generated by the model is shown. To perform the simulation, the values
A1 = 20 A, A2 = 0 A, V0 = 7 V, tc = 210 s, and tτ = 2600 s have been used.

8
Experimental Data
7 Model

6
Stack voltage (V)

3 Depolarization Relaxation

0
0 1000 2000 3000 4000 5000
Time (s)
Figure 8. Comparison between the experimental data from 20 to 0 A and the simulation of the model.
Membranes 2021, 11, 379 12 of 17

In Figure 9, one can observe the comparison between the curve of the experimental
data from 30 to 0 A and the curve generated by the model. To develop this simulation, the
values A1 = 30 A, A2 = 0 A, V0 = 6.9 V, tc = 190 s, and tτ = 2400 s have been used.

7
Experimental Data
Model
6

5
Stack voltage (V)

3 Depolarization Relaxation

0
0 500 1000 1500 2000 2500 3000 3500 4000 4500
Time (s)
Figure 9. Comparison between the experimental data from 30 to 0 A and the simulation of the model.

Finally, in Figure 10, the comparison between the curve of the experimental data from
35 to 0 A and the curve generated by the model is shown. To perform the simulation, the
values A1 = 35 A, A2 = 0 A, V0 = 7.1 V, tc = 170 s, and tτ = 2300 s have been used.

8
Experimental Data
7 Model

6
Stack voltage (V)

3 Depolarization Relaxation

0
0 1000 2000 3000 4000 5000
Time (s)
Figure 10. Comparison between the experimental data from 35 to 0 A and the simulation of the model.
Membranes 2021, 11, 379 13 of 17

4.1. Discussion
To facilitate the comparison between the experimental data and the simulated data, a
summary of the relative error and mean error is presented in Table 4. On the one hand, a
relative error Er of less than 5% is presented for the tests of 10–0, 20–0, and 30–0 A. The
test with 35–0 A obtained a relative error of 5.69%, this being the highest relative error
obtained within these comparisons. On the other hand, a mean error Em of less than 0.075
V was obtained for each one of the experimental tests. Therefore, in general, there was an
average relative error of Er = 4.6% and an average mean error of Em = 0.0603 V for the
experimental tests, which validates the model.

Table 4. Summary of the relative and mean errors.

Input Current Er Em
10–0 A 3.55% 0.0528 V
20–0 A 4.43% 0.0489 V
30–0 A 4.70% 0.0745 V
35–0 A 5.69% 0.0648 V

As can be seen, the behavior of the self-discharge voltage curve was expected to
be asymptotically stable for large periods. Furthermore, the behavior presented by the
self-discharge voltage in a PEM electrolyzer has great similarity with the curve described
by Verhulst’s equation. Besides, as highlighted in [41], the self-discharge phenomenon
in supercapacitors does not obey the usual empirical equations. For this reason, the
developed model was based on physical reasoning, taking into consideration the real
physical phenomena governed by supercapacitors’ and electrolyzers’ operation. As a
result, the model achieved high accuracy.
Finally, recommendations to prevent the phenomenon of self-discharge voltage in
a PEM electrolyzer are presented in the next subsection as a result of the analysis of the
experimental data.

4.2. Self-Discharge Prevention

Over the last few years, research has been intensified to develop different solutions to
reduce the self-discharging of supercapacitors by providing improvements to the electrodes,
separators, or electrolytes [42,43]. Indeed, as highlighted in Section 2, the self-discharge of
supercapacitors is an important issue due to rapid self-discharging, low energy efficiency,
and loss of stored energy. Like supercapacitors, self-discharge in PEM electrolyzers is a
major concern given that gas crossover through the membrane and leakage currents may
decrease the lifespan of the electrolyzer and lead to the degradation of the membrane
(e.g., pinholes), as reported for PEM fuel cells [32,34]. Besides, when the electrolyzer
is reconnected to a power supply with low OCV (as a result of a self-discharge), high
overshoot may occur, likely causing deterioration of the PEM electrolyzer [44]. In Figure 11,
a voltage overshoot (around 9 V, 12.5% higher than the rated voltage of the electrolyzer
(i.e., 8 V)) is shown from when reconnecting the DC power supply (current step from 0
to 20 A). At the beginning of the test, it can be highlighted that the OCV was very low
(close to 0 V) as a result of the self-discharge. To prevent the self-discharge issue in PEM
electrolyzers, several potential solutions can be adopted:
1. A compromise must be found in the thickness of the membrane to reduce the self-
discharging. A thin membrane leads to lower resistance, consequently increasing
the gas crossover and leakage currents. An increase of the membrane thickness
(improving the permeability of the membrane against gas crossover) leads to higher
losses in the membrane, and as a result a decrease in energy efficiency [15,35].
2. The operating temperature must be as low as possible to enhance the protective
function of the membrane against gas crossover. A higher operating temperature
Membranes 2021, 11, 379 14 of 17

leads to lower resistance and consequently contributes to gas crossover, as highlighted

in previous works focused on the effect of the temperature on Faraday’s efficiency [11].
3. The operating pressure must be as small as possible to limit the gas crossover.
4. To avoid the limitation of the operating conditions (pressure, temperature) while
keeping a thinner membrane, the self-discharge can be compensated by supplying
the PEM electrolyzer with a small current (contributing to a constant OCV).

Figure 11. Stack voltage overshoot when connecting to the DC power supply as a result of a self-
discharge.

5. Conclusions
The main objective of this work consisted of introducing to the research field the
phenomenon of self-discharge voltage that occurs in a PEM electrolyzer. Additionally, to
predict the behavior of the self-discharge voltage, an accurate model has been developed.
This model has been developed based on models that have already been reported in the
literature for supercapacitors and batteries (i.e., due to the similarities that these devices
have with PEM electrolyzers).
The model presented in this work has been validated for self-discharge voltage in a
PEM electrolyzer. Besides, this model presents different characteristics and can be useful in
different investigations. Indeed, the model has been developed based on physical reasoning,
taking into consideration the real physical phenomena governing a PEM electrolyzer’s
operation. As a result, this model performs well. However, the model presented in this
work can be improved using different mathematical tools, such as functional analysis or
control theory. Thus, this work is a basis for future research on the development of more
complex models of PEM electrolyzers.
From the analysis performed on the experimental data, recommendations for lowering
self-discharge voltage in a PEM electrolyzer have been proposed, taking into consideration
many variables, such as temperature, membrane, and pressure.
Membranes 2021, 11, 379 15 of 17

Author Contributions: Conceptualization, Á.H.-G., V.R., D.G. and B.S.; methodology, Á.H.-G., V.R.,
D.G. and B.S.; validation, Á.H.-G. and D.G.; investigation, Á.H.-G. and D.G.; writing—original draft
preparation, Á.H.-G. and D.G.; writing—review and editing, Á.H-G., V.R., D.G. and B.S. All authors
have read and agreed to the published version of the manuscript.
Funding: This research received no external funding.
Institutional Review Board Statement: Not applicable.
Informed Consent Statement: Not applicable.
Data Availability Statement: The data presented in this study are available on request from the
corresponding author. The data are not publicly available due to their current utilization for future
works involving the authors of this paper.
Acknowledgments: The authors would like to express their gratitude to the GREEN laboratory of
the University of Lorraine for their constant encouragement and support toward developing research
cooperation between France and Mexico.
Conflicts of Interest: The authors declare no conflict of interest.

Abbreviations
The following abbreviations are used in this manuscript:

OCV Open-Circuit Voltage

PEM Proton Exchange Membrane
RES Renewable Energies Sources
SLPM Standard Liter Per Minute
SO Solid Oxide

References
1. Abbasi, T.; Abbasi, S. ‘renewable’ hydrogen: Prospects and challenges. Renew. Sustain. Energy Rev. 2011, 15, 3034–3040.
doi:10.1016/j.rser.2011.02.026.
2. Mohammadi, A.; Mehrpooya, M. A comprehensive review on coupling different types of electrolyzer to renewable energy
sources. Energy 2018, 158, 632–655. doi:10.1016/j.energy.2018.06.073.
3. Ursúa, A.; Gandía, L.M.; Sanchis, P. Hydrogen production from water electrolysis: Current status and future trends. Proc. IEEE
2012, 100, 410–426. doi:10.1109/JPROC.2011.2156750.
4. Abdin, Z.; Zafaranloo, A.; Rafiee, A.; Mérida, W.; Lipiński, W.; Khalilpour, K.R. Hydrogen as an energy vector. Renew. Sustain.
Energy Rev. 2020, 120, 109620. doi:10.1016/j.rser.2019.109620.
5. Lee, C.-Y.; Chen, C.-H.; Li, S.-C.; Wang, Y.-S. Development and application of flexible integrated microsensor as real-time monitor-
ing tool in proton exchange membrane water electrolyzer. Renew. Energy 2019, 143, 906–914. doi:10.1016/j.renene.2019.05.071.
6. Ruuskanen, V.; Koponen, J.; Sillanpää, T.; Huoman, K.; Kosonen, A.; Niemelä, M.; Ahola, J. Design and implementation of a power-
hardware-in-loop simulator for water electrolysis emulation. Renew. Energy 2018, 119, 106–115. doi:10.1016/j.renene.2017.11.088.
7. Corengia, M.; Torres, A.I. Two-phase dynamic model for pem electrolyzer. In 13th International Symposium on Process Systems
Engineering (PSE 2018); Eden, M.R., Ierapetritou, M.G., Towler, G.P., Eds.; Volume 44 of Computer Aided Chemical Engineering;
Elsevier: Amsterdam, The Netherlands, 2018; pp. 1435–1440. doi:10.1016/B978-0-444-64241-7.50234-2.
8. Nie, J.; Chen, Y.; Cohen, S.; Carter, B.D.; Boehm, R.F. Numerical and experimental study of three-dimensional fluid flow in the
bipolar plate of a PEM electrolysis cell. Int. J. Therm. Sci. 2009, 48, 1914–1922. doi:10.1016/j.ijthermalsci.2009.02.017.
9. Abdin, Z.; Webb, C.J.; Gray, E.M. Modelling and simulation of a proton exchange membrane (PEM) electrolyser cell. Int. J.
Hydrogen Energy 2015, 40, 13243–13257. doi:10.1016/j.ijhydene.2015.07.129.
10. Rahim, A.H.A.; Tijani, A.S.; Kamarudin, S.K.; Hanapi, S. An overview of polymer electrolyte membrane electrolyzer for hydrogen
production: Modeling and mass transport. J. Power Sources 2016, 309, 56–65. doi:10.1016/j.jpowsour.2016.01.012.
11. Ulleberg, O. Modeling of advanced alkaline electrolyzers: A system simulation approach. Int. J. Hydrogen Energy 2003, 28, 21–33.
doi:10.1016/S0360-3199(02)00033-2.
12. Sánchez, M.; Amores, E.; Rodríguez, L.; Clemente-Jul, C. Semi-empirical model and experimental validation for the performance
evaluation of a 15 kW alkaline water electrolyzer. Int. J. Hydrogen Energy 2018, 43, 20332–20345. doi:10.1016/j.ijhydene.2018.09.029.
13. Guilbert, D.; Vitale, G. Experimental validation of an equivalent dynamic electrical model for a proton exchange membrane
electrolyzer. In Proceedings of the 2018 IEEE International Conference on Environment and Electrical Engineering and 2018
IEEE Industrial and Commercial Power Systems Europe (EEEIC/I&CPS Europe), Palermo, Italy, 12–15 June 2018; pp. 1–6.
doi:10.1109/EEEIC.2018.8494523.
Membranes 2021, 11, 379 16 of 17

14. Hernández-Gómez, Á.; Ramirez, V.; Guilbert, D.; Saldivar, B. Development of an adaptive static-dynamic electrical model based on
input electrical energy for pem water electrolysis. Int. J. Hydrogen Energy 2020, 45, 18817–18830. doi:10.1016/j.ijhydene.2020.04.182.
15. Schalenbach, M.; Carmo, M.; Fritz, D.L.; Mergel, J.; Stolten, D. Pressurized PEM water electrolysis: Efficiency and gas crossover.
Int. J. Hydrogen Energy 2013, 38, 14921–14933. doi:10.1016/j.ijhydene.2013.09.013.
16. Koponen, J.; Kosonen, A.; Ruuskanen, V.; Huoman, K.; Niemelä, M.; Ahola, J. Control and energy efficiency of PEM water
electrolyzers in renewable energy systems. Int. J. Hydrogen Energy 2017, 42, 29648–29660. doi:10.1016/j.ijhydene.2017.10.056.
17. Schalenbach, M. Corrigendum to ‘pressurized PEM water electrolysis: Efficiency and gas crossover’ [Int. J. Hydrogen Energy 2013,
38, 14921–14933]. Int. J. Hydrogen Energy 2016, 41, 729–732. doi:10.1016/j.ijhydene.2015.11.009.
18. Conway, B.E.; Pell, W.; Liu, T.-C. Diagnostic analyses for mechanisms of self-discharge of electrochemical capacitors and batteries.
J. Power Sources 1997, 65, 53–59. doi:10.1016/S0378-7753(97)02468-3.
19. Andreas, H.A. Self-discharge in electrochemical capacitors: A perspective article. J. Electrochem. Soc. 2015, 162, A5047–A5053.
doi:10.1149/2.0081505jes.
20. de Fazio, R.; Cafagna, D.; Marcuccio, G.; Visconti, P. Limitations and characterization of energy storage devices for harvesting
applications. Energies 2020, 13, 783. doi:10.3390/en13040783.
21. Olaszi, B.D.; Ladanyi, J. Comparison of different discharge strategies of grid-connected residential pv systems with energy
storage in perspective of optimal battery energy storage system sizing. Renew. Sustain. Energy Rev. 2017, 75, 710–718.
doi:10.1016/j.rser.2016.11.046.
22. Ke, B.-R.; Lin, Y.-H.; Chen, H.-Z.; Fang, S.-C. Battery charging and discharging scheduling with demand response for an electric
bus public transportation system. Sustain. Energy Technol. Assessments 2020, 40, 100741. doi:10.1016/j.seta.2020.100741.
23. Xiao, J.; Guo, J.; Zhan, L.; Xu, Z. A cleaner approach to the discharge process of spent lithium ion batteries in different solutions.
J. Clean. Prod. 2020, 255, 120064. doi:10.1016/j.jclepro.2020.120064.
24. Lin, P.; Jin, P.; Hong, J.; Wang, Z. Battery voltage and state of power prediction based on an improved novel polarization voltage
model. Energy Rep. 2020, 6, 2299–2308. doi:10.1016/j.egyr.2020.08.014.
25. Madani, S.S.; Schaltz, E.; Kær, S.K. A review of different electric equivalent circuit models and parameter identification methods
of lithium-ion batteries. ECS Trans. 2018, 87, 23–37. doi:10.1149/08701.0023ecst.
26. Diab, Y.; Venet, P.; Gualous, H.; Rojat, G. Self-discharge characterization and modeling of electrochemical capacitor used for
power electronics applications. IEEE Trans. Power Electron. 2009, 24, 510–517. doi:10.1109/TPEL.2008.2007116.
27. Wang, B.; Wang, C.; Hu, Q.; Zhang, L.; Wang, Z. Modeling the dynamic self-discharge effects of supercapacitors using a controlled
current source based ladder equivalent circuit. J. Energy Storage 2020, 30, 101473. doi:10.1016/j.est.2020.101473.
28. Yassine, M.; Fabris, D. Performance of commercially available supercapacitors. Energies 2017, 10, 1340. doi:10.3390/en10091340.
29. Navarro, G.; Nájera, J.; Torres, J.; Blanco, M.; Santos, M.; Lafoz, M. Development and experimental validation of a supercapacitor
frequency domain model for industrial energy applications considering dynamic behaviour at high frequencies. Energies 2020, 13,
1156. doi:10.3390/en13051156.
30. Zhang, G.; Wan, X. A wind-hydrogen energy storage system model for massive wind energy curtailment. Int. J. Hydrogen Energy
2014, 39, 1243–1252. doi:10.1016/j.ijhydene.2013.11.003.
31. Ray, A.; Korkut, D.; Saruhan, B. Efficient flexible all-solid supercapacitors with direct sputter-grown needle-like mn/mnox@graphite-
foil electrodes and ppc-embedded ionic electrolytes. Nanomaterials 2020, 10, 1768. doi:10.3390/nano10091768.
32. Morin, B. Hybridation d’une pile à Combustible par des Supercondensateurs: Vers une Solution Passive et Directe, Ph.D. Thesis,
Institut National Polytechnique de Toulouse-INPT, Toulouse, France, 2013. Available online: https://tel.archives-ouvertes.fr/tel-
00821123 (accessed on 21 May 2021).
33. Inaba, M.; Kinumoto, T.; Kiriake, M.; Umebayashi, R.; Tasaka, A.; Ogumi, Z. Gas crossover and membrane degradation in
polymer electrolyte fuel cells. Electrochim. Acta 2006, 51, 5746–5753. doi:10.1016/j.electacta.2006.03.008.
34. Weber, A.Z. Gas-crossover and membrane-pinhole effects in polymer-electrolyte fuel cells. J. Electrochem. Soc. 2008, 155, B521.
doi:10.1149/1.2898130.
35. Papakonstantinou, G.; Sundmacher, K. H2 permeation through N117 and its consumption by IrOx in PEM water electrolyzers.
Electrochem. Commun. 2019, 108, 106578. doi:10.1016/j.elecom.2019.106578.
36. Li, A.; Pelissier, S.; Venet, P.; Gyan, P. Fast Characterization Method for Modeling Battery Relaxation Voltage. Batteries 2016, 2, 7.
doi:10.3390/batteries2020007.
37. An, F.; Zhao, H.; Li, P.; Self-discharge rates in cells have a critical effect on the cycle life of parallel lithium-ion batteries. RSC Adv.
2018, 8, 30802. doi.org/10.1039/C8RA05403G.
38. Hinaje, M.; Raël, S.; Noiying, P.; Nguyen, D.A.; Davat, B. An equivalent electrical circuit model of proton exchange membrane
fuel cells based on mathematical modelling. Energies 2012, 5, 2724–2744. doi:10.3390/en5082724.
39. Tijani, A.S.; Rahim, A.H.A. Numerical modeling the effect of operating variables on Faraday efficiency in PEM electrolyzer.
Procedia Technol. 2016, 26, 419–427. doi:10.1016/j.protcy.2016.08.054.
40. Hernández-Gómez, Á.; Ramirez, V.; Guilbert, D.; Saldivar, B. Cell voltage static-dynamic modeling of a pem elec-
trolyzer based on adaptive parameters: Development and experimental validation. Renew. Energy 2021, 163, 1508–1522.
doi:10.1016/j.renene.2020.09.106.
41. Satpathy, S.; Dhar, M.; Bhattacharyya, B.K. Why supercapacitor follows complex time-dependent power law (tα ) and does not
obey normal exponential (exp[−t/( RC )]) rule? J. Energy Storage 2020, 31, 101606. doi:10.1016/j.est.2020.101606.
Membranes 2021, 11, 379 17 of 17

42. Wang, Y.-Z.; Shan, X.-Y.; Wang, D.-W.; Cheng, H.-M.; Li, F. Mitigating self-discharge of carbon-based electrochemical
capacitors by modifying their electric-double layer to maximize energy efficiency. J. Energy Chem. 2019, 38, 214–218.
doi:10.1016/j.jechem.2019.04.004.
43. Subramanian, S.; Johny, M.A.; Neelanchery, M.; Ansari, S. Self-discharge and voltage recovery in graphene supercapacitors. IEEE
Trans. Power Electron. 2018, 33, 10410–10418. doi:10.1109/TPEL.2018.2810889.
44. Guida, V.; Guilbert, D.; Vitale, G.; Douine, B. Design and realization of a stacked interleaved dc–dc step-down converter for pem
water electrolysis with improved current control. Fuel Cells 2020, 20, 307–315. doi:10.1002/fuce.201900153.