International Journal of Electrochemistry

International Journal of Electrochemistry / 2011 / Article
Special Issue

Direct Alcohol Fuel Cell

View this Special Issue

Research Article | Open Access

Volume 2011 |Article ID 204937 | 9 pages | https://doi.org/10.4061/2011/204937

An Improved Tank in Series Model for the Direct Methanol Fuel Cell

Academic Editor: Changwei Xu
Received09 Mar 2011
Accepted27 Apr 2011
Published07 Jul 2011

Abstract

This study presents an improved TSR model for Direct Methanol Fuel Cell (DMFC) with mass and energy balance equations taking into account gas evolution in the anode channels. The TSR model includes a modified charge balance equation suitable for potentiostatic fuel cell operation mode. Polarization curves calculated with the improved TSR model agree with experimental data from the literature. The developed TSR model is able to approximate two-dimensional concentration, temperature, and current density profiles in DMFC with parallel flow field.

1. Introduction

The direct methanol fuel cell is a potential energy source to replace batteries for portable electronics. The main attractive features of DMFC include high energy density and nearly zero recharge time. Numerous 1D and 3D models are available for DMFC in the literature. Kulikovsky presented a 1D isothermal model for DMFC [1]. Murgia et al. developed 1D multicomponent steady-state isothermal model for liquid-feed DMFC [2]. Hyun et al. predicted the DMFC performance for different anode flow field designs using computational simulations [3]. Danilov et al. developed the two-phase model with a submodel for interphase transfer. They used CFD-based DMFC model to visualize and analyze the gas evolution and flow patterns in the anode channels [4–6]. In contrast to the CFD technique with highly complex process description, lumped parameter models with a reactor network are computationally fast, approximating flow patterns and concentration profiles in fuel cells. Sundmacher et al. [7–10] developed CSTR and TSR models for DMFC operating in galvanostatic mode. They used a tank in series reactor model to investigate the influence of the anode flow fields on the steady-state and dynamic behavior of DMFC with parallel, spot, and rhomboidal designs. Our previous studies focused on TSR model development for solid oxide fuel cells [11, 12]. The objective of this research is to modify this TSR model for DMFC operating in potentiostatic mode.

2. Model Formulation

2.1. TSR Model for Cocurrent Flow DMFC

Fuel and oxidant flow inside the fuel cell with parallel channels can be approximated as a reactor network complying with a tank in series model. Figure 1 illustrates the application of a tank in series reactor (TSR) model for cocurrent flow mode in a planar fuel cell with parallel flow field design.

The developed TSR model for DMFC is based on the following assumptions.(i)Each anode and cathode compartment is treated as a TSR.(ii)Electrochemical reactions occur at the electrode/membrane interfaces.(iii)Fuel cell operates under the potentiostatic operating mode and constant pressure.(iv)Negligible contact resistances between components.(v)Negligible evaporation and condensation in cathode channels.(vi)Uniform distribution of fuel and oxidant in channels.

Balance equations are composed for channel compartment and catalyst layer in each tank. For fuel cells with cocurrent flow mode, the mass balance equations for the anode and cathode channels are written as follows: 𝑉𝐴𝑖,𝑗𝑑𝐶(𝑘)𝐴,𝑖,𝑗𝑑𝑡=𝐶(𝑘)𝐴,𝑖,𝑗−1𝐹𝐴𝑖,𝑗−1−𝐶(𝑘)𝑖,𝑗𝐹𝐴𝐿,𝑖,𝑗−𝑁(𝑘)𝐴,𝑖,𝑗,𝑘=MeOH,CO2,H2𝑉O,𝐶𝑖,𝑗𝑑𝐶(𝑘)𝐶,𝑖,𝑗𝑑𝑡=𝐶(𝑘)𝐶,𝑖,𝑗−1𝐹𝐶𝑖,𝑗−1−𝐶(𝑘)𝐶,𝑖,𝑗𝐹𝐶𝑖,𝑗−𝑁(𝑘)𝐶,𝑖,𝑗,𝑘=O2,H2O,N2.(1) Gas evolution in the anode channels is defined by the following balance equation for the gas phase: 𝑉𝐴𝑖,𝑗𝑑𝜑𝐴𝐺,𝑖,𝑗𝑑𝑡=𝜑𝐴𝐺,𝑖,𝑗−1𝐹𝐴𝑖,𝑗−1−𝜑𝐴𝐺,𝑖,𝑗𝐹𝐴𝑖,𝑗−𝑟𝐴desorp,𝑖,𝑗𝜌𝐴mol,𝐿,(2) In cocurrent flow mode, the fuel and oxidant outgoing streams from tank 𝑖,𝑗−1, are the inlet streams into tank 𝑖,𝑗. Therefore, the inlet flow rate for each tank depends on the outgoing flow rate of the previous tank: 𝐹𝐴𝑖,𝑗=𝐹𝐴𝑖,𝑗−1−𝑁MeOH𝐴,𝑖,𝑗+𝑁CO2𝐴,𝑖,𝑗+𝑁H2O𝐴,𝑖,𝑗𝐶MeOH𝐴,𝑖,𝑗+𝐶CO2𝐴,𝑖,𝑗+𝐶H2O𝐴,𝑖,𝑗,𝐹𝐶𝑖,𝑗=𝐹𝐶𝑖,𝑗−1−𝑁O2𝐶,𝑖,𝑗+𝑁H2O𝐶,𝑖,𝑗𝐶O2𝐶,𝑖,𝑗+𝐶H2O𝐶,𝑖,𝑗+𝐶N2𝐶,𝑖,𝑗,(3) Here 𝑉𝐴𝑖,𝑗, 𝑉𝐶𝑖,𝑗 represent the volume of anode and cathode channels in tank (𝑖,𝑗), 𝑉𝐴𝑖,𝑗=𝑉𝐴/(𝑛𝑖𝑛𝑗), 𝑉𝐶𝑖,𝑗=𝑉𝐶/(𝑛𝑖𝑛𝑗); 𝑆𝑖,𝑗 electrode area in tank (𝑖,𝑗), 𝑆𝑖,𝑗=𝑆cell/(𝑛𝑖𝑛𝑗).

Component balance equations for the anode and cathode catalyst layers reflect changes due to consumption and production of species via reaction and mass transfer: 𝑉𝐴cat,𝑖,𝑗𝑑𝐶(𝑘)𝐴,cat,𝑖,𝑗𝑑𝑡=𝑁(𝑘)𝐴,𝑖,𝑗−𝑟(𝑘)𝐴,𝑖,𝑗,𝑘=MeOH,CO2,H2𝑉O,(4)𝐶cat,𝑖,𝑗𝑑𝐶(𝑘)𝐶,cat,𝑖,𝑗𝑑𝑡=𝑁(𝑘)𝐶,𝑖,𝑗−𝑟(𝑘)𝐶,𝑖,𝑗,𝑘=O2,H2O.(5) Here 𝑉𝐶cat,𝑖,𝑗, 𝑉𝐴cat,𝑖,𝑗 correspond to the volume of anode and cathode catalyst layer in tank (𝑖,𝑗), 𝑉𝐶cat,𝑖,𝑗=𝑉𝐶cat/(𝑛𝑖𝑛𝑗), 𝑉𝐴cat,𝑖,𝑗=𝑉𝐴cat/(𝑛𝑖𝑛𝑗).

The energy balances for the channels and the MEA structure in tank 𝑖,𝑗 are written as follows: 𝑉𝐴𝑖,𝑗𝜌𝐴mol,𝐿𝐶𝐴𝑝,𝐿𝑑𝑇𝐴𝐿,𝑖,𝑗=𝜌𝑑𝑡𝐴mol,𝐿𝐹𝐴𝑖,𝑗−1Î”â„Žğ´ğ¿,𝑖,𝑗−1−𝜌𝐴mol,𝐿𝐹𝐴𝑖,ğ‘—Î”â„Žğ´ğ¿,𝑖,𝑗+ğ‘žğ´ğ¿,𝑖,𝑗+ğ‘žğ´coll,𝑖,𝑗,𝑉(6)𝐶𝐺,𝑖,𝑗𝜌𝐶mol,𝐺𝐶𝐶𝑝,𝐺𝑑𝑇𝐶𝐺,𝑖,𝑗=𝜌𝑑𝑡𝐶mol,𝐺𝐹𝐶𝑖,𝑗−1Î”â„Žğ¶ğº,𝑖,𝑗−1−𝜌𝐶mol,G𝐹𝐶𝑖,ğ‘—Î”â„Žğ¶ğº,𝑖,𝑗+ğ‘žğ¶ğº,𝑖,𝑗+ğ‘žğ¶coll,𝑖,𝑗,𝑉(7)𝑆𝑖,𝑗𝜌𝐶𝑆𝑝𝑑𝑇𝑆𝑖,𝑗𝑑𝑡=ğ‘žğ‘†ğ‘–,ğ‘—âˆ’ğ‘žğ´ğ¿,𝑖,ğ‘—âˆ’ğ‘žğ¶ğº,𝑖,𝑗.(8) Here ğ‘žğ¶ğº,𝑖,𝑗, ğ‘žğ´ğ¿,𝑖,𝑗 represent the convective heat transferred from the MEA to the channel in tank 𝑖,𝑗; ğ‘žğ¶coll,𝑖,𝑗, ğ‘žğ´coll,𝑖,𝑗 represent heat source due to heat transfer from channel to the collector; ğ‘žğ‘†ğ‘–,𝑗 represents the heat generation in MEA in tank 𝑖,𝑗.

The charge balance equations for anode and cathode electrode/membrane interfaces are given by 𝐶𝐴𝑑𝑙𝑑𝜂𝐴𝑖,𝑗𝑑𝑡=𝐼cell𝑖,𝑗−𝐼𝐴𝑖,𝑗,𝐶𝐶𝑑𝑙𝑑𝜂𝐶𝑖,𝑗𝑑𝑡=−𝐼cell𝑖,𝑗−𝐼𝐶𝑖,𝑗+𝐼par𝑖,𝑗.(9)

For j = 1 tanks, representing the inlet sections of the anode and cathode channels, we obtain the following balance equations: 𝑉𝐴𝑖,1𝑑𝐶(𝑘)𝐴,𝑖,1𝑑𝑡=𝐶(𝑘)𝐴,𝑖,in𝐹𝐴𝑖,in−𝐶(𝑘)𝐴,𝑖,1𝐹𝐴𝑖,1−𝑁(𝑘)𝐴,𝑖,1,𝑘=MeOH,CO2,H2𝑉O,(10)𝐶𝑖,1𝑑𝐶(𝑘)𝐶,𝑖,1𝑑𝑡=𝐶(𝑘)𝐶,𝑖,in𝐹𝐶𝑖,in−𝐶(𝑘)𝐶,𝑖,1𝐹𝐶𝑖,1−𝑁(𝑘)𝐶,𝑖,1,𝑘=O2,H2O,N2,𝑉(11)𝐴𝑖,1𝑑𝜑𝐴𝐺,𝑖,1𝑑𝑡=𝜑𝐴𝐺,𝑖,in𝐹𝐴𝑖,in−𝜑𝐴𝐺,𝑖,1𝐹𝐴𝑖,1−𝑟𝐴desorp,𝑖,1𝜌𝐴mol,𝐿𝐹,(12)𝐴𝑖,1=𝐹𝐴𝑖,in−𝑁MeOH𝐴,𝑖,1+𝑁CO2𝐴,𝑖,1+𝑁H2O𝐴,𝑖,1𝐶MeOH𝐴,𝑖,1+𝐶CO2𝐴,𝑖,1+𝐶H2O𝐴,𝑖,1𝐹,(13)𝐶𝑖,1=𝐹𝐶𝑖,in−𝑁O2𝐶,𝑖,1+𝑁H2O𝐶,𝑖,1𝐶O2𝐶,𝑖,1+𝐶H2O𝐶,𝑖,1+𝐶N2𝐶,𝑖,1,(14) where 𝐹𝐴𝑖,in=𝐹𝐴in/𝑛𝑖 is the inlet anode volumetric flow rate for (𝑖,𝑗) tank with j = 1; 𝐹𝐶𝑖,in=𝐹𝐶in/𝑛𝑖 is the inlet cathode volumetric flow rate for (𝑖,𝑗) tank with j = 1.

For 𝑗=𝑛𝑗 tanks, corresponding to the outlet section of fuel and oxidant gas channels, we define mean outlet variables as follows: 𝐹𝐴out=𝑛𝑖𝑖=1𝐹𝐴𝑖,𝑛𝑗,𝐶MeOHout=∑𝑛𝑖𝑖=1𝐶MeOH𝐴,𝑖,𝑛𝑗𝐹𝐴𝑖,𝑛𝑗𝐹𝐴out,𝑇𝐴𝐿,out=∑𝑛𝑖𝑖=1𝑇𝐴𝐿,𝑖,𝑛𝑗𝐹𝐴𝑖,𝑛𝑗𝐹𝐴out,𝐹𝐶out=𝑛𝑖𝑖=1𝐹𝐶𝑖,𝑛𝑗,𝐶O2out=∑𝑛𝑖𝑖=1𝐶O2𝐶,𝑖,𝑛𝑗𝐹𝐶𝑖,𝑛𝑗𝐹𝐶out,𝑇𝐶𝐺,out=∑𝑛𝑖𝑖=1𝑇𝐶𝐺,𝑖,𝑛𝑗𝐹𝐶𝑖,𝑛𝑗𝐹𝐶out.(15) The developed mathematical model includes the following phenomena:(i)electrochemical oxidation of methanol at the anode electrode/membrane interface,(ii)electrochemical reduction of oxygen at the cathode electrode/membrane interface,(iii)charge balances at anode and cathode electrode/membrane interfaces,(iv)energy balances in gas channels and in MEA.

2.2. The Electrochemistry Submodel
2.2.1. Electrode Current

The rate of electrochemical reactions is defined by Butler-Volmer equation. The anode reaction rate is I𝐴𝑖,𝑗=𝐼𝐴0𝐶MeOHcat,𝑖,𝑗𝐶MeOHref0.5×𝛼exp𝐴𝐴𝐹𝑅𝑇𝑖,𝑗𝜂𝐴𝑖,𝑗−𝜂𝐴eq−𝛼−exp𝐴𝐶𝐹𝑅𝑇𝑖,𝑗𝜂𝐴𝑖,𝑗−𝜂𝐴eq,(16) where 𝐼𝐴0 is the anode exchange current density.

The cathode reaction rate is 𝐼𝐶𝑖,𝑗=𝐼𝐶0𝐶O2cat,𝑖,𝑗𝐶O2ref𝛼exp𝐶𝐴𝐹𝑅𝑇𝑖,𝑗𝜂𝐶𝑖,𝑗−𝜂𝐶eq−𝛼−exp𝐶𝐶𝐹𝑅𝑇𝑖,𝑗𝜂𝐶𝑖,𝑗−𝜂𝐶eq,(17) where 𝐼𝐶0 is the cathode exchange current density.

Parasitic current is calculated as methanol molar flux from anode catalyst layer to cathode catalyst layer with a linear profile in the membrane: 𝐼par𝑖,𝑗=𝑛𝐴𝑒𝐹𝐷MeOH𝑚𝛿𝑚𝐶MeOH𝐴,cat,𝑖,𝑗−0.(18)

Source terms in the component balance equations reflect changes due to the consumption or production of species via reaction or mass transfer. For component balance equations in anode catalyst layer (4), the rate of electrochemical reaction is given by 𝑟MeOH𝐴,𝑖,𝑗=𝑆𝑖,𝑗𝜈MeOH𝑛𝐴𝑒𝐹𝐼𝐴𝑖,𝑗,𝑟CO2𝐴,𝑖,𝑗=𝑆𝑖,𝑗𝜈CO2𝑛𝐴𝑒𝐹𝐼𝐴𝑖,𝑗,𝑟H2O𝐴,𝑖,𝑗=𝑆𝑖,𝑗𝜈H2O𝑛𝐴𝑒𝐹𝐼𝐴𝑖,𝑗−𝑆𝑖,𝑗𝑛𝑑,𝑖,𝑗𝐼𝐴𝑖,𝑗𝐹,(19) where 𝑛𝑑,𝑖,𝑗 represents the electro-osmotic drag coefficient.

For balance equations in cathode catalyst layer (5), the rate of electrochemical reaction is 𝑟O2𝐶,𝑖,𝑗=𝑆𝑖,𝑗𝜈O2𝑛𝐶𝑒𝐹||𝐼𝐶𝑖,𝑗||,(20) The source term for water on the cathode side includes electrochemical reaction and transfer from anode side 𝑟H2O𝐶,𝑖,𝑗=𝑆𝑖,𝑗𝜈H2O𝑛𝐶𝑒𝐹||𝐼𝐶𝑖,𝑗||+𝑆𝑖,𝑗𝑛𝑑,𝑖,𝑗||𝐼C𝑖,𝑗||𝐹,(21) An empirical equation for calculating electro-osmotic drag coefficient 𝑛𝑑 is taken from Ren et al. [13]. For the inert nitrogen component, the corresponding source term is equal to zero (𝑟N2𝐶,𝑖,𝑗=0).

2.2.2. Electrolyte Current

For fuel cells operating under potentiostatic mode (𝐸cell=const), the current density in PEM electrolyte media is estimated from the voltage equation: 𝐼cell𝑖,𝑗=𝐸OCV𝑖,𝑗−𝐸cell−𝜂𝐴act,𝑖,𝑗+𝜂Cact,𝑖,𝑗𝑅Ohmic𝑖,𝑗,(22) where 𝜂𝐴act,𝑖,𝑗 is the anode activation overpotential, 𝜂𝐴act,𝑖,𝑗=𝜂𝐴𝑖,𝑗−𝜂𝐴eq; 𝜂𝐶act,𝑖,𝑗 is the cathode activation overpotential, 𝜂𝐶act,𝑖,𝑗=𝜂𝐶𝑖,𝑗−𝜂𝐶eq; 𝜂𝐴eq, 𝜂𝐶eq is the anodic and cathodic equilibrium potential differences; 𝜂𝐴𝑖,𝑗, 𝜂𝐶𝑖,𝑗 is the anode and cathode potential differences at the electrode/membrane interfaces; 𝑅Ohmic𝑖,𝑗 is the ohmic resistance.

As shown in Appendix B, voltage (22) is applicable for calculating the current in PEM under a linear membrane phase potential profile. Membrane conductivity is given by the following empirical relation [14]: ğœŽğ‘š=0.0005139𝐶𝑚1−0.000326exp1268−1303𝑇.(23)

2.3. Heat Transfer Submodel

The source terms in energy balance equations for anode (6) and cathode channels (7) are defined as follows: ğ‘žğ´ğ‘–,𝑗=𝛼𝐴𝑆𝑖,𝑗𝑇𝑆𝑖,𝑗−𝑇𝐴𝑖,ğ‘—î€¸ğ‘ž,(24)𝐶𝑖,𝑗=𝛼𝐶𝑆𝑖,𝑗𝑇𝑆𝑖,𝑗−𝑇𝐶𝑖,𝑗.(25) The source term in energy balance equation (8) for MEA is ğ‘žğ‘†ğ‘–,𝑗=𝛿𝑚𝑆𝑖,𝑗𝑅Ohmic𝑖,𝑗𝐼cell𝑖,𝑗2−Δ𝐻𝑅𝑁H2O𝐶,𝑖,𝑗+𝑆𝑖,𝑗𝐸cell𝐼cell𝑖,𝑗+𝑆𝑖,𝑗𝛼𝐶+𝑘𝐶𝑝(𝑘)𝑁(𝑘)𝐶,𝑖,𝑗𝑇𝐶𝑖,𝑗−𝑇𝑆𝑖,𝑗+𝑆𝑖,𝑗𝛼𝐴+𝑘𝐶𝑝(𝑘)𝑁(𝑘)𝐴,𝑖,𝑗𝑇𝐴𝑖,𝑗−𝑇𝑆𝑖,𝑗.(26) Additional source terms in energy equations (6) and (7) take into account heat transfer from channels to the anode and cathode collectors ğ‘žğ´coll,𝑖,𝑗=𝛼𝐴coll𝑆𝑖,𝑗𝑇coll−𝑇𝐴𝑖,ğ‘—î€¸ğ‘ž,(27)𝐶coll,𝑖,𝑗=𝛼𝐶coll𝑆𝑖,𝑗𝑇coll−𝑇𝐶𝑖,𝑗.(28) Convective heat transfer coefficients 𝛼𝐶 and 𝛼𝐴 are calculated from empirical correlations for laminar heat transfer in channels. Heat transfer coefficients 𝛼coll are found by a technique proposed by Siegel et al. [15].

2.4. Mass Transfer Submodel

Component molar fluxes for species on the cathode side are defined by mass transfer in gas phase: 𝑁(𝑘)𝑖,𝑗=𝑆𝑖,𝑗𝜌mol,𝐺𝛽(𝑘)𝐺,eff𝑦(𝑘)𝑖,𝑗−𝑦(𝑘)cat,𝑖,𝑗,𝑘=O2,H2O,(29) where 𝛽(𝑘)𝐺,eff is the effective mass transfer coefficient in gas phase for k component; 𝑦(𝑘)𝑖,𝑗,𝑦(𝑘)cat,𝑖,𝑗 are mole fraction of k component in (𝑖,𝑗) tank in channel and catalyst surface, respectively.

The multicomponent mixture in gas phase includes carbon dioxide, methanol and water in equilibrium with the liquid phase. Component molar fluxes for species on the anode side are defined by mass transfer in liquid phase: 𝑁(𝑘)𝑖,𝑗=𝑆𝑖,𝑗𝜌mol,𝐿𝛽(𝑘)𝐿,eff𝑥(𝑘)𝑖,𝑗−𝑥(𝑘)cat,𝑖,𝑗,𝑘=MeOH,CO2,H2O.(30)

2.5. Gas Evolution

Gas evolution results from interphase mass transfer of carbon dioxide in the anode channels. The source term in balance equation for gas phase (2) describes the rate of desorption and absorption processes in the anode channels. The conventional submodel for estimating the source term is based on mass transfer equation 𝑟desorp,𝑖,𝑗=𝑆𝑖,𝑗𝑘𝑉𝜌mol,𝐺𝑦CO2𝑖,𝑗−𝑦CO2sat,𝑖,𝑗,(31) where 𝑘𝑉 is an empirical volumetric mass transfer coefficient.

It is generally assumed that the multicomponent gas phase is in equilibrium with the liquid phase. Using the equilibrium condition, Danilov et al. [4–6] proposed the following equation for estimating the source term in anode channels: 𝑟desorp,𝑖,𝑗=𝛾𝜓+𝑟CO2𝐴,𝑖,𝑗,(32) where 𝛾 is the splitting factor; 𝜓 is the coefficient. The derivation of the auxiliary equation for coefficient 𝜓 is given in [4–6]. The local splitting factor 𝛾 is found from solving the equilibrium flash equation [4–6]. It should be noted that new submodel (32) corresponds to an equilibrium model of multicomponent mass transfer between liquid and gas in the anode channels [4–6].

According to the physical meaning of the mass balance equations, molar component concentration is defined as mixture concentrations for gas-liquid flow in channels. Component mole fraction in liquid phase (𝑥(𝑘)) can be expressed through mixture concentration (𝐶(𝑘)) with known gas volume fraction (𝜑𝐺). As shown in Appendix C, the condition of thermodynamic equilibrium in the multicomponent gas-liquid mixture gives the next relationship for the specie mole fraction in liquid phase 𝑥(𝑘)=𝐶(𝑘)𝐾(𝑘)𝜑𝐺𝜌mol,𝐺+1−𝜑𝐺𝜌mol,𝐿,(33) where 𝐾(𝑘) is the distribution coefficient for k-component, 𝐾(𝑘)=𝑦(𝑘)/𝑥(𝑘).

3. Results and Discussion

Modeling and simulations are valuable tools for improving our understanding of the heat and mass transfer processes in fuel cells. To validate the improved TSR model, we used experimental data reported by Murgia et al. [2] for a 25 cm2 DMFC operating with 1.5 M and 1 M aqueous methanol feeds at 90°C. Operating conditions for this DMFC are listed in Table 1.


Parameter

Fuel flow rate, cm3/min 15
Oxidant flow rate, cm3/min 1000
M e O H M in fuel1.5 M; 1 M
C O 2 , m.fr. in fuel1.0 × 10−4
O2 mol fraction in oxidant 0.99
Temperature, C 90
Pressure, atm 1

The number of tanks is set as 𝑛𝑖=4 and 𝑛𝑗=4, corresponding to a TSR model with 272 nonlinear coupled first-order ordinary differential equations. The accepted number of tanks corresponds to the laminar flow regime in the channels with parallel flow field design. The developed TSR model was implemented in MATLAB, and it was initialized with the feed composition and temperature. The geometry and electrochemical parameters of DMFC are listed in Table 2.


ParameterAnodeCathode

Exchange current density 𝐼 0 , A m−247.10.042
Reference concentration 𝐶 r e f , M20.034
Charge transfer coefficient 𝛼 𝐴 0.50.8
Charge transfer coefficient 𝛼 𝐶 0.51.2
Channel height, m1×10−3
Electrode area 𝑆 , m225×10−4
Membrane width 𝛿 𝑚 , m2×10−4
Catalyst width 𝛿 c a t , m1×10−5

Figure 2 compares the experimental and calculated polarization curves for DMFC with operating conditions in Table 1. The solid line represents a DMFC model prediction with the improved TSR model.

Figure 3 displays histograms with steady-state simulation results predicted by the TSR model for DMFC with 1.0 M feed of fuel at the anode and pure oxygen feed at the cathode. It should be noted that each tank is characterized by complete mixing, and distribution of concentration, temperature, and current density is presented by step change for tanks in series. Concentration of methanol in channel is decreased due to the anode electrochemical reaction. For the anode channels, carbon dioxide is the product of the anode electrochemical reaction. Gas content in the anode channels is increased from inlet to the outlet following the trend in molar CO2 concentration. The two-dimensional histograms indicate a strong coupling between the reactants and the current density distribution. The predicted current density profile is following the trend in methanol concentration in anode channels. Murgia et al. [2] obtained experimental polarization data by circulating aqueous methanol from a reservoir where CO2 was released to the atmosphere. The inlet CO2 concentration in the liquid fuel in Table 1 corresponds to the gas-liquid equilibrium condition in the fuel reservoir. The predicted temperature profile in MEA is following the main trend in the current density distribution.

Figure 4 presents histograms with steady-state simulation results predicted by the TSR model for DMFC with 1.5 M feed of fuel at the anode and pure oxygen feed at the cathode.

Gas management greatly influences the performance of the fuel cell. On the anode side, carbon dioxide is produced by electrochemical oxidation of the methanol. Inefficient removal of CO2 bubbles may block anode channels and decrease efficiency of the fuel cells due to limited mass transport and maldistribution of reactants. According to the conventional submodel, the interface flux is completely dependent on empirical mass transfer coefficient (𝑘𝑉). The predicted gas content in the anode channels can vary from 0 to 90% depending on the value of the mass transfer coefficient. Simulation results reveal that high values of the mass transfer coefficient (𝑘𝑉>6000s−1) predict gas evolution in the anode channels with high gas content (𝜑>40%). For low mass transfer coefficients (𝑘𝑉<6000s−1), on the other hand, 𝜑<40%. The volumetric mass transfer coefficient 𝑘𝑉=6000s−1 used in gas evolution submodel (31) provides the condition 𝜑≈40% in the anode channels, which is close to the equilibrium condition given by the flash equation.

As shown in Figure 5, empirical model (31) and new submodel (32) for gas evolution predict similar trends in the gas content distribution. In contrast to nonequilibrium submodel (31) with empirical mass transfer, new rate expression (32) determines the gas content under equilibrium conditions without empirical coefficients.

Comparison of DMFC performance with different fuel composition is given in Table 3. Changing the fuel composition from 1 M to 1.5 M gives a small improvement in fuel cell performance. The mass transfer coefficient is presented in the form of limiting current density. Average current density approaches the value of limiting current densities under cell voltage 𝐸cell=0.2V. The possible way of DMFC performance improvement is to intensify mass transfer from the channels to the catalyst layer.


ParameterOutlet gasMean parasiticMean limitingMean anodeFuelOxidant
contentcurrentcurrent currentutilization utilization
𝜑 𝐴 o u t 𝐼 p a r 𝐼 𝐴 l i m 𝐼 𝐴 F U A U
m3/m3A/m2A/m2A/m2%%

1.0 M0.2739.170096905.20.018.1
1.5 M0.36329.8105149819.90.0111.8

4. Conclusions

This study presents a TSR model for DMFC operating in potentiostatic mode. The developed TSR model includes mass and energy balance equations in anode and cathode compartments together with gas evolution in anode channels. Modified charge balance equations are defined as interface boundary conditions. The possibility of evaluating two-dimensional profiles in the DMFC is one of the main advantages of the developed TSR model. The simulation results indicate the strong coupling between concentration, temperature and current density distribution in coflow DMFC. Taking into account the assumption of uniform distribution of fuel and oxidant in channels, the TSR model predicts the limiting performance of the DMFC under the given flow regime (number of tanks), mass transfer and catalyst activity. The improved TSR model allows studying the influence of different parameters on the DMFC performance. Results can be used to better understand and investigate the effects of various parameters and operating conditions on DMFC performance.

Appendices

A. Derivation of Charge Balance Equation

The electric potential fields are governed by the charge conservation equations. The charge balance at the interface between electron-conducting and ion-conducting media is given by [16] 𝜕𝑄𝜕𝑡+∇⋅𝑖s=𝐼1−𝐼2,(A.1) where 𝐼1 is the current in electron-conducting media normal to the boundary, 𝐼2 is the current in ion-conducting media normal to the boundary, 𝑖𝑠 is the superficial current density, 𝑄is the charge. For the TSR model, the net charge flux is zero (∇⋅𝑖𝑠=0). The interfaces between ionic and electronic media behave like a capacitor in which the charge density is a function of potential difference across the double layer. Charge or discharge rate at the electrode-electrolyte double layer can be represented as 𝜕𝑄𝜕𝑡=𝐶dl𝜕𝜂𝜕𝑡.(A.2)

For the difference of potential of the electron conducting media 𝜙 and the potential of the electrolyte phase 𝜙𝑚, the following charge conservation equation is valid: 𝜕𝜂=1𝜕𝑡𝐶dl𝐼1−𝐼2,(A.3) where 𝜂 is potential differences or overpotential, 𝜂=𝜙−𝜙𝑚.

The improved fuel cell models with new charge balance equation provide a better understanding of main phenomena governing electrochemical reactions in fuel cells [4–6, 11, 12].

B. Electrolyte Current

By definition, the electrolyte current is written as follows: 𝐼𝑚=âˆ’ğœŽğ‘šğœ•ğœ™ğ‘šğœ•ğ‘›.(B.1) Here 𝜙𝑚 represents the potential in the ionic conducting media; 𝑛 is the normal to the interface; ğœŽğ‘š is the membrane conductivity. Using linear approximation of the potential profile in the ionic conducting media, the electrolyte current is defined as follows: ğ¼ğ‘šâ‰ˆâˆ’ğœŽğ‘šî‚µğœ™ğ¶ğ‘šâˆ’ğœ™ğ´ğ‘šğ›¿ğ‘šî‚¶,(B.2) where 𝛿𝑚 is the membrane thickness; 𝜙𝐶𝑚 is the membrane phase potential at cathode electrode/membrane interface; 𝜙𝐴𝑚 is the membrane phase potential at anode electrode/membrane interface. The values of potential in ionic conducting media at electrode/membrane interfaces can be expressed through potential difference 𝜂 given by the definition of the activation overpotential 𝜂𝐴act=𝜙𝐴−𝜙𝐴𝑚−𝜂𝐴eq,𝜂𝐶act=𝜙𝐶−𝜙𝐶𝑚−𝜂𝐶eq,(B.3) where 𝜙𝐶 is the potential at cathode electrode/membrane interface in the electron conducting media; 𝜙𝐴 is the potential at anode electrode/membrane interface in the electron conducting media.

We note that the reversible Nernst potential at the electrode/membrane interface is linked with anodic and cathodic equilibrium potential differences: 𝜂𝐶eq−𝜂𝐴eq=𝐸OCV.(B.4) It is known that the difference of the anodic and cathodic electric potential is equal to the cell voltage 𝐸Cell=𝜙𝐶−𝜙𝐴.(B.5) The values of potential in ionic conducting media at electrode/electrolyte interfaces can be expressed from (B.3) as follows: 𝜙𝐴𝑚=𝜙𝐴−𝜂𝐴act−𝜂𝐴eq,𝜙𝐶𝑚=𝜙𝐶−𝜂𝐶act−𝜂𝐶eq.(B.6) Taking into account equations (B.2)–(B.6), we obtain the next expression for the electrolyte current density: 𝐼𝑚=𝐸OCV−𝐸cell−𝜂𝐴act+𝜂𝐶act𝑅Ohmic,(B.7) where 𝑅Ohmic is the ohmic resistance, 𝑅Ohmic=𝛿𝑚/ğœŽğ‘š.

Application of new expression for electrolyte current with fuel cell models is shown in our papers [4–6, 11, 12].

C. Definition of Mixture Concentration

Mass balance equations for channels and catalyst layer are defined with molar component concentration. In view of gas-liquid flow in channels, the physical meaning of component concentration corresponds to mixture concentration. By definition, the mixture concentration is 𝐶∗=𝐶∗𝐺𝜑𝐺+𝐶∗𝐿1−𝜑𝐺,(C.1) where 𝐶∗ is mixture concentration, kg/m3; 𝐶∗𝐺 is concentration in gas phase, kg/m3; 𝐶∗𝐿 is concentration in liquid phase, kg/m3. Dividing both parts by molecular weight, we transfer to molar mixture concentration 𝐶∗𝑀=𝐶∗𝐺𝜑𝐺𝑀+𝐶∗𝐿1−𝜑𝐺𝑀(C.2) or 𝐶(𝑘)=𝑦(𝑘)𝜑𝐺𝜌mol,𝐺+𝑥(𝑘)1−𝜑𝐺𝜌mol,𝐿,(C.3) where 𝐶 is mixture concentration, mol/m3; 𝐶𝐺 is concentration in gas phase, mol/m3; 𝐶𝐿 is concentration in liquid phase, mol/m3. The condition of thermodynamic equilibrium in multicomponent gas-liquid mixture gives the relationship among species concentrations in phases 𝑥(𝑘)=𝐶(𝑘)𝐾(𝑘)1−𝜑𝐺𝜌mol,𝐺+1−𝜑𝐺𝜌mol,𝐿.(C.4) Equation (C.4) is used for calculating the driving mass transfer force in mass transfer equation.

Abbreviations

Latin Letters
C:Molar concentration/mol m−3
𝐶dL:Double layer capacitance
Cp:Specific heat/J mol−1 K−1
Cm:Water content in membrane/mol m−3
𝐸cell:Cell voltage/V
𝐸ocv:Open circuit voltage/V
FA,FC:Anode and cathode volumetric flow rate/m3 s−1
F:Faraday’s constant/96485 C mol−1
K:Distribution coefficient
kV:Volumetric mass transfer coefficient/s−1
h:Enthalpy/J mole−1
Δ𝐻𝑅:Reaction enthalpy/J mol−1
I:Current density/A m−2
𝐼0:Exchange current density/A m−2
MEA:Membrane electrode assembly
N:Component molar flux/mol s−1
n:Number of tanks
nd:Net drug coefficient
ne:Number of electrons
P:Pressure/Pa
r:Mass source term/mol s−1
R:Ideal gas constant/J mol−1 K−1
𝑅ohmic:Ohmic resistance/Ω m2
q:Energy source term/J s−1
S:Electrode area/m2
t:Time/s
T:Temperature/K
TSR:Tank in series reactor
V:Volume/m3
x:mole fraction in liquid phase
y:mole fraction in gas phase.
Greek Symbols
𝛼:Heat transfer coefficient/W m−2 K−1
𝛼𝐴𝐴:Anodic charge transfer coefficients for anode
𝛼𝐴𝐶:Cathodic charge transfer coefficients for anode
𝛼𝐶𝐴:Anodic charge transfer coefficients for cathode
𝛼𝐶𝐶:Cathodic charge transfer coefficients for cathode
𝛽eff:Effective mass transfer coefficient/m s−1
𝜈:Stoichiometry coefficient
𝜂:Potential difference/V
𝜌mol:Molar density/mol m−3
ğœŽ:Conductivity/Ω−1m−1
𝛿𝑚:Membrane thickness/m
𝛿cat:Catalyst thickness/m
𝜑𝐺:Gas content/m3m−3.
Subscripts/Superscripts
in:Inlet
out:Outlet
eff:Effective
i,j:i,j tank
A:Anode
C:Cathode
cell:Fuel cell
G:Gas phase
lim:Limiting
mol:Molar
m:Membrane
par:Parasitic
ref:Reference
S:Interface; solid
sat:Saturated
MeOH:Methanol
O2:Oxygen
H2O:Water
N2:Nitrogen
CO2:Carbon dioxide.

Acknowledgment

The authors are grateful to IWT Vlaanderen, for financial support for OCPEC “Optimised Chemical Production with Electricity Cogeneration”.

References

  1. A. A. Kulikovsky, “1D + 1D model of a DMFC: localized solutions and mixedpotential,” Electrochemistry Communications, vol. 6, no. 12, pp. 1259–1265, 2004. View at: Publisher Site | Google Scholar
  2. G. Murgia, L. Pisani, A. K. Shukla, and K. Scott, “A numerical model of a liquid-feed solid polymer electrolyte DMFC and its experimental validation,” Journal of the Electrochemical Society, vol. 150, no. 9, pp. A1231–A1245, 2003. View at: Publisher Site | Google Scholar
  3. M. S. Hyun, S. K. Kim, D. Jung et al., “Prediction of anode performances of direct methanol fuel cells with different flow-field design using computational simulation,” Journal of Power Sources, vol. 157, no. 2, pp. 875–885, 2006. View at: Publisher Site | Google Scholar
  4. V. A. Danilov, J. Lim, I. L. Moon, and K. H. Choi, “Gas management in flow field design using 3D direct methanol fuel cell model under high stoichiometric feed,” Korean Journal of Chemical Engineering, vol. 23, no. 5, pp. 753–760, 2006. View at: Publisher Site | Google Scholar
  5. V. A. Danilov, J. Lim, I. L. Moon, and H. Chang, “Three-dimensional, two-phase, CFD model for the design of a direct methanol fuel cell,” Journal of Power Sources, vol. 162, no. 2, pp. 992–1002, 2006. View at: Publisher Site | Google Scholar
  6. H. Kim, V. A. Danilov, J. Lim, and I. Moon, “Development of 3D DMFC model for flow field design,” The Korean Journal of Chemical Engineering, vol. 45, pp. 93–102, 2007. View at: Google Scholar
  7. K. Sundmacher, T. Schultz, S. Zhou, K. Scott, M. Ginkel, and E. D. Gilles, “Dynamics of the direct methanol fuel cell (DMFC): experiments and model-based analysis,” Chemical Engineering Science, vol. 56, no. 2, pp. 333–341, 2001. View at: Publisher Site | Google Scholar
  8. K. Sundmacher, A. Kienle, H. J. Pesch, J. F. Berndt, and G. Huppmann, Eds., Molten Carbonate Fuel Cells Modeling, Analysis, Simulation, and Control, Wiley-VCH, Weinheim, Germany, 2007.
  9. U. Krewer, M. Pfafferodt, A. Kamat, D. F. Menendez, and K. Sundmacher, “Hydrodynamic characterisation and modelling of anode flow fields of Direct Methanol Fuel Cells,” Chemical Engineering Journal, vol. 126, no. 2-3, pp. 87–102, 2007. View at: Publisher Site | Google Scholar
  10. U. Krewer, Y. Song, V. John et al., “Direct methanol fuel cell (DMFC): analysis of residence time behavior of anodic flow bed,” Chemical Engineering Science, vol. 59, no. 1, pp. 119–130, 2004. View at: Google Scholar
  11. S. Hosseini, V. A. Danilov, P. Vijay, and M. O. Tadé, “An improved tank in series model for the planar solid oxide fuel cell,” Industrial and Engineering Chemistry Research, vol. 50, no. 2, pp. 1056–1069, 2011. View at: Publisher Site | Google Scholar
  12. V. A. Danilov, P. Vijay, and M. O. Tade, “An improved TSR model for tubular solid oxide fuel cell stacks,” Chemical Engineering & Technology, vol. 334, no. 5, pp. 737–745, 2011. View at: Google Scholar
  13. X. Ren, T. E. Springer, T. A. Zawodzinski, and S. Gottesfeld, “Methanol transport through nafion membranes electro-osmotic drag effects on potential step measurements,” Journal of the Electrochemical Society, vol. 147, no. 2, pp. 466–474, 2000. View at: Publisher Site | Google Scholar
  14. T. E. Springer, T. A. Zawodzinski, and S. Gottesfeld, “Polymer electrolyte fuel cell model,” Journal of the Electrochemical Society, vol. 138, no. 8, pp. 2334–2342, 1991. View at: Google Scholar
  15. N. P. Siegel, M. W. Ellis, D. J. Nelson, and M. R. von Spakovsky, “A two-dimensional computational model of a PEMFC with liquid water transport,” Journal of Power Sources, vol. 128, no. 2, pp. 173–184, 2004. View at: Publisher Site | Google Scholar
  16. R. Bansal, Handbook of Engineering Electromagnetics, Marcel Dekker, New York, NY, USA, 2004.

Copyright © 2011 Valery A. Danilov et al. This is an open access article distributed under the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.


More related articles

1043 Views | 489 Downloads | 5 Citations
 PDF  Download Citation  Citation
 Download other formatsMore
 Order printed copiesOrder

Related articles

We are committed to sharing findings related to COVID-19 as quickly and safely as possible. Any author submitting a COVID-19 paper should notify us at help@hindawi.com to ensure their research is fast-tracked and made available on a preprint server as soon as possible. We will be providing unlimited waivers of publication charges for accepted articles related to COVID-19. Sign up here as a reviewer to help fast-track new submissions.