Location via proxy:   [ UP ]  
[Report a bug]   [Manage cookies]                

cac412f5d30628abe1b91469dc46fc85

Download as pdf or txt
Download as pdf or txt
You are on page 1of 14

This article appeared in a journal published by Elsevier.

The attached copy is furnished to the author for internal non-commercial research and education use, including for instruction at the authors institution and sharing with colleagues. Other uses, including reproduction and distribution, or selling or licensing copies, or posting to personal, institutional or third party websites are prohibited. In most cases authors are permitted to post their version of the article (e.g. in Word or Tex form) to their personal website or institutional repository. Authors requiring further information regarding Elseviers archiving and manuscript policies are encouraged to visit: http://www.elsevier.com/copyright

Author's personal copy

Renewable Energy 44 (2012) 128e140

Contents lists available at SciVerse ScienceDirect

Renewable Energy
journal homepage: www.elsevier.com/locate/renene

Using electrical analogy to describe mass and charge transport in PEM fuel cell
P. Noiying a, M. Hinaje a, P. Thounthong b, S. Ral a, *, B. Davat a
a b

Groupe de Recherche en Electrotechnique et Electronique de Nancy, 2 Avenue de la Fort de Haye, 54516 Vandoeuvre-ls-Nancy, France King Mongkuts University of Technology North Bangkok, 1518 Piboolsongkram Road, Bangsue, Bangkok, Thailand

a r t i c l e i n f o
Article history: Received 30 May 2011 Accepted 10 January 2012 Available online 9 February 2012 Keywords: PEM fuel cells Dynamic modeling Mass transport Electrical analogy Large signal equivalent circuit model

a b s t r a c t
This article deals with a PEMFC (proton exchange membrane fuel cell) electrical model based on a 1D analog representation of mass transport phenomena, and taking into account the inuence of gas supply conditions on fuel cell voltage. By using an electrical analogy to describe transport of gaseous species in GDLs (gas diffusion layers), and water distribution in the membrane, the model can be directly implemented in standard simulation softwares used in electrical engineering (such as Saber, in our case), so that it can be easily employed for the simulation of fuel cell electrical systems. The paper explains how the analog model is obtained from mass transport equations, then it presents validation tests carried out in both steady-state and transient regimes on a PEM single cell bench. 2012 Elsevier Ltd. All rights reserved.

1. Introduction The use of fuel cells in electrical circuits needs an understanding of internal effects for an efcient management of the system, as well as an understanding of dynamic behavior in order to design and control the entire energy production unit with fuel cells, power electronic converters, and storage units. It is especially crucial to describe accurately the strong inuence of operating conditions, such as temperature, gas ows, pressure and humidity, on fuel cell performances. For example, it is widely accepted that one of the key weak point of PEM fuel cell systems is their dynamic limitation. Indeed, the fuel cell system time constant is dominated by the compressor and the membrane hydration level, and may be several hundred of milliseconds. Therefore, fast load demands will cause signicant voltage drop, particularly due to air starvation. This phenomenon is illustrated in Fig. 1, which presents the voltage experimental response of a 500 W PEM fuel cell system to a current step, obtained by means of a current controlled boost converter [1]. In this test, fuel ows are adapted to the fuel cell current, and the effect of the ancillaries time constant (namely the fact that fuel supply and delivered electrical current do not coincide in transient), can be clearly seen on the voltage curve. Within the heart of a fuel cell three key processes that are strongly coupled take place [2]: mass transport in electrodes and electrolyte, electron conduction in electrodes and proton migration

* Corresponding author. Tel.: 33 383 59 56 54; fax: 33 383 59 56 53. E-mail address: Stephane.Rael@ensem.inpl-nancy.fr (S. Ral). 0960-1481/$ e see front matter 2012 Elsevier Ltd. All rights reserved. doi:10.1016/j.renene.2012.01.081

in electrolyte, and electrochemical kinetics at reactive sites. Many mathematical models have already been developed in order to describe locally these phenomena, through partial differential equations involving space and time. One of the earlier studies, by T. E. Springer et al. [3,4], led to an isothermal, one-dimensional, and steady-state model, relative to a PEMFC with a 117 Naon membrane. This model focused on water transport in the membrane through diffusion and electro-osmotic drag, and introduced variable hydration as well as hydration dependent protonic conductivity. In another signicant early work [5], D. M. Bernardi and M. W. Verbrugge developed a one-dimensional isothermal cell model to investigate factors that limit performance. Some other studies improved previous works, by including heat transfer description [6] or conservation equations [7] (i.e. transient modeling) in the general one-dimensional formulation. Many articles also focus on fuel cell two-dimensional mathematical modeling, the aim being often to investigate different aspects of water and heat management [8,9], of ow modes (coow or counterow) [10], or of gas distribution along channels [11]. According to results presented in the literature, mathematical fuel cell models are obviously accurate, and enable to study specic local effects. However, they are hardly used in a system approach, in which fuel cell models that are easy to handle are required. Moreover, they are most often only steady-state, which is clearly not sufcient to design a fuel cell based generation system comprising power electronic converters and storage units. For the aforementioned reasons, many electrical circuit models have been developed. Among them, we rst nd models that have been obtained by extending a static currentevoltage description to

Author's personal copy

P. Noiying et al. / Renewable Energy 44 (2012) 128e140

129

Fig. 1. Fuel cell voltage response to a current step when fuel ows are adapted to fuel cell current.

transient regimes, often through adding a capacitor associated with the electric double layer phenomenon. This is seen for example in the case of works published by C. Wang et al. [12], and by S. Pasricha et al. [13]. In Ref. [14], such a circuit is used to establish a state model of the fuel cell, and then to achieve a control algorithm of the boost converter connected with the fuel cell. Some authors propose original solutions to describe fuel cell static or transient behaviors. In Ref. [15], the static characteristic is obtained by means of electronic components (bipolar transistors, diodes, resistances), so that the model has been implemented in a simulation software dedicated to electrical and electronic circuits (PSPICE). S. Lazarou et al. [16] use an ideal transformer and two saturable inductors to describe fuel cell dynamic responses to load changes. These kinds of models can give accurate results, but they are often empirical, and with the exception of fuel cell current, they can hardly describe the inuence of operating conditions (temperature, gas ows, pressure, and humidity). Secondly, many dynamic models of fuel cells (and of electrochemical devices in general) are based on Randles equivalent scheme associated with harmonic responses, obtained by the usual EIS technique under various loading conditions. The aim is to t, usually by means of a numerical identication method (such as least square algorithm), a non-linear equivalent circuit with experimental impedance spectra. For example, A. M. Dhirde et al. have recently proposed such an equivalent circuit [17]. In this work, distributed elements, such as Warburg impedance and constantphase element, are used to describe diffusion phenomena which take place within the fuel cell. In Ref. [18], authors employ fractional differentiation to approximate the Nernst diffusion transfer function in Laplace domain. I. Sadli et al. have shown that this diffusion impedance could be correctly represented by two RC cells [19]. Modeling fuel cells through AC impedance studies is most often physics-based, but fundamentally, theses studies generates only small signal models, which are theoretically not available in large signal conditions, because of non-linearities of electrochemical devices. In addition, include all operating conditions leads to a quite long and difcult parameter identication procedure.

Lastly, some works, such as [20e22], present fuel cell models that associate a standard description through electrical circuits, and macroscopic conservation equations of mass and energy. This enables prediction of transient evolution of gas ows and pressures, and to evaluate their inuence on electrical behavior. In this article, we focus on a PEM fuel cell electrical model implemented in simulation software commonly used in electrical engineering to design systems. This model is isothermal and includes transient description, double layer phenomenon, and 1D mathematical representation of mass transport phenomena in GDLs (gas diffusion layers) and membrane. For this purpose, we use an electrical analogy of gas and water diffusion equations, so that the model can be directly implemented in simulation software dedicated to electrical circuit. It can be therefore easily employed for the simulation of fuel cell electrical systems. It will be shown moreover how gas supply conditions and associated phenomena (inuence of over-stoichiometric conditions on fuel cell voltage, transient air starvation) can be described. In the rst part, mass and charge transport equations, electrochemical kinetic equations, and boundary conditions of our fuel cell model are recalled. Following this, electrical analogy of mass transport equations in electrodes and membrane, and of associated boundary conditions, is detailed, as well as the model assembly. Finally, an experimental validation carried out on a single cell system is presented. 2. One-dimensional PEM fuel cell mathematical model Fig. 2 shows a schematic representation of the studied single PEM fuel cell. It is composed of a membrane sandwiched between two catalyst layers and two gas diffusion layers. Hydrogen (humidied or dry) is fed to the anodic compartment, and humidied air to the cathodic one, by means of gas distribution channels. In this section, we detail equations that describe the distribution of gaseous species in GDLs, the water content in the membrane, electrochemical overvoltages at membraneeelectrode interfaces, and ohmic voltage drop across the membrane. The aim

Author's personal copy

130

P. Noiying et al. / Renewable Energy 44 (2012) 128e140

2.1.1. Gas transport If both Knudsen and StefaneMaxwell diffusions are taken into account, 1D gas mixture transport in GDLs is governed for each gas i of the mixture by the following law [23,24]:

0 1 N X Pj $Ji Pi $Jj vPi RT @ Ji A; $ vx Acell Di;eff j 1;jsi P$Dij;eff

(1)

where N is the number of species (Na 2 and Nc 3), Ji, Pi and Di,eff are the molar ow, the partial pressure and the effective Knudsen diffusion coefcient of the gas i, Dij,eff is the effective StefaneMaxwell diffusion coefcient of gases i and j, and P is the total pressure. The subscript i is H2 or H2O at the anode, O2 or N2 or H2O at the cathode. If is the GDL porosity, the balance equation of gas i can be written as:

vPi RT vJi $ : $Acell vx vt


Fig. 2. Internal structure of a single fuel cell.

(2)

is to calculate the fuel cell voltage in transient conditions. Assumptions and simplications adopted in the present model are as follows: 1) the cell temperature remains constant and homogeneous throughout the cell 2) the ideal gas mixture is in a single phase 3) mass and charge transports are one-dimensional 4) the membrane is gas-tight, and homogeneous 5) all electrical contact resistances are neglected 6) catalyst layers are assumed like interfaces As a result, only three computational regions will be considered: the anodic GDL, the membrane, and the cathodic GDL. Phenomena and associated laws that are taken into account in our onedimensional transient fuel cell model are listed hereafter: 1) diffusion of species in GDLs (Knudsen diffusion law), and binary diffusion of gas mixtures in GDLs (StefaneMaxwell diffusion law) 2) diffusion (due to concentration gradient) and electro-osmotic dragging (due to the membrane ionic current) of liquid water in the membrane 3) conductive transport (Ohm law) of electrons in electrodes, and of ions in the membrane As mentioned before, electrochemical reactions that take place in catalyst layers are computed as boundary conditions at membrane/electrode interfaces. 2.1. Gas diffusion layers The aim of mass transport calculation in GDLs is to determine partial pressures of hydrogen and oxygen at membraneeelectrode interfaces, in order to evaluate electrode overvoltages. Secondly, partial pressures of water at membraneeelectrode interfaces are needed to calculate the membrane water content at these boundaries. Thus, anodic and cathodic gas diffusion layers content the following unknowns: partial pressures of gases (PH2 and PH2 Oa at the anode, PO2 , PN2 and PH2 Oc at the cathode), molar uxes of gases (JH2 and JH2 Oa at the anode, JO2 , JN2 and JH2 Oc at the cathode), and electrode potentials (denoted fs).

2.1.2. Charge transport Charge transport in GDLs is governed by Ohms law. Thus, the GDL electronic current Is and the electrical potential fs are linked by:

Is vf ss;eff $ s ; Acell vx

(3)

where ss,eff is the effective electronic conductivity of electrodes. In a 1D approach, the electronic current Is only depends on time, and is necessarily equal to the fuel cell current Icell. As a result, voltage drop across GDL is simply described by a constant resistance. 2.1.3. Boundary conditions for gas transport Boundary conditions at membraneeelectrode interfaces (x da and x da dm) are linked to hydrogen and oxygen consumption, to water production, and to water molar ow continuity. This leads at the membraneeanode interface to:

8 <

Ia JH2 da ; 2F :J H2 Oa da JH2 Om da

(4)

JH2 Om being the membrane water molar ow, and Ia the anodic Faraday current. It should be noticed here that this current is different from Icell in transient, because of the capacitive component due to the electric double layer that exists at membraneeanode interface. At the membraneecathode interface, if Ic is the cathodic Faraday current, boundary conditions are:

8 > J d d Ic > O2 a m > < 4F JN2 da dm 0 : > > Ic > :J H2 Oc da dm JH2 Om da dm 2F

(5)

In an actual experimental setup, both gas partial pressures (gas mixture composition) and gas molar ows are imposed at GDL inlets. For example, these boundary conditions can be expressed versus gas relative humidities (HRa, HRc), humidier water temperatures (Thum,a, Thum,c), gas mixture pressures (Pa,in, Pc,in), and gas supply conditions (reference current Iref, reactant stoichiometries za and zc). At the anode side, the saturation pressure being denoted Psat, this leads to:

Author's personal copy

P. Noiying et al. / Renewable Energy 44 (2012) 128e140

131

&

PH2 Oa ;in HRa $Psat Thum;a ; PH2 ;in Pa;in PH2 Oa ;in

(6)

expression of M. Whr [26], with Thum in Kelvin, Psat in Pa unit, A 8.073, B 1656.39 K, and C 226.86 K:

and:

8 > > JH <

Psat Thum 100$10 za $ Iref 2F : PH2 Oa ;in $J PH2 ;in H2 ;in

ACT

B hum 273

(14)

2 ;in

>J > H2 Oa ;in :

(7)

For binary diffusion coefcients, we will use the empirical correlation given by M. F. Serican and S. Yesilyurt [27]:

At the cathode side (x da dm dc), boundary conditions are:

 1:5 P T Dij T; P D0 T0 ; P0 $ 0 $ ; ij P T0

(15)

8 $P < PH2 Oc ;in HRc sat Thum;c 0:21$ Pc;in PH2 Oc ;in ; P : O2 ;in PN2 ;in 0:79$ Pc;in PH2 Oc ;in
and:

(8)

where T and P are the local fuel cell temperature and pressure. Table 1 hereafter details values of D0 T0 ; P0 [27]. To account for ij GDL porosity s, the effective diffusion coefcient is calculated with Bruggemans relation [28]:

8 I >J > O2 ;in zc $ ref > > 4F > > < PH2 Oc ;in JH2 Oc ;in $J PO2 ;in O2 ;in : > > > > PN2 ;in > >J $J : N2 ;in PO2 ;in O2 ;in

Dij;eff T; P Dij T; P$1:5 : s

(16)

(9)

Last, it should be noticed that binary diffusion coefcients of gas i and gas j are linked by a reciprocal relation, called in this case Onsagers reciprocal relation [29]:

Dij;eff T; P Dji;eff T; P:
2.2. Membrane

(17)

Nevertheless, as there is no GDL outlet in 1D modeling, gas molar ows and gas partial pressures cannot be imposed simultaneously at GDL inlets. Most often, inlet partial pressures are imposed as boundary conditions. As a result, gases are necessarily supplied under strict stoichiometric conditions [7,25], what does not correspond to actual operating conditions. Indeed, reactants are usually supplied under over-stoichiometry conditions, what has a signicant inuence on fuel cell voltage. On the contrary, as shown in Fig. 1, air starvation phenomena may occur, especially during fast transient power demands [1]. In our 1D model, whether gas partial pressures, or gas molar ows, are used as boundary conditions at GDL inlets, depending on gas supply conditions. Thus, in the case of an over-stoichiometry gas supply, partial pressures are imposed at GDL inlets. Then, using Eqs. (6) and (8), boundary conditions at the anodic inlet (x 0) are as follows:

&

PH2 Oa 0 PH2 Oa ;in ; PH2 0 PH2 ;in

(10)

and at the cathodic inlet (x da dm dc):

8 < PH2 Oc da dm dc PH2 Oc ;in P da dm dc PO2 ;in : : O2 PN2 da dm dc PN2 ;in

(11)

In the case of an under-stoichiometry gas supply, gas molar ows are imposed at GDL inlets, using Eqs. (7) and (9) as follows:

Fuel cell membrane behaves like an acid solution. It contains xed sulfonic sites SO that dissociate water molecules in protons 3 H. As a result, both water and proton transports in the membrane strongly depends on the membrane hydration state. The aim of water transport calculation in the fuel cell membrane is to determine the membrane water concentration, in order to evaluate the voltage drop across the membrane. Secondly, water concentration and water molar ow at membraneeelectrode interfaces are needed to calculate partial pressures of water at these boundaries. Thus, the fuel cell membrane, in which water and ions are transported, has the following unknowns: water concentration, water molar ux and membrane potential, denoted cH2 O , JH2 Om and fm, respectively. However, the membrane hydration state is often represented, in particular in parametric laws, by the membrane water content l. This quantity is dened as the ratio between the number of water molecules and the number of sulfonic sites SO available in the 3 membrane polymer. In practice, it approximately varies between 2 and 22. Water content can also be calculated versus membrane proton exchange capacity Xm (number of available sulfonic sites in the membrane per mass unit), membrane density rm and water concentration as follows:

&

JH2 0 JH2 ;in ; JH2 Oa 0 JH2 Oa ;in

c H2 O : Xm $rm

(18)

(12)
2.2.1. Water transport The water transport in the membrane is a combination of two competing mechanisms [9,30]. One is due to the proton displacement from the anode to the cathode. As protons are solvated, they drag some water molecules with them. This phenomenon is called
Table 1 Binary diffusivities at P0 1 atm [27]. Diffusivity name Ref. temperature T0 [K] 307.1 308.1 293.2 307.5 Diffusivity value [m2 s1] 9:15,105 2:82,105 2:20,105 2:56,105

and:

8 < JO2 da dm dc JO2 ;in J da dm dc JH2 Oc ;in : : H2 Oc JN2 da dm dc JN2 ;in

(13)

It will be explained in Section 3 how to compute such boundary conditions, when using electrical analogy. 2.1.4. Parametric laws The saturation pressure of water vapor, which is needed to calculate water molar fractions at GDL inlets, depends on humidier temperature. In our model, it is computed using the empirical

D0 2 ;H2 O H D0 2 ;H2 O O D0 2 ;N2 O D0 2 ;H2 O N

Author's personal copy

132

P. Noiying et al. / Renewable Energy 44 (2012) 128e140

electro-osmotic drag. The other mechanism is water diffusion that generally occurs from the cathode to the anode. This water ux results from the water concentration gradient created in the membrane by the electro-osmotic drag and the water produced by the redox reaction at the cathode. Therefore, the total molar ow of water transported inside the membrane is given by:

2.2.4. Parametric laws We will make use of parametric laws established by some authors for Gore-Select membranes. Indeed, our fuel cell test bench is equipped with this kind of membrane. Early studies of T. E. Springer et al. [3] have proposed the following linear empirical law for the electro-osmotic drag coefcient versus water content:

JH2 Om

vcH2 O nd $Im Acell $DH2 Om $ ; F vx

(19)

nd

2:5 $l: 22

(25)

where Im is the ionic current owing through the membrane, nd is the electro-osmotic drag coefcient (number of water molecules dragged per proton), and DH2 Om is the water diffusion coefcient. The conservation of water quantity in the membrane can be written as:

More recently, this coefcient has been measured for a Gore membrane to the following constant value, over a wide range of water activity (from 40% to 95%) [31]:

nd 1:07:

(26)

vcH2 O 1 vJH2 Om $ : Acell vx vt

(20)

In the same article, diffusion properties of Gore membrane are studied, and authors propose for water diffusion coefcient versus membrane water content and temperature the following empirical relation [31]:

2.2.2. Charge transport The membrane is assumed to be homogeneous. Therefore, as sulfonic sites are xed, there is no proton concentration gradient in the membrane body, where local electroneutrality is veried. As a consequence, proton transport in the membrane is only governed by Ohms law. Thus, the ionic current Im and the electrical potential fm are linked by:

8 > DH O fD l$gD T$1010 > 2 m > > > > 2 3 > < 2:563 0:33$l 0:0264$l 0:000671$l fD l : 2 > >    > > > > g T exp 2416$ 1 1 > D : 303 T

(27)

Im vf sm $ m ; Acell vx

(21)

To evaluate the membrane ionic conductivity, we will use the following empirical law, established by Neubrand [32]:

where sm is the ionic conductivity of the membrane that greatly depends on membrane hydration. In a 1D approach, the ionic current Im only depends on time, and is necessarily equal to the fuel cell current Icell. As a result, voltage drop across the membrane is described by a non-linear resistance:

8 > sm fs l$gs T > < 2 3 fs l 0:2658$l 0:0298$l 0:0013$l  :   > > gs T exp 2640exp0:6l 1183 1 1 : 303 T

(28)

hm

Icell $ Acell

dZ dm a da

dx

sm

Last, the sorption phenomenon at membraneeelectrode interfaces will be described by empirical sorption curves of Hinatsu et al. [33]. At 30  C and 80  C, these curves are dened as follows:

(22)

&

lsorp 30  C 0:043 17:81$a 39:85$a2 36$a3 : lsorp 80  C 0:3 10:8a 16a2 14:1a3

(29)

2.2.3. Boundary conditions for water transport At membraneeelectrode interfaces (x da and x da dm), the continuity of water molar ow is considered. Thus, according to Eqs. (4) and (5), we obtain:

To estimate interface water contents at an intermediate temperature, a linear interpolation between these two experimental formulas is made.

8 < JH2 Om da JH2 Oa da : JH2 Om da dm JH2 Oc da dm

Ic ; 2F

(23)

2.3. Voltage, overvoltages and currents Fuel cell voltage is given versus electrode equilibrium potentials Ea and Ec, electrode overvoltages ha and hc, membrane voltage drop hm, and GDL electrical resistances ra and rc by:

Ic being the cathodic Faraday current (different from the fuel cell current Icell in transient, because of the capacitive component due to the electric double layer at membraneecathode interface). Moreover, a phase change takes place for water (vapor phase in GDLs, liquid phase in the membrane) at these interfaces. This phenomenon is called sorption. Assuming thermodynamic equilibrium between water vapor in the backing layers and liquid water in polymer, sorption curves enable the calculation of water contents versus water activity a PH2 O =Psat at membraneeelectrode interfaces, as follows:

Vcell Ec hc Ea ha hm ra rc $Icell :

(30)

Electrode overvoltages are evaluated using simplied expressions based on ButlereVolmer equation, in order to facilitate algorithm convergence. The anodic overvoltage expression supposed that both oxidation and reduction transfer coefcients are equal to 0.5, so that it can be written:

  8 PH2 Oa da > ld l > < a sorp Psat Tcell :  PH2 Oc da dm > > ld d l : a m sorp Psat Tcell

ha
(24)

  PH2 ;in RT Ia ; $ $a sinh F PH2 da 2$Acell $j0a

(31)

where j0a is the anodic exchange current. The cathodic overvoltage expression supposed that the oxidation component of Ic is negligible, compared to reduction component of Ic. However, the logarithm

Author's personal copy

P. Noiying et al. / Renewable Energy 44 (2012) 128e140

133

argument theoretically obtained is slightly modied, so that hc is dened if Ic 0. Finally, it gives:

hc

  PO2 ;in RT 1 Ic ; $ln 1 $ $ F 2ac PO2 da dm Acell $j0c

(32)

where j0c is the cathodic exchange current, and ac is the reduction transfer coefcient. At membraneeelectrode interfaces, where redox reactions take place, fuel cell current Icell is split into a Faraday component verifying ButlereVolmer equation, and a capacitive component due to the electric double layers that exist at these interfaces. This results in:

8 dEa ha >I I < a cell Cdl;a $ dt : dEc hc > : Ic I Cdl;c $ cell dt


3. PEMFC modeling using electrical analogy 3.1. Gas diffusion layers

(33)

Space discretization of Eqs. (1) and (2) leads to the following system, for the description of gas mixture diffusion in GDLs:

8 > > P x Dx P x < i i

Ri

N P j 1;jsi

! Rij x $Ji x

N P j 1;jsi

Vij x

Fig. 3. Equivalent coupled circuits of gas mixture diffusion in an anodic GDL element.

> > : J x Dx J x C$dPi i i dt


with:

(34) 8 > > R RT$Dx > i > > Acell $Di;eff > > > > > > > > R x RT$Dx $Pj x > ij < Acell $Dij;eff Px ; > > > > Vij x RT$Dx $Pi x$Jj x > > > Acell $Dij;eff Px > > > > > > > C s $Acell $Dx : RT

(35)

Dx being the space discretization step. This non-linear system can be computed using an electrical analogy, for which pressures and molar ux are assimilated to voltages and currents, respectively. For illustration, Fig. 3 presents the equivalent electrical scheme of gas mixture diffusion in an element of fuel cell anode GDL. This equivalent scheme is composed of two electrical circuits (three in the case of the cathodic GDL), coupled to one another by partial pressures (resistances Rij in Eq. (35)) and molar ows (voltage sources Vij in Eq. (35)). To compute these couplings, as depicted in Fig. 3, each circuit of the equivalent electrical scheme (for example, circuit describing hydrogen transport in an element of anode GDL) receives from the other circuits their input partial pressures and molar ows (in our example: PH2 O x and JH2 O x, respectively), what enables to calculate coupled components of the considered circuit (in our example: RH2 ;H2 O and VH2 ;H2 O ), according to Eq. (35). In return, it feeds the other circuits with its own input partial pressure and molar ow (in our example: PH2 x and JH2 x), what enables to calculate coupled components of other circuits (in our example: RH2 O;H2 and VH2 O;H2 ), according to Eq. (35). Gas mixture diffusion in fuel cell GDLs is then described by associating in series such equivalent coupled circuits. In the analogical model presented here, a space discretization in 10 elements of both GDLs is used (Dxa da/10 and Dxc dc/10).

Figs. 4 and 5 depict inputs, outputs and boundary conditions of the resulting anodic and cathodic models. Model inputs, i.e. current reference Iref, Faraday current Ia or Ic, and membrane water molar ow JH2 Om da or JH2 Om da dm at membraneeelectrode interfaces, are required for setting boundary molar ow conditions. Model outputs enable fuel cell voltage calculation (overvoltage ha or hc) or water sorption computation (water partial pressure PH2 Oa da or PH2 Oc da dm at membraneeelectrode interfaces). Boundary conditions (Eqs. (4) and (5) for molar ows at membraneeelectrode interfaces, Eqs. (6)e(9) for molar ows or partial pressures at GDL inlets) are described by means of current sources and voltage sources. It should also be noticed here that for more clarity, couplings between circuits within gas transport models are not represented. At GDL inlets, the use of diode functions enable to distinguish the over-stoichiometry gas supply conditions (for which partial pressures are imposed), from the under-stoichiometry gas supply conditions (for which molar ows are imposed). By this way, air or hydrogen starvation phenomena, as well as the inuence of gas ows on fuel cell performances, are included in the model, despite this one is 1D. 3.2. Membrane Space discretization of Eqs. (19) and (20) leads to the following system, for the description of water transport in the membrane:

8 < lx Dx lx Rw $JH2 Om x Vw ; dl : JH2 Om x Dx JH2 Om x Cm $ dt


with:

(36)

8 > Dx > > > Rw > > Acell $DH2 Om $Xm $rm < ; n > Vw d $Rw $Im > > F > > > : C A $X $r $Dx m m m cell

(37)

Author's personal copy

134

P. Noiying et al. / Renewable Energy 44 (2012) 128e140

Fig. 4. Electrical analogy of gas mixture diffusion in fuel cell anode.

where Dx is the space discretization step. Once again, this system can be computed using an electrical analogy, for which water content and molar ux are assimilated to voltages and currents, respectively. Fig. 6 presents the two equivalent electrical schemes of water transport in membrane that are implemented in our model. The rst (Fig. 6a) is used for the left part of the membrane (anode side), and the second (Fig. 6b) for the right part of the membrane (cathode side). Water transport in the membrane is then described by associating in series these equivalent circuits. In the analogical model presented here, a space discretization in 10 elements of the membrane (5 elements for the left part of the membrane, and as many for the right part) is used (Dx dm/10). Fig. 7 depicts inputs, outputs and boundary conditions of the resulting membrane model. Anodic and cathodic water partial pressures at membraneeelectrode interfaces, i.e. PH2 Oa da and PH2 Oc da dm are required for setting boundary water contents. The ionic current Im owing through the membrane is needed to calculate the electro-osmotic term in water transport (voltage term Vw in Eq. (37)), and to evaluate the voltage drop hm across the membrane (Eq. (22)). The model outputs JH2 Om da and JH2 Om da dm , which are water molar ows at membraneeelectrode interfaces, are used as inputs of anodic and cathodic models (cf. Figs. 4 and 5) to set water ow conditions at these boundaries. The two other model outputs, i.e. the membrane resistance Rm and the voltage drop hm, simply enable the calculation of membrane electrical performances and fuel cell voltage. There are computed as follows:

3.3. Fuel cell model Fig. 8 presents the complete fuel cell model that has been implemented in Saber software. It associates: 1) analogical transport models previously described, which calculate gas ows and partial pressures in GDLs, water content and water ow in the membrane, electrochemical overvoltages at membraneeelectrode interfaces (Eqs. (31) and (32) and voltage drop across the membrane (Eq. (38)), 2) a standard electrical scheme that compute the fuel cell output voltage as a function of electrode equilibrium potentials, electrode overvoltages, membrane voltage drop and GDL electrical resistances (Eq. (30)), and the fuel cell output current as a function of Faraday currents and double layer capacitors (Eq. (33)). As depicted in Fig. 8, these two parts of the model are coupled, from transport models toward electrical model through electrochemical overvoltages and membrane voltage drop, and from electrical model toward transport models through electrochemical currents. 4. Simulations and experimental validation 4.1. Fuel cell experimental setup Experiments were carried out on a single proton exchange membrane fuel cell, based on a Gore Primea MEA. Membrane thickness is about 15 mm, and GDL thickness is equal to 400 mm. The cell active area is 100 cm2. It is supplied with pure dry hydrogen from cylinders. Two pressure regulators decrease hydrogen pressure from 200 bars to 3 bars, and a third pressure regulator enables

8 > < > :

Rm

Dx

hm

2$Acell Rm $Im

Nm X

k1

 1 1 sm xk sm xk Dx ;

(38)

Nm being the number of membrane elements. Last, membrane boundary conditions on water contents (Eq. (24)) are described by means of voltage sources.

Author's personal copy

P. Noiying et al. / Renewable Energy 44 (2012) 128e140

135

Fig. 5. Electrical analogy of gas mixture diffusion in fuel cell cathode.

to adjust hydrogen pressure at anode outlet. At anode inlet, hydrogen ow is set by a ux regulator. The fuel cell cathode is supplied with air through a compressor and a pressure regulator that decreases air pressure down to 3 bars. As for anode, another pressure regulator enables to adjust air pressure at cathode outlet, and air ow is set at cathode inlet by

means of a ux regulator. Moreover, air humidity is controlled by bubbling the gas through a water bath, which is maintained at a set-point temperature. Inlet air temperature is measured, in order to calculate its water molar fraction (according to Eq. (14), inlet air being supposed water saturated), and then its relative humidity (at stack temperature).

Fig. 6. Equivalent circuits of water transport in a membrane element.

Author's personal copy

136

P. Noiying et al. / Renewable Energy 44 (2012) 128e140

Fig. 7. Electrical analogy of water transport in fuel cell membrane.

Fig. 8. Electrical model of a PEM fuel cell.

Author's personal copy

P. Noiying et al. / Renewable Energy 44 (2012) 128e140

137

The fuel cell rated current is 50 A. A cooling system keeps the stack temperature constant. An electronic load enables to impose a constant current to the fuel cell, up to 100 A. A second one, made in our laboratory by means of power bipolar transistors working in linear regime (rated current: 40 A, cut-off frequency: 10 kHz), enables to impose fast varying currents (such as current steps), and to perform impedance spectroscopy. Last, a "real time" electronic card (dSPACE DS1104 card) is used in combination with MatlabSimulink mathematical environment for all necessary control functions such as reference setting (inlet gas ows, outlet pressures, and load current). Inlet gas ows are calculated versus stoichiometric coefcients za and zc, and reference current Iref that can be set by an external signal (in particular when using an adapted ow supply), as follows:

Table 3 Model parameters. Parameter GDL thicknesses Membrane thickness Cell active area GDL porosity S.-M. diffusion coefcients Electrode conductivity GDL resistances Dry membrane density Membrane proton exchange capacity Electro-osmotic drag coefcient Water diffusion coefcient Membrane ionic conductivity Anode exchange current density Cathode exchange current density Cathode reduction transfer coefcient Double layer capacitors Water saturation pressure Water sorption curves Symbol Value or equation 400 mm 15 mm 100 cm2 0.6 (16)e(18) 1000 S m1 0 2050 kg m3 0.91 mol kg1 (27) (28) (29) 5000 A m2 0.5 A m2 0.5 2F (15) (30) Reference e e e [34] [27e29] [35] Neglected [36] [37] [31] [31] [32] [38] Estimated [38] Measured [26] [33]

da, dc dm
Acell s Dij,eff

ss
ra, rc

rm
Xm nd DH2 Om

sm ac
j0a j0c

8 > > dH Iref $RT0 $za $1000$60 < 2 2F P0 > > d Iref $RT0 $ 1 $zc $1000$60 : air 4F P0 0:21

L:min1 ;

(39)

Cdl,a, Cdl,c Psat

lsorp

with T0 273 K and P0 1:013,105 ,Pa. Other operating conditions are detailed in Table 2. 4.2. Model parameters Geometric and physical parameters required for model simulation are recalled or given in Table 3. It should be noticed here that Knudsen diffusion phenomenon will not be taken into account. GDL ohmic resistances ra and rc (that can easily be calculated to 40 mU) have also been neglected. Anodic and cathodic double layer capacitors are assumed to be equal to twice the value obtained, by means of usual EIS technique, for the global double layer capacitor of the fuel cell. 4.3. Simulations and experimental validation in steady-state Steady-state characteristics plotted hereafter have been obtained with constant gas ows, set by Iref 40 A. The rst test (cf. Fig. 9a) is associated with a stoichiometry 1-2 (za 1 at the anode side, and zc 2 at the cathode side). It can be observed that there is quite good agreement between simulation results and experimental results. Thus, the model correctly describes the actual static behavior of the fuel cell when air is supplied with a double stoichiometry. However, there is an increasing difference between experiments and simulation, when the fuel cell current is near the limiting current imposed by hydrogen supply (i.e. 40 A in this case). At zc 1, model and experiment give quite different results, as shown in Fig. 9b. Indeed, if simulation obviously ts experiments for currents less than 20 A, it can be observed for higher currents that the measured fuel cell voltage drops more sharply than the simulated one. It seems that the fuel cell suffers from air starvation. This phenomenon should normally occur for a fuel cell current approximately equal to Iref. This is clearly not the case, and further investigation on the experimental setup has demonstrated that gas

ow imposed at cathode inlet is signicantly different than gas ow actually entering cathodic gas diffusion layer. Indeed, for the cell technology used here, gas ows in channels, parallel to the GDL from the inlet to the outlet. As a result, a part of inlet gas (approximately 37% at the cathode) ows in channels to the outlet, without entering the GDL. This part of inlet gas cannot react, it is simply wasted. A small but actual discrepancy (approximately 5%) is also observed at the anode, which results in the increasing difference between experiment and simulation near the anodic limiting current (cf. Fig. 9a). Fig. 10 presents membrane resistance obtained in steady-state with constant gas ows, set by Iref 40 A, za 2 and zc 4. The

Table 2 Operating conditions. Parameter Cell temperature Humidier temperature Air inlet temperature Anode relative humidity Cathode relative humidity Anode outlet pressure Cathode outlet pressure Symbol T Thum,c Tair,in HRa HRc Pa,in Pc,in Value 333 333 320 0% 53.2% 1 1 Unit K K K e e atm atm

Fig. 9. Fuel cell static characteristics. (a): Iref 40 A, za 1 and zc 2. (b): Iref 40 A, za 1 and zc 1.

Author's personal copy

138

P. Noiying et al. / Renewable Energy 44 (2012) 128e140

4.4. Simulations and experimental validation in transient Fig. 12a presents experimental and simulated voltage responses to a current step at constant gas ows, set by Iref 40 A, za 1 and zc 2. The initial current is 0 A, and the step amplitude is 30 A. As stated previously, it can be observed that steady-state is correctly described by the model. However, one can also notice a signicant difference between experiment and simulation in short times (the rst 500 ms following the current step). Indeed, simulated voltage quickly drops to 0.607 V (obtained at t 52 ms) before increasing to the new steady-state (i.e. 0.662 V), whereas measured voltage slowly decreases to 0.657 V (obtained at t 510 ms) before increasing to nearly the same steady-state. As can be seen in Fig. 12b, which presents the simulated response of the membrane resistance to the considered current step, this difference is mainly due to the description of water diffusion in the membrane, the dynamic of which is clearly too high. The so-called fuel or air starvation phenomenon occurs during fast power demands (such as current steps), when gas ows are set by current level. It is mainly due to the slow time constant of the air ux regulator. This time constant is about 1 s in the case of our test bench. Fig. 13 shows experimental and simulated voltage responses to a current step, from 10 A to 30 A. Gas ows are set by the actual fuel cell current Icell (t), stoichiometries being za 2 and zc 4 (as recommended by the cell manufacturer). When comparing step voltage responses given in Fig. 12 (obtained with constant gas ows) to Fig. 13, it can be concluded that our model includes the description of air starvation phenomenon. Nevertheless, as shown previously by the step response at constant gas ows, it is obvious that the accuracy has to be improved, notably by an actual parameter identication procedure that is still under study.

Fig. 10. Fuel cell membrane resistance versus current. Iref 40 A, za 2, za 4.

measurement has been achieved by means of current step technique (measure of the immediate voltage drop caused by a current step added to a given steady-state DC regime). The experimental curve is compared with two simulated curves: one is calculated with gas ow conditions imposed at electrode inlets (Iref 40 A, za 2, zc 4), the other with corrected gas ow conditions (Iref 40 A, za 2 0.95, zc 4 0.63). Except for low current levels, there is quite good agreement between simulation and experiment, as far as air waste (due to direct air ow from cathode inlet to cathode outlet) is taken into account. Last, we point out that, as for an actual experimental setup, the model describes fuel cell voltage dependence versus air ow conditions. Fig. 11 presents static characteristics simulated with zc 1 and zc 4. It can be observed that the voltage obtained with zc 4 is greater than the one corresponding to zc 1. This is rst due to a higher oxygen partial pressure at membraneecathode interface, implied by a greater air stoichiometry. The effect on output voltage increases with fuel cell current. In the example treated in Fig. 11, it represents 60% of the voltage gain at Icell 20 A, and 92% at Icell 35 A. Second, increasing air stoichiometry also implies a higher water partial pressure at membraneecathode interface, and therefore a better membrane hydration in open circuit condition. However, this increase of membrane water content is partially or totally balanced (depending on the relative humidity of inlet air) by electro-osmotic drag phenomenon. As a result, its effect on output voltage decreases with fuel cell current. In the example treated in Fig. 11, it represents 40% of the voltage gain at Icell 20 A, and only 8% at Icell 35 A. These simulation results have still to be validated by experiment.

Fig. 11. Fuel cell static characteristics simulated with various air ow conditions. Iref 40 A, za 1.

Fig. 12. Voltage and membrane resistance responses to a 0 A/30 A current step. Gas ow setting: Iref 40 A, za 1 and zc 2.

Author's personal copy

P. Noiying et al. / Renewable Energy 44 (2012) 128e140

139

Fig. 13. Experimental and simulated voltage responses to a 10 A/30 A current step. Gas ow setting: Iref Icell(t), za 2 and zc 4.

5. Conclusion In this article, we detail a PEM fuel cell electrical model that includes transient description, double layer phenomenon, and 1D mathematical representation of mass transport phenomena in GDLs (gas diffusion layers) and membrane. For this purpose, electrical analogies of gas and water diffusion equations are used, so that the model is directly implemented in simulation software dedicated to electrical circuits and electrical systems design. Simulations have shown that gas supply conditions are taken into account although the model is one-dimensional. As a result, electrical behaviors associated with air stoichiometry, or with transient air starvation, are included in the model. However, validation tests indicate that the model parameters have to be correctly dened, in order to increase accuracy in transient. We are presently working on dening specic tests for parameter identication. And even in steady-state, it seems that we have to investigate fuel cell behavior near limiting current conditions, in particular to predict more accurately cell voltage under transient air starvation condition. References
[1] Thounthong P, Ral S, Davat B. Control strategy of fuel cell and supercapacitors association for a distributed generation system. IEEE Transactions on Industrial Electronics December 2007;54(6):3225e33. [2] Baschuk JJ, Li X. A general formulation for a mathematical PEM fuel cell model. Journal of Power Sources March 2005;142(1e2):134e53. [3] Springer TE, Zawodzinski TA, Gottesfeld S. Polymer electrolyte fuel cell model. Journal of the Electrochemical Society August 1991;138(8):2334e42. [4] Springer TE, Wilson MS, Gottesfeld S. Modeling and experimental diagnostics in polymer electrolyte fuel cells. Journal of the Electrochemical Society December 1993;140(12):3513e26. [5] Bernardi DM, Verbrugge MW. A mathematical model of the solid-polymerelectrolyte fuel cell. Journal of the Electrochemical Society September 1992; 139(9):2477e91. [6] Rowe A, Li X. Mathematical modeling of proton exchange membrane fuel cells. Journal of Power Sources December 2001;102(1e2):82e96. [7] Friede W, Ral S, Davat B. Mathematical model and characterization of the transient behavior of a PEM fuel cell. IEEE Transactions on Power Electronics September 2004;19(5):1234e41. [8] Fuller TF, Newman J. Water and thermal management in solid-polymerelectrolyte fuel cells. Journal of the Electrochemical Society May 1993; 140(5):1218e25. [9] Nguyen TV, White RE. A water and thermal management model for proton exchange membrane fuel cells. Journal of the Electrochemical Society August 1993;140(8):2178e86. [10] Ge SH, Yi BL. A mathematical model for PEMFC in different ow modes. Journal of Power Sources October 2003;124(1):1e11. [11] Dannenberg K, Ekdunge P, Lindbergh G. Mathematical model of the PEMFC. Journal of Applied Electrochemistry December 2000;30(12):1377e87. [12] Wang C, Nehrir MH, Shaw SR. Dynamic models and model validation for PEM fuel cells using electrical circuits. IEEE Transactions on Energy Conversion June 2005;20(2):442e51.

[13] Pasricha S, Shaw SR. A dynamic PEM fuel cell model. IEEE Transactions on Energy Conversion June 2006;21(2):484e90. [14] Andujar JM, Segura F, Vasallo MJ. A suitable model plant for control of the set fuel cell DC/DC converter. Renewable Energy April 2008;33(4):813e26. [15] Yu D, Yuvarajan S. Electronic circuit model for proton exchange membrane fuel cells. Journal of Power Sources March 2005;142(1e2):238e42. [16] Lazarou S, Pyrgioti E, Alexandridis AT. A simple electric circuit model for proton exchange membrane fuel cells. Journal of Power Sources May 2009; 190(2):380e6. [17] Dhirde AM, Dale NV, Salehfar H, Mann MD, Han TH. Equivalent electric circuit modeling and performance analysis of a PEM fuel cell stack using impedance spectroscopy. IEEE Transactions on Energy Conversion September 2010;25(3): 778e86. [18] Usman Iftikhar M, Riu D, Druart F, Rosini S, Bultel Y, Retire N. Dynamic modeling of proton exchange membrane fuel cell using non-integer derivatives. Journal of Power Sources October 2006;160(2):1170e82. [19] Sadli I, Thounthong P, Martin J-P, Ral S, Davat B. Behaviour of a PEMFC supplying a low voltage static converter. Journal of Power Sources May 2006; 156(1):119e25. [20] Pathapati PR, Xue X, Tang J. A new dynamic model for predicting transient phenomena in a PEM fuel cell system. Renewable Energy January 2005;30(1): 1e22. [21] Hernandez A, Hissel D, Outbib R. Modeling and fault diagnosis of a polymer electrolyte fuel cell using electrical equivalent analysis. IEEE Transactions on Energy Conversion March 2010;25(1):148e60. [22] Becherif M, Hissel D, Gaagat S, Wack M. Electrical equivalent model of a proton exchange membrane fuel cell with experimental validation. Renewable Energy October 2011;36(10):2582e8. [23] Curtiss CF, Bird RB. Multicomponent diffusion. Industrial and Engineering Chemistry Research June 1999;38(7):2515e22. [24] Amundson NR, Pan T-W, Paulsen VI. Diffusing with Stefan and Maxwell. AIChE Journal April 2004;49(4):813e30. [25] Hinaje M, Nguyen DA, Ral S, Davat B. Modeling of the proton exchange membrane fuel cell in steady state. Presented at IEEE-PESC 2008, pp. 3550e3556, Rhodes, June 2008. [26] Whr M. Instationres, thermodynamisches Verhalten der Polymermembrane-Brennstoffzelle. Fortschritt-Berichte VDI, Reihe 3, Nr. 630. Dsseldorf: VDI-Verlag, ISBN 3-18-363003-6; 2000. [27] Serican MF, Yesilyurt S. Transient analysis of proton electrolyte membrane fuel cells (PEMFC) at start-up and failure. Fuel Cells November 2006;7(2): 118e27. [28] Bird RB, Stewaert WE, Lightfoot EN. Transport phenomena. 2nd ed. New York: Wiley; 2002. [29] Monroe CW, Newman J. Onsager reciprocal relations for Maxwell-Stefan diffusion. Industrial and Engineering Chemistry Research July 2006;45(15):5361e7. [30] Berg P, Promislow K, St. Pierre J, Stumper J, Wetton B. Water management in PEM fuel cells. Journal of the Electrochemical Society January 2004;151(3):A341e53. [31] Ye X, Wang C-Y. Measurement of water transport properties through membrane-electrode assemblies. Journal of the Electrochemical Society May 2007;154(7):B676e82. [32] Neubrand W. Modellbildung und Simulation von Elektromembranverfahren. Berlin: Logos, ISBN 3-89722-216-7; 1999. [33] Hinatsu JT, Mizuhata M, Takenaka H. Water uptake of peruorosulfonic acid membranes from liquid water and water vapour. Journal of the Electrochemical Society June 1994;141(6):1493e8. [34] Wang Y, Wang C-Y. Dynamics of polymer electrolyte fuel cells undergoing load changes. Electrochimica Acta May 2006;51(19):3924e33. [35] Mahinpey N, Jagannathan A, Idem R. The effects of mass transfer parameters on the modelling of a PEM fuel cell, presented at EIC Climate Change Technology Conference, Ottawa, May 2006. [36] Ramousse J, Deseure J, Lottin O, Didierjean S, Maillet D. Modeling of heat, mass and charge transfer in a PEMFC sigle cell. Journal of Power Sources August 2005;145(2):416e27. [37] Liu F, Lu G, Wang C-Y. Water transport coefcient distribution though the membrane in polymer electrolyte fuel cell. Journal of Membrane Science January 2007;287(1):126e31. [38] Shimpalee S, Ohashi M, Van Zee JW, Ziegler C, Stoeckmann C, Sadeler C, et al. Experimental and numerical studies of portable PEMFC stack. Electrochimica Acta April 2009;54(10):2899e911.

Nomenclature
Latin letters a: water activity [-] A: active surface [m2] c: concentration [mol m3] C: capacitor [F] D: diffusion coefcient of species [m2 s1] Eeq: equilibrium potential [V] F: Faradays constant, 96472 [C mol1] j0: exchange current density [A m2] J: molar ow [mol s1] I: current [A]

Author's personal copy

140 n: charge number [-] N: number of species in a gas mixture [-] nd: electro-osmotic drag coefcient [-] P: pressure [Pa] R: gas constant, 8.314 [J mol1 K1] RH: relative humidity [-] t: time [s] T: temperature [K] V: voltage [V] x: 1D coordinate [m] X: proton exchange capacity [mol kg1] Greek letters

P. Noiying et al. / Renewable Energy 44 (2012) 128e140

r: density [kg m3] s: electrical conductivity [S m1] f: electrical potential [V]


Superscripts and subscripts a: anode c: cathode cell: fuel cell dl: double layer eff: effective value GDL: gas diffusion layer hum: humidier in: inlet i, j: species (H2, O2, H2O, N2) m: membrane ref: reference condition s: electronic phase sat: saturated value sorp: sorption

a: transfer coefcient [-] d: layer thickness [m]


: porosity [-]

h: surface overvoltage [V] l: membrane water content [-] z: stoichiometry [-]

You might also like