Hidrogenioverde

Pursuant to the DOE Public Access Plan, this document represents the authors' peer-reviewed, accepted manuscript.
international journal of hydrogen energy xxx (xxxx) xxx

The published version of the article is available from the relevant publisher.
Available online at www.sciencedirect.com
ScienceDirect
journal homepage: www.elsevier.com/locate/he
A comprehensive modeling method for proton

exchange membrane electrolyzer development
Zhiwen Ma*, Liam Witteman, Jacob A. Wrubel, Guido Bender

National Renewable Energy Laboratory, 15013 Denver West Parkway, Golden, CO 80401, USA
highlights graphical abstract
Modeling scopes and methods are

presented to facilitate PEMEC
development.
Electrochemical, thermal-fluid,
and transport processes are
integrated.
A PEMEC modeling tool imple-
ments the add-on modules in
ANSYS/Fluent.
Modeling outcomes provide design
and performance insights.
article info abstract
Article history: Hydrogen attracts significant interests as an effective energy carrier that can be derived
Received 29 December 2020 from renewable sources. Hydrogen production using a proton-exchange membrane (PEM)
Received in revised form electrolyzer can efficiently convert renewable power via water splitting in wide scalesd-
20 February 2021 from large, centralized generation to on-site production. Mathematical models with mul-
Accepted 22 February 2021 tiple scales and fidelities facilitate the continuing improvements of PEM electrolyzer
Available online xxx development to improve performance, cost, and reliability. The model scopes and methods
are presented in this paper, which also introduces a comprehensive PEM electrolysis
Keywords: modeling tool based on computational fluid dynamics (CFD) software, ANSYS/Fluent. The
Low temperature electrolysis water modeling tool incorporates electrochemical model of a PEM electrolysis cell to simulate the
splitting performance of coupled thermal-fluid, species transport, and electrochemical processes in
Proton exchange membrane a product-scale cell or stack by leveraging the powerful meshing generation and CFD solver
electrolysis cell of ANSYS/Fluent. The thermal-fluid modeling includes liquid water/gas two-phase flow
Hydrogen production and simulates a PEM electrolysis cell by using Fluent user-defined functions as add-on
Electrochemical modeling modules accounting for PEM-specific species transport and electrochemical processes.
The modeling outcomes expediate PEM electrolyzer scaling up from basic material devel-
opment and laboratory testing.
© 2021 Hydrogen Energy Publications LLC. Published by Elsevier Ltd. All rights reserved.
* Corresponding author.
E-mail address: [email protected] (Z. Ma).
https://doi.org/10.1016/j.ijhydene.2021.02.170
0360-3199/© 2021 Hydrogen Energy Publications LLC. Published by Elsevier Ltd. All rights reserved.
Please cite this article as: Ma Z et al., A comprehensive modeling method for proton exchange membrane electrolyzer development,
International Journal of Hydrogen Energy, https://doi.org/10.1016/j.ijhydene.2021.02.170
Pursuant
2 to the DOE Public Access Plan, this document represents the authors' peer-reviewed, accepted manuscript.
overall water-splitting process is H2O(l) / H2(g) þ 1/2O2(g),

Introduction which results in the coexistence of liquid water and gas pha-
ses on both sides. PEMEC components include flow distribu-
Fuel cell systems have been deployed for various applications, tion channels inside separate plates, porous transport layers,
primarily in the distributed power generation and trans- and a membrane-electrode assembly (MEA) comprised of a
portation sectors [1,2]. The deployment of fuel cell electric proton exchange membrane, anode catalyst layer, and cath-
vehicles and transition to a hydrogen economy have attracted ode catalyst layer. The PEMEC MEA often uses the per-
great attention in obtaining carbon-free hydrogen supplies fluorosulfonic acid polymer Nafion® membrane as the solid
[3,5,6]. Hydrogen is an attractive energy carrier for carbon-free proton-conducting electrolyte. Protons are transported
power generation and transportation [3], and it enables long across the membrane from anode to cathode. Water also
driving distances for long-haul trucks or heavy-duty vehicles crosses the membrane via electroosmotic drag or diffusion,
[1]. Various technologies are available or under development which are both modeled in this work.
for hydrogen production from different energy sources Oxygen is generated at the anode side and mixes with the
including conventional hydrocarbon fuels such as coal or liquid water and water vapor. Electrons are derived from the
natural gas, or carbon-free sources such as electrolysis water electrochemical oxygen evolution reaction (OER) at the anode,
splitting by using renewable electricity. Of various water and travel through an external circuit back to the cathode.
splitting methods, a proton-exchange membrane (PEM) elec- Electrons and protons recombine at the cathode to produce
trolyzer is a promising device that enables a wide range of hydrogen gas. Hydrogen produced at the cathode side will be
hydrogen production scalesdfrom centralized generation to separated from water crossover through the membrane.
distributed supply. However, the cost and performance of PEM Experimentally probing the multi-physics processes within an
electrolyzer limit their broad economic applications, and operating PEMEC, such as species distributions, spatially
continuing improvements are needed for cost reduction and varying current densities, and thermal fluid flow, is nearly
performance improvements [7,8]. impossible. Therefore, modeling tools in various scales are
The PEM electrolyzer technology is on the path for useful to complement experiments. Modeling results can
commercialization, and its performance is under sustaining provide insight and understanding of the component design
improvements. Carmo et al. [8] have provided a comprehen- and material properties on PEMEC performance. Computa-
sive review of the PEM electrolysis technologies and the scope tional modeling can simulate the transport and electro-
of the research status. The review ranges from electrolyzer chemical processes to aid the development of PEMECs
material, components, and modeling, and it discusses the efficiently and cost effectively. Modeling outcomes in various
method and role of the modeling in improving PEM electro- scopes provide details inside an electrolyzer and help opti-
lyzer performance. Fig. 1 shows a cross-section of a single mize designs and operating conditions through so-called
channel and details the different species of transport inside computational experiments.
an electrolysis cell. In an operating electrolyzer cell, two-
phase flow occurs when liquid water enters the cell, and
gaseous hydrogen and oxygen are generated and mixed
within the liquid- or gas-phase water from vaporization and Overview of PEMEC modeling
water crossover. The gas/liquid two-phase flow benefits the
thermal management in PEM electrolyzer but causes PEMEC modeling includes multi-physics and multiscale ap-
complexity in modeling species transport and electrochemical proaches, spanning from membrane/electrode electro-
processes. chemical and catalytic processes to cell or stack performance.
Fig. 1 shows the mass transport and electrochemical pro- To analyze the PEMEC designs from a single cell to a large
cesses within a typical PEM electrolysis cell (PEMEC). The stack, PEMEC models involving thermal-flow distributions can
Fig. 1 e Schematic of a PEMEC with reactant flow and electrochemical processes.
international journal of hydrogen energy xxx (xxxx) xxx 3
be built upon multi-dimensionality, ranging from a lumped Kim et al. [19] developed the first 1-D dynamic model of a high-
parameter or 0-dimensional (0-D) models, 1-dimensional (1- pressure PEM electrolyzer system. Both 0-D and 1-D model can
D), 2-dimensional (2-D), to 3-dimensional (3-D) models. be used to predict the PEMEC system behavior, but unable to
While 1-D and 2-D models can characterize physical processes simulate product level stack performance. The modeling
in some details, 3-D models describe the components and method introduced in this paper using computational fluid
system in their entirety. Furthermore, the transport phe- dynamics (CFD) and electrochemical model built upon
nomena in the MEA components and flow fields are inherently ANSYS/Fluent software, however, leverages the commercial
3-D. thermal fluid solver, and has the potential to solve product-
PEMEC modeling approaches can be also divided into scale 3-D PEMEC stack, even in transient operation.
steady-state or dynamic simulations. A steady-state model Rahim et al. [20] overviewed modeling work on a PEMEC
assumes that all state variables are stable locally and provides electrolyzer and showed the modeling types and scales of
a general assessment of the cell and stack performance. Most various modeling methods. Three approaches summarized in
material analyses and property simulations are steady-state developing PEMEC chemical-physical models include analyt-
models and are often numerically simple and relatively sta- ical modeling, semi-empirical modeling, and mechanistic
ble to solve. They help develop physical and chemical bases modeling based on the temporal and spatial dimensions.
for other modeling scales and scopes. One important focus for Analytical modeling uses simplified assumptions and ap-
modeling is the transport phenomena in the proton- proximations. It represents a basic process model but does not
conductive membrane, which is the essential component of give an accurate prediction of transport processes that occur
PEMECs and has been thoroughly studied by Weber's group within the cell. For example, voltage losses are approximated
[9e14]. A microscale physical-chemical model may facilitate to be due only to the activation and ohmic overpotentials that
material discovery for membranes of high ionic conductivity, are calculated from the electrode/electrolyte overpotentials
tolerance to higher temperature, mechanical durability and without considering the effect of the distributed species con-
chemical stability. As an example of a steady-state stack centrations and cell temperatures in a large-area PEMEC.
model, Chandesris et al. [15] developed numerical 1-D steady- A semi-empirical model can be theoretically derived from
state modeling to study the effects of current density and differential or algebraic equations and characterize experi-
temperature on membrane degradation. mental results in evaluating electrolyzer performance. Harri-
The challenge and uniqueness of PEMEC modeling is the son et al. [16] developed a semi-empirical model of 0-D SS
multilayer structures with multiphase phenomena through conditions to characterize the performance of a 20-cell PEMEC
multiscale processes including the proton-conducting mem- stack [16]. Dale et al. [21] also developed a semi-empirical 0-D
brane, porous transport layers for multispecies transport, and SS model based on thermodynamic principles for determining
interfacial reactions. The gas-liquid interactions are relevant 6-kW PEM electrolysis stack characteristics by curve-fitting
to surface morphology and wettability with bubble formation experimental data to determine model parameters [21].
hindering the gas transport. In addition, large-scale cells and These works used the stack experimental results to derive a
stacks need thermal management associated with the fluid current-voltage relationship on anode and cathode over-
flow. Mathematical modeling is ideal for exploring the PEMEC potential and membrane conductivity to evaluate electrolyzer
designs due to the complexity and non-linearity of the performance on a system level.
transport phenomena and structural characteristics [10,14]. Mechanistic models represent the underlying physics or
Well-developed and verified models provide optimal oper- chemistry principles or characteristics and provide general
ating conditions and conduct sensitivity analysis of design algebraic expressions or differential equations of underline
and operation tradeoffs. phenomena or processes. PEMECs usually consist of several
For renewable hydrogen production, PEMECs can be con- functional layers with different structures and transport and
nected to renewable power that is subject to uncertainty and surface properties. Marangio et al. [22] used an experimental
generation fluctuations. This leads to transient operating analysis of a high-pressure PEMEC stack to validate the
conditions. Thus, using dynamic model to simulates PEMEC detailed theoretical model of the PEMEC system. Their 0-D SS
transient behavior analyzes the changes in uncertain oper- model included activation, ohmic, and diffusion over-
ating conditions. The objective is to evaluate hydrogen pro- potentials and supported the experimental test. They also
duction associated with the electrolyzer operation responding analyzed the resistances of electrodes and plates along with
to electric variations. Early works on PEMEC modeling focused membrane resistance. The water transport through PEM was
on electrolyzer operation, and electrochemical system models accounted for by diffusion, electro-osmotic drag, and perme-
quantitatively elaborate on electrochemical phenomena ation due to the pressure difference between the anode and
€ ¨ n [17] developed a dynamic
occurring in cells [16,17]. Gorgu cathode. Onda et al. [23] developed a mathematical model
PEME model by using Simulink to model 0-D mass and energy comparing efficiency between atmospheric and high-pressure
conservations at the anode and cathode. The model was based electrolysis based on 1-D SS conditions. The model inter-
on simulations and supported experiments under transient preted the voltage-current relationship such that cell voltage
dynamic behavior of a PEM electrolyzer. The 0-D dynamic was described as the sum of the Nernst voltage, resistance
model is useful in characterizing the PEM electrolyzer stack overpotential, and anode and cathode overpotentials, and it
and system performance and can further be used in simu- simulated the effects of temperature and pressure on
lating the condition under grid integration (Awasthi et al. [18]). enthalpy and Gibbs free energy. Mechanistic modeling
Increasing modeling dimension provides a refined spatial methods are useful to characterize a small-scale PEM cell from
resolution of the electrolyzer performance. Toward that goal, basic test data.
Pursuant
Table 1 shows the PEMEC modeling scope from basic
Ohmic losses, membrane water uptake, ionic resistance
Electrode characterization and optimization, activation

components of the electrodes, membrane, porous transport
Cell heat mass balance and parameter distributions.

layer (PTL), cell assembly, stack, and electrolysis system.
Gas species transport. Concentration overpotential
Size of the stack module and the BOP components

Models are used to establish fundamental characteristics that
Optimization of stack thermal, electrochemical

Water flow, crossover, and humidity inside the
occur in PEMECs. The various modeling scopes support
Use test data to determine cell V-i relations.

comprehensive component, cell and stack design, and system
Reaction process and Nernst potential
performance and evaluate their behaviors under different

Model Outcome
Component and system simulations

operating conditions.
Stack component life prediction

The modeling scope can be plotted in a graphic represen-
tation evolved further from Rahim's structural plot and
reconstructed into an interconnected relationship as seen in
Single cell performance
Fig. 2. Fig. 2 illustrates the various paths of the electrochemical
cell and stack modeling. The paths of various combination of
modeling through a block in each ring involving multi-
overpotential
performance
membrane
dimension, Multiphysics, multi-scale, thermal-fluid, trans-

port processes, chemical, and electrochemical relations. The
connection between various modeling layersde.g., a various-
dimensional modeldintegrates one or more types of
chemical-physical models, in analytical, mechanistic, or
3-D model of a PEMEC stack to predict steady-state and
Integrated electrochemical, thermal-flow and energy
empirical expressions for static or dynamic analysis. Fig. 2

Ohmic resistance and membrane ionic conductance
3-D model of current density, temperature, and gas
Electrolyzer integrated into a hydrogen production

Electrode performance model using Butler-Volmer
Nernst potential affected by the species transport
Gas/liquid water flow, diffusion, permeation, and
maps out the various electrochemical cell and stack

3-D component stress and mechanical reliability
modeling approaches to emphasize the numerous permuta-

tions possible when defining a modeling scope. The high-
lighted green blocks are the modeling scope introduced in this
Transport model with porous media
paper.
Features
Fig. 2 shows the modeling applications to complement the

scope of the PEMEC development. Broad studies have been
conducted on PEM electrolyzer component and system
transient characteristics
PEMEC stack operation
development by modeling and testing [27,29e31,33e48].

electro-osmotic drag.
composition profiles
Complementary modeling and testing can support the effec-

tive development of PEM electrolyzers by bridging materials
testing and material development and predicting the scaling-
equations.
equations
up performance. The modeling tool focuses on macroscale

system
modeling cell and stack designs and subsequently

Mechanical reliability model of a PEMEC components in
Ionic conductance in membrane, electric and interfacial
Multi-dimension thermal-flow and current density

Electrochemical model of electrode processes with
Species transport through porous diffusion layers
Stack integration with the balance of plant (BOP)

Water balance through diffusion and transport
System configuration and operating conditions

Electrolyzer stack with different flow field and
CFD model of cells for thermal management

Fundamental electrochemical process and
Table 1 e PEMEC modeling scopes and approaches.
Objectives
thermodynamic model
catalyst kinetics
configurations
distributions
contact
a stack
Material Properties
Model [16,26,27,32]
[8,9,15,22,23,26,27]
Module and System

[17,19,21,28,29]
Modeling and
Membrane and
[5,8e10,20]
References
Stack Model
Scope and
Electrode
Cell Model
Fig. 2 e Map of the electrochemical cell and stack modeling

scopedthe various path combinations for a multi-
dimension, multi-scale, and multi-physics model.
understands the operation with parameters distributions among electrochemical reactions, fluid dynamics, mass
often unable to be measured by experiments. For example, transfer, and multi-phase flow that are coupled simulta-
micro-scale water transport inside the PTL/microporous layer neously by extending the ANSYS/Fluent models with the
(MPL) can only be solved by numerical methods [7,8]. The UDFs, which provides greater flexibility and scope incorpo-
macroscale models also solved the flow field in conjunction rating the material and electrochemical properties than the
with MEA performance. With the multi-scale/multi- built-in electrolyzer model inside the Fluent software [56]. Ma
dimensional modeling being proved for PEM fuel cells et al. [57e60] used this approach in designing a commercial
[49e54], the modeling PEMEC cell or stack using an integrated molten carbonate fuel cell (MCFC) stack. Prior to the MCFC
CFD electrochemical model provides the understanding of modeling, stack thermal and mechanical optimization often
basic transport and thermal-flow effects to improve the took several iterations and required several different testing
performances. stacks to be produced; the model-aided design approach only
Some early CFD models of a PEM electrolyzer lacked the took one verification stack to get to the product, shortening
incorporation of electrochemistry or adequate modeling of the implementation of new components, catalyst, and elec-
gas/liquid water two-phase flow. For example, Nie et al. [16] trode designs. The CFD model can be used to study all the
modeled a two-phase gas-liquid flow-field plate restricted to effects in an electrolyzer and provide effective design simu-
the anode side of a PEMEC. They used numerical 3-D dynamic lation from the constituent materials to cell and stack designs.
simulations to examine individual components of PEMEC and
were more concerned with the flow patterns inside an elec-
trolyzer cell to find the pressure and velocity distribution. PEMEC model description
Since 2010, more details have been included in the simula-
tions, such as a two-phase flow study and the effect on flow To simulate the PEMEC, the mixture model in Fluent was used
patterns [9], thermal balance, and temperature distribution to solve the multiphase flow in the electrode components and
across the cell [10], as well as comprehensive models that was coupled to customized UDFs that model the electro-
simulated both two-phase flow and heating effects to study chemical processes. UDFs are add-on modules that can inte-
the flow-field patterns [11]. However, none of the models grate with the Fluent solver and are used to define source
included a detailed electrochemical model, and typically, pa- terms and material properties. Using UDFs leverages the ca-
rameters such as current density or temperatures were set as pacity and robustness of the Fluent software, while extending
constant or not included in the study. its PEMEC modeling capability.
PEMEC modeling can provide in-depth information from Fig. 3a shows a schematic of the 25 cm2 flow field and
thermal-flow optimization, electrochemical performances, PEMEC that is similar to a laboratory test cell, which was
and cell and stack operation conditions including the reli- studied for the convenience of using experimental data to
ability analysis under a pressurized operation. The modeling validate the model. The component dimensions are included
results can predict the effects of the species flow, operating in the schematic labels in Fig. 3a. A close view of the sand-
current and voltage, the effect of temperature and pressure, wiched cell structure is shown in Fig. 3b. The multiphase flow
and more. Defining the modeling scope and reviewing the was solved throughout the channels and the PTLs, while the
current modeling approaches indicate that an ideal modeling MEA was treated as a thin film boundary. The electrochemical
platform should be flexible and capable of covering the blocks reactions, ionic resistance, and water transport in the MEA
as needed. The blue blocks in Fig. 2 show a path of the current were accounted analytically through a UDF. Using analytical
modeling scope of 3-D static, mechanistic, thermal, and models for the electrochemical processes improves the nu-
electrochemical processes. The modeling method presented merical stability of the model and significantly reduces
in this paper integrated an electrochemical model with user- computational times. Fig. 3c shows the swept mesh which
defined functions (UDFs) and leverages a CFD solver in was used to efficiently discretize the geometry. The mesh
ANSYS-Fluent software. This approach toward a combined consists of just under 500,000 mostly quadrilateral elements
modeling platform accommodates multidimensional with a maximum aspect ratio less than 50 and a skewness
thermal-fluid, multiphase flow, multi-physics processes, and close to 1. The mesh was generated using the ANSYS meshing
the ability to incorporate chemical and electrochemical pro- tool, which can facilitate industry-scale electrolysis cell
cesses within an integrated modeling environmentdall using analysis [60].
UDFs. Such a model can be scaled from a single-channel
PEMEC representative of a full electrolysis cell to a Thermal-fluid and species-transport processes
commercial-scale electrolyzer. The model can be extended to
thermal-mechanical analysis for further investigating the The mixture model in Fluent was chosen due to its accuracy
reliability, durability, and mechanical performance of a and simplicity compared to the Eulerian or Volume of Fluid
PEMEC component. A Multiphysics approach can help opti- methods [61]. In the mixture model, liquid water is consid-
mize the stack design for lower cost and higher performance, ered as a primary (continuous) phase and gas is considered as
which accounts for about 60% of a PEMEC system cost as the secondary (dispersed) phase. The gas is a multicompo-
shown by Babic et al. [55]. nent mixture, comprised of oxygen and water vapor in the
The modeling method presented in this paper focuses anode side, and hydrogen and water vapor in the cathode
primarily on the cell and stack scale toward a goal of side.
comprehensively simulating the performance of an electro- The mixture model is derived by performing Eulerian
lyzer. The model development integrates the relationship averaging on the Navier-Stokes equations to solve a single set
Pursuant
Fig. 3 e Schematics of a) the complete computational geometry, b) zoomed in view of the domains representing various
PEMEC components, and c) the swept mesh used in the solution of the model.
of continuity, momentum, and energy equations for the Momentum equation

mixture [61]. The volume fraction of the secondary (i.e. non- The momentum equations for the mixture are obtained by
continuous) phase is obtained by applying a separate conti- summing the momentum equations for the primary and
nuity equation to the gas mixture and using a set of algebraic secondary phases, as given by Eq. (4). A momentum source
closure equations for the relative velocities. A summary of the term, Smom , is included to account for porous domains, and mm
governing equations is presented in Table 2, with more detail is the mass-averaged viscosity of the mixture:
provided in the subsequent sections. X
mm ¼ ak mk (5)
k
Continuity equation
The continuity equation for the liquid-gas mixture is given by Lastly, the drift velocity, !
v dr;k , is related to the slip velocity,
!
Eq. (1), where ! v m is the mass-averaged velocity v kg ; which is the velocity of phase k relative to secondary
! 1 X
vm¼ ak rk !
vk (2) phase g:
rm k
! X !
v dr;k ¼ !
v kg xk v kg (6)
!
v k is the velocity of phase k, ak is the volume fraction, rk is the k
density, and rm is the mixture density, expressed as:

where xk is the mass fraction of phase k. The slip velocity
X formulation from Manninen et al. is expressed as [61]:
rm ¼ ak rk (3)
k
Table 2 e Governing equations used in the mixture model.

Equation Eq. #
Continuity V,ðrm !
v m Þ ¼ Sm (1)
!
Momentum P (4)
V ,ðrm !
v m! Vp þV ,ðmm ðV!
v m þ! ak rk !
v dr;k !
T
v mÞ ¼ v m ÞÞ þ V , v dr;k þSmom
k
Phase Volume Fraction V,ðag rg !

v mÞ ¼ V,ðag rg !
v dr;g Þ þ m_ lg m_ gl (11)
!
Energy P ! PP ! (12)
hj;k J j;k þðteff !
¼
V, ðak v k ðrk Ek þpÞÞ ¼ V , keff VT v Þ þSh
k k j
Species V,ðctot !
v yi Þ ¼ V,Ji þ Si (15)
effective thermal conductivity of the mixture, calculated

! tp rg rm !
v kg ¼ a (7) as
fdrag rg
X
where tp is the particle relaxation time, keff ¼ ak kk (13)
k
rg d2g Sh In Eq. (12) is a volumetric generation term that is defined

tp ¼ (8)
18mk in section. For a compressible gas, the internal energy Ek is
and dq is the particle diameter of bubbles of phase q, and fdrag is expressed as:
the drag function from Schiller and Naumann [62]:
p v2k
Ek ¼ hk þ (14)
fdrag ¼ 1 þ 0:15Re0:687 (9) rk 2
The acceleration, !
where hk is the sensible enthalpy for phase k.
a , is expressed as:
An energy source term is included and defined in Table 3, that
!
a¼ !
vm ,V !
vm (10) accounts for entropic heating from the electrochemical re-
actions and the latent heat associated with the phase change of
In the porous regions, i.e. the PTLs, a momentum source water. At the outside of the anode and cathode bipolar plates, a
term is included that accounts for the mixture's interactions Dirichlet boundary condition was applied, setting the tempera-
with the solid phase. The momentum source term, defined in ture was set to the cell temperature. Heat transfer in solid portion
Table 3, includes a viscous loss term (i.e. a Darcy's law term) of the cell is through heat conduction only and interacts with
and an inertial loss term. fluid phase through thermal boundary layer that are solved as a
No-slip boundary conditions were applied at all solid conjugate heat transfer problem by Fluent.
boundaries, e.g. the channel walls and bipolar plate ribs. The
given cell flow rates were prescribed at the channel inlets, and Species-transport equations
pressure boundary conditions were set at the outlets. Species transport within the gas phases through the PTLs on the
anode and cathode sides respectively is modeled by a modified
Volume fraction equation convection-diffusion equation and solved in whole domain
The volume fraction of the secondary phase, ag , can be including the transport processes inside the flow channels. In
found by applying a separate continuity equation to the Eq. (15), the species concentrations are related to their partial
gas phase, given by Eq. (11). The volume fraction equation pressures as ci ¼ pi ðRTÞ 1 , and the mole fractions are defined as
describes the transport of the secondary phase due to the ! 1
P
bulk convection of the mixture, as well as the individual yi ¼ pi pk . The diffusive flux is represented by Fick's law
convection of the gas phase. The m_ lg and m_ gl terms reflect k
the (volumetric) mass transfer from the liquid phase to the through the species concentration gradient,
gas phase, and from the gas phase to the liquid, respec-
Ji ¼ ctot Di;eff Vyi (16)
tively. With only two phases, the volume fraction of the
P
liquid phase can be found as al ¼ 1 ag . where ctot ¼ i ci , and Di;eff is the effective diffusion coefficient
of species i in the mixture, which is modified to account for the
Energy equation multicomponent interactions.
The energy equation for the mixture, given by Eq. (12), is ob- Species source terms are also included and defined in Table
tained by summing the energy conservation equations for 3, that account for species production from the electro-
each phase. The energy equation considers the conjugate heat chemical reactions.
transfer within the solid portions of the cell as well as the
multiphase fluid regions. Membrane water transport
!
In Eq. (12), hj;k is enthalpy of species j in phase k, J j;k is Water from the anode channel supplies the reactant for
the diffusive flux of species j in phase k, and keff is the electrolysis and hydrates the PEM membrane. Water transport
Table 3 e Source terms implemented in conservation equations.

Source Terms Anode Cathode
Mass: liquid phase i Sm ¼ N_ H2 O;m N_ H2 O;evap
Sm ¼ N_ H2 O;m N_ H2 O;evap MH O
2F 2
Mass: vapor phase i i
Sm ¼ MO þ N_ H2 O;evap Sm ¼ MH þ N_ H2 O;evap
4F 2 2F 2
Momentum m!
Smom ¼ v
k
Energy N_ H2 O;evap
Sh ¼ ðVcell 1:48VÞi DHvap
MH2 O
Species: H2 SH 2 ¼ 0 iMH2
SH 2 ¼
2F
Species: O2 i SO 2 ¼0
SO 2 ¼ MO
4F 2
Pursuant
through the membrane is driven by diffusion, electro-osmotic

drag, and pressure gradients [22]: RT 1 i
hAact ¼ sinh (23)
aF 2RACL i0;OER
Jw;tot ¼ Jw;diff þ Jw;eo þ Jw;p (17)
The diffusion contribution arises from gradients in the RT 1 i

hCact ¼ sinh (24)
activity of water across the membrane. The diffusive flux of aF 2RCCL i0;HER
water from the anode to the cathode electrode can be
where a are the charge transfer coefficients, RACL and RCCL are
expressed by:
the roughness factors for the ACL and CCL, respectively, and
Dw i0;OER and i0;HER are the exchange current densities for the OER
Jw;diff ¼ cC cAH2 O;mem (18)
dmem H2 O;mem and HER, respectively. The roughness factors represent the
ratio of actual surface area to geometric surface area of the
where Dw is the diffusion coefficient of water in the mem-
catalyst layers and were determined from fitting experimental
brane, and dmem is the membrane thickness.
data of PEMEC performance described in Result and
Electro-osmotic drag refers to the phenomena by
discussion. While the full Butler-Volmer equation could be
which water molecules are drawn through the mem-
used to model the kinetics, the benefit of Eqs. (23) and (24) is
brane by migrating protons. The flux of water is related
that they can be solved analytically for the overpotential. The
to the current density (which is assumed to be provided
exchange current densities, represent the intrinsic kinetics of
entirely by proton migration) through the electro-osmotic
the electrochemical reactions, were obtained from the litera-
drag coefficient, nd , as:
ture as:
i
Jw;eo ¼ nd (19) 28920:95 1 1
F i0;OER ¼ 2:83792 10 7
exp (25)
8:314 T 303:15
The electro-osmotic drag coefficient refers to the number
of water molecules dragged by each proton, and is a function 17000 1 1
of the membrane water content as [63]: i0;HER ¼ 2:15 10 2
exp (26)
8:314 T 303:15
l from Refs. [24,68], respectively.
nd ¼ 2:5 (20)
22
The membrane process considered PEMECs that are fed Concentration overpotential ðhconc Þ
liquid water on both flow channels adjacent to the anode and The concentration overpotential, hconc , is caused by mass-
cathode. Thus, the membrane is assumed to remain fully transport resistances through the CLs and PTLs. The electro-
hydrated, i.e. l ¼ 22, during operation, which affects the chemical reactions produce gases that must be transported
membrane conductivity in Eq. (32) below. out of the catalyst layers through the PTLs and into the flow
channels [25]. The accumulation of gaseous species in the
PEMEC electrochemical model catalyst layers of anode and cathode leads to an overpotential
that can be expressed as:
To model the cell performance in terms of voltage-current !
relation, the overall cell voltage applied from an electric RT cO
hAconc ¼ ln 0 2 (27)
source is broken down to various terms and equal to the sum 4F cO2
of individual overpotentials:
!
RT cH
V ¼ EOCV þ hact þ hconc þ hohm (21) hCconc ¼ ln 0 2 (28)
2F cH2
where EOCV is the open-circuit potential, hact is the activation
overpotential, hconc is the concentration overpotential, and hohm where the c0O2 and c0H2 terms denote equilibrium conditions,
is the ohmic overpotential. The definitions of each term in Eq. usually corresponding to the conditions at which the ex-
(21) describe the PEMEC processes and provide an electro- change current density was measured. The local species
chemical modeling basis. concentrations, cO2 and cH2 , are found from the solution of Eq.
In the absence of gas crossover and/or other nonidealities, (15) by the Fick's law of species diffusion.
the open circuit voltage (OCV) can be expressed as [64]:
! Ohmic overpotential ðhohm Þ
1=2
RT pH2 pO2 Ohmic overpotentials, hohm , stem from the electric and ionic
EOCV ¼ 1:229 0:9 10 ðT 3
298:15Þ þ ln (22)
2F pH2 O resistances that the current in the cell experiences while
traveling from one current collector to the other. Each
Activation overpotential ðhact Þ component in the cell occurs in series, and so the total resis-
To model the activation overpotential, we assume symmetric tance of the cell, Rtot , is equal to the sum of the individual
charge transfer coefficients, which allows the Butler-Volmer resistances. Ohm's law was used to find the overpotential as a
equation to be simplified as [18,19,22,65e67]: function of the total resistance and current density:
hohm ¼ IRtot ¼ IðRe þ RI Þ (29)
where the total resistance includes electronic ðRe Þ and ionic

ðRI Þ contributions.
The method for calculating the total electronic resistance
through the cell components follows the approach of Mar-
angio et al. [22], who developed an equivalent circuit to
represent the various electrode components. The equivalent
circuit includes contributions from the catalyst layer, PTL, and
flow plate, and represents the total electronic resistances as
Re ¼ RAe þ RCe þ Rcontact (30)
The ionic resistance accounts for the voltage drop required

to conduct protons across the membrane, and can be
expressed in terms of the membrane conductivity as:
dmem
RI ¼ (31)
smem A
where dmem is the membrane thickness and A is the cross-
sectional area of the cell. The ionic conductivity of the mem-
brane is a function of the water uptake, l, and temperature
with respect to a reference temperature (303 K), and can be
calculated as [63]:
1 1
smem ¼ ð0:005139l 0:00326Þexp 1268 (32)
303 Tavg
Higher membrane water content facilitates proton trans-

port and results in higher conductivity values.
Source terms in the conservation equations from electrochemical

processes
Source terms are volumetric scalar quantities that are incor-
porated into the conservation equations. Table 3 defines the
source-term expressions integrated in the mass, species, en- Fig. 4 e Flow chart of the numerical algorithm used for
ergy, and momentum source terms in the conservation solving the coupled electrochemistry and CFD equations.
equations of Eqs. (4),(11),(12) and (15).
The source terms connect the electrochemical processes
with thermal-fluid and transport model and balance the This allows the spatially varying current distribution to be
conservation equations. They were incorporated within the captured over the length of the channel.
Fluent solver through the UDFs and solved simultaneously Once the current density is obtained, the mass, mo-
with the conservation equations. mentum, and energy equations, are iteratively solved. Lastly,
the species equations are solved, and the fluid properties are
Numerical solution method updated. Iterations continue until all convergence criteria are
A flow chart of the numerical algorithm is presented in Fig. 4. met.
First, all variables in the computational domain were initial- Generally, in an electrochemical system, the response of
ized and computed without electrochemical modeling to current versus applied voltage is rapid compared to the
establish a stable thermal-flow field. Then, the electro- thermal-fluid processes. Therefore, the electrochemical
chemical equations were solved using a numerical algorithm components can be modeled as quasi-static, and the dy-
written in the form of a UDF. In the UDF, Eq. (21) is expressed namic equations only applied to the thermal-fluid regions.
in residual form, according to Eq. (33): This work only demonstrates steady state modeling; how-
ever, the method can be readily extended to transient
E þ hact ðiÞ þ hohm ðiÞ þ hdiff ðiÞ V y0 (33)
conditions.
Newton's method is then used to solve for the current The model was developed in ANSYS/Fluent version 19.2. A
density distribution, i, while the applied voltage, V, is applied coupled pressure-velocity solver was used with a first-order
as an input in each cell operating condition. The electro- upwind scheme for all transport equations. To improve
chemical UDF is fully coupled to the mixture model equations computational time and convergence, an algebraic multi-grid
and is called at each iteration and applied to each element in F-cycle method with biconjugate gradient stabilization was
the electrodes. Likewise, the source terms, which are a func- utilized for all equations. The main criteria for convergence
tion of the current density, are also updated at every iteration. were that the average solved current density stabilized for 200
Pursuant
Table 4 e Nominal operating and simulation parameters.

Parameter Value
Low Cell Voltage 1:45 V
High Cell Voltage 1:90 V
Inlet temperature 60 C or 80 C
Inlet water flowrate 50 mL/min
Outlet Pressure 101325 Pa
RACL 1
573:0 m
RCCL 1
32:9 m
Rcontact 8:99 10 3
U
iterations and the total mass imbalance was under 0.5% for
the example case.
Results and discussion
The electrochemical parameters used in the CFD model were

obtained by fitting experimental polarization curves obtained
at 60 C and 80 C. Experimental conditions used for the vali-
dation studies in this work are summarized in Table 4.
The parameters in the electrochemical model were derived
from a least-squares fitting method to fit Eq. (33) to the
experimental data with the following parameters as variables:
RACL (anode catalyst layer roughness factor), RCCL (cathode
catalyst layer roughness factor), and Rcontact (contact resis-
tance). Theses parameters were then used directly in the 3D
CFD model to generate polarization curves and to validate the
obtained parameters. Polarization curves from the 3D model
are shown in Fig. 5, and the electrochemical parameters are Fig. 6 e Overpotential breakdown at cell operating
listed in Table 4. Fig. 5 shows that the obtained parameters temperatures of a) 60 C and b) 80 C.
result in a good match to the experimental data, even when
translated from the electrochemical model to the 3D cell
model. the 60 C data shown in Fig. 6a and the 80 C data shown in
To gain more insight into PEMEC performance, the polari- Fig. 6b. The anode activation overpotential for OER is the
zation curves in Fig. 5 can be expanded to show the individual largest contributor to the cell overpotential, followed by ohmic
overpotential contributions. Fig. 6 shows the contributions overpotential. However, while the activation losses are rela-
from each overpotential at different current densities, with tively constant after 0:5 A=cm2 , ohmic losses continue to grow
linearly. Therefore, reducing the ionic and electrical re-
sistances, including interfacial contact resistances, can
significantly improve high current density operation.
A main benefit of a modeling approach is that it can be
used to evaluate the current, temperature, and phase distri-
butions in an operating electrolysis cell. These results are
difficult or impossible to measure experimentally and can
provide useful insight for the design of cell components.
Nominal operating conditions are shown in Table 4; all results
presented herein use these values unless specifically
mentioned otherwise.
The current-voltage relation shown in Fig. 5 was obtained
by averaging the current density over the total geometric area.
However, the model solves for the spatial distributions of this
parameter, which can be observed to gain insight to the cell
performance. Fig. 7 shows the spatially varying current distri-
butions across the anode side MEA for low applied voltage
(1.45 V, Fig. 7a) and a high applied voltage (1.90 V, Fig. 7b) cases,
Fig. 5 e Simulated voltage-current density relation both operated at 80 C. At both voltages, the regions near the
obtained by fitting the electrochemical model to inlet exhibits higher current densities because of the initial
experimental data. gas-free water flow. After the inlet region, the current density
Fig. 7 e Current density distribution at 80 C across the MEA for the a) low cell potential case, b) high cell potential case; the
same results are shown in c) and d) but with the color scale zoomed in; concentration dependent OCVs are also shown for e)
low cell potential, and f) high cell potential. (For interpretation of the references to color in this figure legend, the reader is
referred to the Web version of this article.)
gradually decreases along the flow channel. The colormap regions of the MEA under the lands. Also, the current density
ranges in Fig. 7a and b reflect the extremes of the current tends to be higher underneath the outside paths of the triple
density variations, which obscures some of the more subtle serpentine pattern. This is most likely because the inner paths
current variations. Therefore, Fig. 7c and d also plot the same accumulate more the gas phase because they are exposed to
current density distributions, but at more refined ranges. This more of the MEA. This can be thought of as a boundary layer
provides a more detailed view of the current distribution, type of effect, in which the channel accumulates the gas phase
especially at the higher voltage. For instance, Fig. 7d shows that through convection. The accumulation of gas phase contrib-
the current is higher under the channels compared to the utes to the current density variation via a concentration effect.
Pursuant
This is quantified by the OCV, Eq. (22), which is plotted in Fig. 7e represent a heat sink. On the other hand, the high potential
and f. As noted, the inlet contains the least amount of gas (1.9 V) case is operated above the thermoneutral voltage, so
phase, and so has the lowest OCV. Note that the spatial varia- the electrochemical reactions represent a heat source. This,
tion of the OCV matches the variation of the current density combined with the boundary layer effect of accumulating gas
shown in Fig. 7a and b. These results demonstrate that a key phase, is why the temperature increases in the straight sec-
function of the flow channels is to transport the gas phase away tions of Fig. 8b.
from the MEA promptly to maintain high water concentration Figs. 7 and 8 demonstrate how key performance parame-
or the partial pressure of reactants. The increase in current ters depend on the distribution of the gas and liquid phases.
density near the flow outlet in Fig. 7d is likely due to the tem- Fig. 9 plots the gas phase distribution in both the flow chan-
perature increase, which improves ionic conductance. nels and PTLs. Fig. 9a and b are surface plots for the low and
Fig. 8a and b shows the temperature distribution along the high voltage cases, respectively, and Fig. 9c and d are volume
MEA for the low voltage and high voltage cases, respectively. rendered averages for the low and high voltage cases,
For the low voltage case, Fig. 8a, the temperature change is respectively. The volume rendered plots are a way to visualize
minimal, due to the low current density. For the higher voltage the average quantity throughout the volume, which can be
case, Fig. 8b, the mixture temperature increases by about 3 C obscured in surface plots.
by the time it reaches the outlet, due to the heat generated by In general, the gas volume fraction increases from as the
the electrochemical reactions. In Fig. 8a, the temperature is channels progress from inlet to outlet, corresponding to the
slightly higher near the edges of the cell (i.e., under the bends variations in current density and OCV observed in Fig. 7. In the
of the flow channels), whereas in Fig. 8b the temperature is high voltage case, the mixture is almost 90% gas phase at the
slightly lower. These cold and hot spots correspond to the outlet, corresponding to the significant consumption of liquid
regions of moderately higher current density shown in Fig. 7c and production of gas. The gas phase also tends to accumulate
and d. Note that the maximum temperature in Fig. 8a is equal near the edges of the cell, since gas that is produced there has
to the inlet temperature. The slight cooling effect observed in fewer directions in which it can escape. Similarly, in Fig. 9c
the straight sections is because the low cell potential (1.45 V) is and d we can observe the gas accumulation under the lands in
lower than the thermoneutral voltage for water splitting the PTLs, as this gas must be transported in-plane before it can
(1.48 V), which means that the electrochemical reactions leave via a channel. These results demonstrate the
Fig. 8 e Temperature profiles across the MEA for the a) high cell potential case, and b) low cell potential case, as well as the
total mixture temperature in the flow channels for the c) high cell potential case, and d) low cell potential case.
Fig. 9 e Surface plots of the gas volume fraction distribution for the a) high cell potential case, b) low cell potential case, and
volume rendered (averaged) gas volume fractions for c) high cell potential, and d) low cell potential.
Fig. 10 e 2D cross-sectional distribution of the gas phase volume fraction in the anode channel and PTL at a) low voltage,
and b) high voltage, with gas phase velocity streamlines superimposed.
Pursuant
importance of PTL design. Thinner PTLs with high perme- the assumptions and simplifications of the electrochem-
abilities can improve device performance by enhancing the istry and two-phase flow still provide adequate accuracy
removal of gaseous products from the cell. Removing the required to predict cell performance. This modeling
products more efficiently can help reduce concentration los- framework can facilitate component designs e including
ses and allow water to reach the MEA more easily, thereby the electrochemical outcomes from cell electrodes in con-
reducing ohmic losses. tacting polymer membranes, porous transport layers, and
A fundamental result of the mixture model is the distri- separate plates with fluid field e to cell and stack perfor-
bution of gas phase volume fraction, which is shown in Fig. 10 mance optimization. The model provides a comprehensive
at different voltages, plotted along cross section of an inlet analysis tool that can be used for future PEMEC develop-
channel in Fig. 3b. Superimposed on the surface plots are ment and component optimization.
streamlines of the gas phase velocity. Near the inlet, the PTL is The model shows prediction capabilities on parameter
mostly filled with liquid water. Further down the channels the distributions that are difficult or impossible to measure
oxygen produced by the OER accumulates more, and results in experimentally but provide cell performance insights with the
the development of a gas phase boundary layer. The wavelike capability to model different flow-field configurations. The
structures indicate a type of Kelvin-Helmholtz instability, analysis of model simulation results identified key areas for
which is a type of shear instability driven by the velocity dif- future research in supporting the operation at high current
ference between the liquid and gas phases. The normal densities. The model can be further advanced for simulating
laminar boundary layer development is interrupted by the product stack operation and thermal-mechanical aspects of a
consumption of liquid water in the anode catalyst layer, PEMEC design in accelerating material advancement, PEMEC
which requires liquid water to be transported through the scaling up, and product improvements.
boundary layer. This gives rise to small ripples in the bound-
ary layer, which are then sheared toward the channel outlet.
The ripples in the gas phase velocity streamlines demonstrate Declaration of competing interest
the competition between inertial forces in the vertical direc-
tion and shear forces in the horizontal direction. At higher The authors declare that they have no known competing
voltages, the wave structures are larger, and somewhat more financial interests or personal relationships that could have
stable. As the voltage is increased, the mixture in the outlet appeared to influence the work reported in this paper.
tends to be mostly gas (by volume), due to the increased OER
production.
The 25 cm2, three-serpentine laboratory-scale PEMEC Acknowledgements
model demonstrated a basic case of the combined electro-
chemical process with thermal-flow and transport phenom- This work was authored by the National Renewable Energy
ena. The modeling tool can be applied to full scale cell and Laboratory (NREL), operated by Alliance for Sustainable En-
stack development to predict the performance of a PEMEC ergy, LLC, for the U.S. Department of Energy (DOE) under
electrolyzer. The model-facilitated product design can expe- Contract No. DE-AC36-08GO28308. The work was partially
dite the industry development, which often employs a trial supported by the NREL H2@scale Laboratory Directed
and error approach that can take a longer time and more re- Research and Development (LDRD) led by Dr. Jennifer Kurtz
sources. In addition, the ANSYS software seamlessly connects and Dr. Bryan Pivovar. The authors gratefully acknowledge
thermal modeling outcomes with mechanical analysis, and research support from the HydroGEN Consortium, established
thus can facilitate the component mechanical reliability for under the U.S. DOE, Office of Energy Efficiency and Renewable
lifecycle and performance degradation evaluation. Energy (EERE), Hydrogen and Fuel Cell Technologies Office
(HFTO). The manuscript was prepared with the funding sup-
port from DOE EERE FCTO FOA Award under contract no. DE-
Conclusion EE-0008426 led by Prof. Fengyuan Zhang at the University of
Tennessee Knoxville. We thank for Dr. Clare Davis-Wheeler
Mathematical modeling provides effective insights in Chin for her literature review and Mr. Will Callahan for
PEMEC and product development on various scales and participating in early model development.
approaches. After comparing various modeling scales and The views expressed in the article do not necessarily
scopes, this paper introduces a computationally efficient represent the views of the DOE or the U.S. Government or any
modeling framework that encompasses adequate physical agency thereof. Neither the United States Government nor
understanding of a PEMEC required to study the effects of any agency thereof, nor any of their employees, makes any
operating conditions and component design on PEM cell or warranty, expressed or implied, or assumes any legal liability
stack performance. The multi-scale, Multiphysics model or responsibility for the accuracy, completeness, or usefulness
incorporates the ANSYS/Fluent CFD solvers and adds elec- of any information, apparatus, product, or process disclosed,
trochemical module to solve transport and chemical pro- or represents that its use would not infringe privately owned
cesses for high-fidelity modeling application. The model rights. The U.S. Government retains and the publisher, by
was applied to a laboratory scale PEMEC with three- accepting the article for publication, acknowledges that the
serpentine flow field to illustrate the ANSYS/Fluent capa- U.S. Government retains a nonexclusive, paid-up, irrevocable,
bility in solving complex geometry. The results show that worldwide license to publish or reproduce the published form
of this work, or allow others to do so, for U.S. Government Constants

purposes. Dw Water diffusion coefficient in membrane,
1.28 10 10 m2/s
Km Water permeability in membrane, 1.58 10 18 m2
Nomenclature nd Electro-osmostic drag coefficient, 2.5 molH2O/molHþ
List of symbols references

A Cell area, m2 or cm2
a Species activity
C Species molar concentration, mol/m3 [1] Rustagi N, Satyapal S. By bryan pivovar, neha rustagi, and
D Diffusion coefficient, m2/s sunita satyapal. 2018.
E Nernst potential, V [2] Satyapal S. DOE hydrogen and fuel cells program: H2@Scale
Activities. In: Natl. Fuel cell hydrog. Washington, DC: Forum;
Ea Activation energy, J/mol
2018. 2018.
F Faraday's constant, 96,485 C/mol [3] Carter M, Ryan MWJ. The fuel cell today industry review. Fuel
h Volumetric Enthalpy, J/m3 Cell Ind Rev 2012:1e52.
i Current density, A/m2 [5] Irena. Hydrogen from renewable power: technology outlook
i0 Exchange current density, A/m2 for the energy transition. Abu Dhabi: International
J Species diffusion flux, kg/(m2.s) Renewable Energy Agency; 2018.
[6] Fraile D, Lanoix J-C, Maio P, Rangel A, Torres A. Overview of
k Thermal conductivity, W/(m.K)
the market segmentation for hydrogen across potential
M Molecular molar mass, kg/kmol
customer groups, based on key application areas. CertifHy
N_ Molar diffusion flux, mol/(m2.s) 2015:1e32. https://doi.org/10.1155/2013/789705.
p Pressure, Pa [7] Schmidt O, Gambhir A, Staffell I, Hawkes A, Nelson J, Few S.
R Gas constant, 8.314 J/(mol.K) Future cost and performance of water electrolysis: an expert
S Volumetric source term elicitation study. Int J Hydrogen Energy 2017;42:30470e92.
T Temperature, C or K https://doi.org/10.1016/j.ijhydene.2017.10.045.
[8] Carmo M, Fritz DL, Mergel J, Stolten D. A comprehensive
V Cell potential, V
review on PEM water electrolysis. Int J Hydrogen Energy
v Velocity, m/s 2013:4901e34. https://doi.org/10.1016/
Y Mass fraction j.ijhydene.2013.01.151.
z Number of electrons transferred in reaction [9] Weber AZ, Borup RL, Darling RM, Das PK, Dursch TJ, Gu W,
et al. A critical review of modeling transport phenomena in
Subscripts polymer-electrolyte fuel cells. J OfThe Electrochem Soc
m Mixture 2014;161:1254e99. https://doi.org/10.1149/2.0751412jes.
k Phase [10] Gostick JT, Weber AZ. Resistor-network modeling of ionic
p Secondary phase conduction in polymer electrolytes. Electrochim Acta
i Species 2015;179:137e45. https://doi.org/10.1016/
j.electacta.2015.03.126.
0 Equilibrium
[11] Abdin Z, Webb CJ, Gray EM. Modelling and simulation of a
w Water
proton exchange membrane (PEM) electrolyser cell. Int J
act Activation polarization Hydrogen Energy 2015;40:13243e57. https://doi.org/10.1016/
avg Average j.ijhydene.2015.07.129.
an Anode [12] Shi S, Weber AZ, Kusoglu A. Structure-transport relationship
cat Cathode of perfluorosulfonic-acid membranes in different cationic
diff Diffusion forms. Electrochim Acta 2016;220:517e28.
[13] Kusoglu A, Weber AZ. New insights into perfluorinated
eff Effective
sulfonic-acid ionomers. Chem Rev 2017;117:987e1104.
evap Evaporation https://doi.org/10.1021/acs.chemrev.6b00159.
mem Membrane [14] Pant LM, Weber AZ. Communication d modeling polymer-
ohm Ohmic resistance electrolyte fuel-cell agglomerates with double-trap kinetics. J
ref Reference state OfThe Electrochem Soc 2017;164:3102e4. https://doi.org/
10.1149/2.0111711jes.
Greek [15] Chandesris M, Medeau V, Guillet N, Chelghoum S, Thoby D,
a Charge-transfer coefficient Fouda-Onana F. Membrane degradation in PEM water
d Thickness, m or mm electrolyzer: numerical modeling and experimental evidence
ε Gas diffusion layer porosity of the influence of temperature and current density. Int J
Hydrogen Energy 2015;40:1353e66. https://doi.org/10.1016/
h Overpotential, V
J.IJHYDENE.2014.11.111.
l Water uptake coefficient of hydration
[16] Harrison KW, Hernandez-Pacheco E, Mann M, Salehfar H.
m Viscosity, Pa.s Semiempirical model for determining PEM electrolyzer stack
y Stoichiometric coefficient of a species characteristics. J Fuel Cell Sci Technol 2006;220:3. https://
s Conductivity, S/m doi.org/10.1115/1.2174072.
r Density, kg/m3 € ¨ n H. Dynamic modelling of a proton exchange
[17] Gorgu
t Stress tensor, Pa membrane (PEM) electrolyzer. Int J Hydrogen Energy
2006;31:29e38. https://doi.org/10.1016/j.ijhydene.2005.04.001.
Pursuant
[18] Awasthi A, Scott K, Basu S. Dynamic modeling and [33] Kang Z, Mo J, Yang G, Li Y, Talley DA, Han B, et al.
simulation of a proton exchange membrane electrolyzer for Performance modeling and current mapping of proton
hydrogen production. Int J Hydrogen Energy exchange membrane electrolyzer cells with novel thin/
2011;36:14779e86. https://doi.org/10.1016/ tunable liquid/gas diffusion layers. Electrochim Acta
j.ijhydene.2011.03.045. 2017;255:405e16. https://doi.org/10.1016/
[19] Kim H, Park M, Lee KS. One-dimensional dynamic modeling j.electacta.2017.09.170.
of a high-pressure water electrolysis system for hydrogen [34] Bromberger K, Ghinaiya J, Lickert T, Fallisch A, Smolinka T.
production. Int J Hydrogen Energy 2013;38:2596e609. https:// Hydraulic ex situ through-plane characterization of porous
doi.org/10.1016/j.ijhydene.2012.12.006. transport layers in PEM water electrolysis cells. Int J
[20] Abdol Rahim AH, Tijani AS, Kamarudin SK, Hanapi S. An Hydrogen Energy 2018;43:2556e69. https://doi.org/10.1016/
overview of polymer electrolyte membrane electrolyzer for j.ijhydene.2017.12.042.
hydrogen production: modeling and mass transport. J Power [35] Van Der Merwe J, Uren K, Van Schoor G, Bessarabov D.
Sources 2016;309:56e65. https://doi.org/10.1016/ Characterisation tools development for PEM electrolysers.
j.jpowsour.2016.01.012. Int J Hydrogen Energy 2014;39:14212e21. https://doi.org/
[21] Dale NV, Mann MD, Salehfar H. Semiempirical model based 10.1016/j.ijhydene.2014.02.096.
on thermodynamic principles for determining 6 kW proton [36] Rakousky C, Reimer U, Wippermann K, Carmo M, Lueke W,
exchange membrane electrolyzer stack characteristics. J Stolten D. An analysis of degradation phenomena in polymer
Power Sources 2008;185:1348e53. https://doi.org/10.1016/ electrolyte membrane water electrolysis. J Power Sources
J.JPOWSOUR.2008.08.054. 2016;326:120e8. https://doi.org/10.1016/
[22] Marangio F, Santarelli M, Calı̀ M. Theoretical model and j.jpowsour.2016.06.082.
experimental analysis of a high pressure PEM water [37] Chandesris M, Medeau V, Guillet N, Chelghoum S, Thoby D,
electrolyser for hydrogen production. Int J Hydrogen Energy Fouda-Onana F. Membrane degradation in PEM water
2009;34:1143e58. https://doi.org/10.1016/ electrolyzer: numerical modeling and experimental evidence
j.ijhydene.2008.11.083. of the influence of temperature and current density. Int J
[23] Onda K, Murakami T, Hikosaka T, Kobayashi M. Hydrogen Energy 2015;40:1353e66. https://doi.org/10.1016/
Performance analysis of polymer-electrolyte water j.ijhydene.2014.11.111.
electrolysis cell at a small-unit test cell and performance [38] Mohamed B, Alli B, Ahmed B. Using the hydrogen for
prediction of large stacked cell. https://doi.org/10.1149/1. sustainable energy storage: designs, modeling, identification
1492287; 2002. and simulation membrane behavior in PEM system
[24] Zenyuk IV, Das PK, Weber AZ. Understanding impacts of electrolyser. J Energy Storage 2016;7:270e85. https://doi.org/
catalyst-layer thickness on fuel-cell performance via 10.1016/j.est.2016.06.006.
mathematical modeling163; 2016. p. 691e703. https://doi.org/ [39] Sartory M, Wallno € fer-Ogris E, Salman P, Fellinger T, Justl M,
10.1149/2.1161607jes. Trattner A, et al. Theoretical and experimental analysis of an
[25] Zenyuk IV, Parkinson DY, Connolly LG, Weber AZ. Gas- asymmetric high pressure PEM water electrolyser up to 155
diffusion-layer structural properties under compression via bar. Int J Hydrogen Energy 2017;42:30493e508. https://
X-ray tomography. J Power Sources 2016;328:364e76. https:// doi.org/10.1016/j.ijhydene.2017.10.112.
doi.org/10.1016/j.jpowsour.2016.08.020. [40] Arbabi F, Kalantarian A, Abouatallah R, Wang R, Wallace JS,
[26] Olesen AC, Frensch SH, Kær SK. Towards uniformly Bazylak A. Feasibility study of using microfluidic platforms
distributed heat, mass and charge: a flow field design study for visualizing bubble flows in electrolyzer gas diffusion
for high pressure and high current density operation of PEM layers. J Power Sources 2014;258:142e9. https://doi.org/
electrolysis cells. Electrochim Acta 2019;293:476e95. https:// 10.1016/j.jpowsour.2014.02.042.
doi.org/10.1016/j.electacta.2018.10.008. [41] Selamet OF, Pasaogullari U, Spernjak D, Hussey DS,
[27] Ito H, Maeda T, Nakano A, Kato A, Yoshida T. Influence of Jacobson DL, Mat MD. Two-phase flow in a proton exchange
pore structural properties of current collectors on the membrane electrolyzer visualized in situ by simultaneous
performance of proton exchange membrane electrolyzer. neutron radiography and optical imaging. Int J Hydrogen
Electrochim Acta 2013;100:242e8. https://doi.org/10.1016/ Energy 2013;38:5823e35. https://doi.org/10.1016/
j.electacta.2012.05.068. j.ijhydene.2013.02.087.
[28] Nie J, Chen Y, Cohen S, Carter BD, Boehm RF. Numerical and [42] Dedigama I, Angeli P, Ayers K, Robinson JB, Shearing PR,
experimental study of three-dimensional fluid flow in the Tsaoulidis D, et al. In situ diagnostic techniques for
bipolar plate of a PEM electrolysis cell. Int J Therm Sci characterisation of polymer electrolyte membrane water
2009;48:1914e22. https://doi.org/10.1016/ electrolysers - flow visualisation and electrochemical impedance
j.ijthermalsci.2009.02.017. spectroscopy. Int J Hydrogen Energy 2014;39:4468e82. https://
[29] Toghyani S, Afshari E, Baniasadi E, Atyabi SA. Thermal and doi.org/10.1016/j.ijhydene.2014.01.026.
electrochemical analysis of different flow field patterns in a [43] Ito K, Maeda Y, Sakaguchi T, Tsukamoto S, Inada A,
PEM electrolyzer. Electrochim Acta 2018;267:234e45. https:// Tsuchiya Y, et al. Analysis and visualization of water flow
doi.org/10.1016/j.electacta.2018.02.078. impact on hydrogen production efficiency in solid polymer
[30] Espinosa-Lopez M, Darras C, Poggi P, Glises R, Baucour P, water electrolyzer under high-pressure condition. Int J
Rakotondrainibe A, et al. Modelling and experimental Hydrogen Energy 2015;40:5995e6003. https://doi.org/10.1016/
validation of a 46 kW PEM high pressure water electrolyzer. j.ijhydene.2015.03.045.
Renew Energy 2018;119:160e73. https://doi.org/10.1016/ [44] Lafmejani SS, Müller M, Olesen AC, Kær SK. Experimental
j.renene.2017.11.081. and numerical study of flow in expanded metal plate for
[31] Barbir F. PEM electrolysis for production of hydrogen from water electrolysis applications. J Power Sources
renewable energy sources. Sol Energy 2005;78:661e9. https:// 2018;397:334e42. https://doi.org/10.1016/
doi.org/10.1016/j.solener.2004.09.003. j.jpowsour.2018.07.032.
[32] Xing X, Lin J, Song Y, Zhou Y, Mu S, Hu Q. Modeling and [45] Aubras F, Deseure J, Kadjo JJA, Dedigama I, Majasan J,
operation of the power-to-gas system for renewables Grondin-Perez B, et al. Two-dimensional model of low-
Integration. A Review 2018;4:168e78. https://doi.org/ pressure PEM electrolyser: two-phase flow regime,
10.17775/CSEEJPES.2018.00260. electrochemical modelling and experimental validation. Int J
Hydrogen Energy 2017;42:26203e16. https://doi.org/10.1016/ [56] Inc Ansys. ANSYS fluent. In: Help files. Knowl creat diffus
j.ijhydene.2017.08.211. util; 2018., Version 19.1.
[46] Oliveira LFL, Jallut C, Franco AA. A multiscale physical model [57] Ma Z, Blanchet S, Venkataraman R, Iaccarino G, Moin P.
of a polymer electrolyte membrane water electrolyzer. Mathematical modeling of an internal-reforming, carbonate
Electrochim Acta 2013;110:363e74. https://doi.org/10.1016/ fuel cell stack. Fuel Cell Sci. Eng. Technol 2004. - 2004.
j.electacta.2013.07.214. [58] Venkataraman R, Farooque M, Ma Z. Cost reduction through
[47] Han B, Mo J, Kang Z, Zhang FY. Effects of membrane thermal management improvements in large scale carbonate
electrode assembly properties on two-phase transport and fuel cells. ECS Trans 2007;5. https://doi.org/10.1149/1.2729037.
performance in proton exchange membrane electrolyzer [59] Ma Z, Venkataraman R, Farooque M. High power internal-
cells. Electrochim Acta 2016;188:317e26. https://doi.org/ reforming direct carbonate fuel cell stack development
10.1016/j.electacta.2015.11.139. through mathematical modeling and engineering
[48] Li H, Nakajima H, Inada A, Ito K. Effect of flow-field pattern optimization. J Fuel Cell Sci Technol 2010;7:051003-1e051003-
and flow configuration on the performance of a polymer- 8. https://doi.org/10.1115/1.4000625.
electrolyte-membrane water electrolyzer at high [60] Fuel cells e molten carbonate fuel cells | modeling. Juergen
temperature. Int J Hydrogen Energy 2018;43:8600e10. https:// garche, chris dye. Patrick moseley, zempachi ogumi. In:
doi.org/10.1016/j.ijhydene.2018.02.171. Ma Z, Venkataraman R, Farooque M, editors. Rand David,
[49] Dutta S, Shimpalee S, Van Zee JW. Three-dimensional numerical Scrosati Bruno, editors. Encycl. Electrochem. Power sources,
simulation of straight channel PEM fuel cells. J Appl Electrochem vol. 2. Amsterdam Elsevier; 2009. p. 519e32.
2000;30:135e46. https://doi.org/10.1023/A:1003964201327. [61] Manninen M, Taivassalo V. On the mixture model for
[50] Dutta S, Shimpalee S, Van Zee JW. Numerical prediction of multiphase flow. VTT Publ 1996:3e67.
mass-exchange between cathode and anode channels in a [62] Schiller L, Naumann ZZ. Ver deutsch ing. 1935. p. 318.
PEM fuel cell. Int J Heat Mass Tran 2001;44:2029e42. [63] Springer TE. Polymer electrolyte fuel cell model. J Electrochem
[51] Shimpalee S, Spuckler D, Van Zee JW. Prediction of transient Soc 1991;2334:138. https://doi.org/10.1149/1.2085971.
response for a 25-cm2 PEM fuel cell. J Power Sources [64] O'Hayre RP. Fuel cell fundamentals. 2006. https://doi.org/
2007;167:130e8. https://doi.org/10.1016/ 10.1017/CBO9781107415324.004. vol. 2nd ed.
j.jpowsour.2007.02.004. [65] Choi P. A simple model for solid polymer electrolyte (SPE)
[52] Shimpalee S, Van Zee JW. Numerical studies on rib & water electrolysis. Solid State Ionics 2004;175:535e9. https://
channel dimension of flow-field on PEMFC performance. Int J doi.org/10.1016/j.ssi.2004.01.076.
Hydrogen Energy 2007;32:842e56. https://doi.org/10.1016/ [66] Santarelli MG, Torchio MF, Cochis P. Parameters estimation
j.ijhydene.2006.11.032. of a PEM fuel cell polarization curve and analysis of their
[53] Shimpalee S, Lilavivat V, Van Zee JW, McCrabb H, Lozano- behavior with temperature. J Power Sources
Morales A. Understanding the effect of channel tolerances 2006;159:824e35. https://doi.org/10.1016/
on performance of PEMFCs. Int J Hydrogen Energy 2011. j.jpowsour.2005.11.099.
https://doi.org/10.1016/j.ijhydene.2011.06.146. [67] Tijani AS, Rahim AHA. Numerical modeling the effect of
[54] Shimpalee S, Greenway S, Van Zee JW. The impact of operating variables on faraday efficiency in PEM electrolyzer.
channel path length on PEMFC flow-field design. J Power Procedia Technol 2016;26:419e27. https://doi.org/10.1016/
Sources 2006;160:398e406. https://doi.org/10.1016/ j.protcy.2016.08.054.
j.jpowsour.2006.01.099. [68] Moore M, Wardlaw P, Dobson P, Boisvert JJ, Putz A, Spiteri RJ,
[55] Babic U, Suermann M, Felix NB, Gubler L, Schmidt TJ. Review d et al. Understanding the effect of kinetic and mass transport
identifying critical gaps for polymer electrolyte water. J processes in cathode agglomerates. J Electrochem Soc
Electrochem Soc 2017;164(4):F387e99. https://doi.org/10.1149/ 2014;161:E3125e37. https://doi.org/10.1149/2.010408jes.
2.1441704jes.

Hidrogenioverde

Uploaded by

Copyright:

Available Formats

Hidrogenioverde

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Hidrogenioverde

Uploaded by

Copyright:

Available Formats

Pursuant to the DOE Public Access Plan, this document represents the authors' peer-reviewed, accepted manuscript.

international journal of hydrogen energy xxx (xxxx) xxx

Available online at www.sciencedirect.com

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

A comprehensive modeling method for proton

Zhiwen Ma*, Liam Witteman, Jacob A. Wrubel, Guido Bender

highlights graphical abstract

Modeling scopes and methods are

article info abstract

overall water-splitting process is H2O(l) / H2(g) þ 1/2O2(g),

Fig. 1 e Schematic of a PEMEC with reactant ﬂow and electrochemical processes.

Table 1 shows the PEMEC modeling scope from basic

Ohmic losses, membrane water uptake, ionic resistance

Electrode characterization and optimization, activation

Cell heat mass balance and parameter distributions.

Gas species transport. Concentration overpotential

Size of the stack module and the BOP components

Optimization of stack thermal, electrochemical

Use test data to determine cell V-i relations.

performance and evaluate their behaviors under different

Component and system simulations

Stack component life prediction

dimension, Multiphysics, multi-scale, thermal-ﬂuid, trans-

empirical expressions for static or dynamic analysis. Fig. 2

3-D model of current density, temperature, and gas

Electrolyzer integrated into a hydrogen production

Gas/liquid water ﬂow, diffusion, permeation, and

maps out the various electrochemical cell and stack

modeling approaches to emphasize the numerous permuta-

Fig. 2 shows the modeling applications to complement the

PEMEC stack operation

development by modeling and testing [27,29e31,33e48].

Complementary modeling and testing can support the effec-

up performance. The modeling tool focuses on macroscale

modeling cell and stack designs and subsequently

Multi-dimension thermal-ﬂow and current density

Species transport through porous diffusion layers

Stack integration with the balance of plant (BOP)

System conﬁguration and operating conditions

CFD model of cells for thermal management

Module and System

Fig. 2 e Map of the electrochemical cell and stack modeling

of continuity, momentum, and energy equations for the Momentum equation

density, and rm is the mixture density, expressed as:

Table 2 e Governing equations used in the mixture model.

Phase Volume Fraction V,ðag rg !

effective thermal conductivity of the mixture, calculated

rg d2g Sh In Eq. (12) is a volumetric generation term that is deﬁned

Table 3 e Source terms implemented in conservation equations.

through the membrane is driven by diffusion, electro-osmotic

The diffusion contribution arises from gradients in the RT 1 i

hohm ¼ IRtot ¼ IðRe þ RI Þ (29)

where the total resistance includes electronic ðRe Þ and ionic

Re ¼ RAe þ RCe þ Rcontact (30)

The ionic resistance accounts for the voltage drop required

Higher membrane water content facilitates proton trans-

Source terms in the conservation equations from electrochemical

Table 4 e Nominal operating and simulation parameters.

Results and discussion