2010 International Conference on Science and Social Research (CSSR 2010), December 5 - 7, 2010, Kuala Lumpur, Malaysia

Heat Transfer Simulation of a Single Channel Air-

Cooled Polymer Electrolyte Membrane Fuel Cell
Stack with Extended Cooling Surface
W.A.N. Wan Mohamed, R. Atan and A. A. Ismail
Alternative Energy Research Centre, Faculty of Mechanical Engineering, UiTM Shah Alam, Malaysia.
[email protected]

Abstract— A Polymer Electrolyte Membrane (PEM) fuel cell is layers, and a proton conductive membrane. When the
an electrical power generator utilizing a hydrogen-based fuel hydrogen gas enters the anode electrode, it comes into contact
reactant and oxygen in a reversed electrolysis reaction, with by- with the catalyst, which splits the gas into positive ions
products of water and heat. The application is sensitive to (hydrogen protons), and electrons. The electrons travel to the
temperature; more power is generated at elevated operating cathode via an external circuit, generating electrical current.
temperatures, but excessive cell temperature causes dehydration
The protons travel through the membrane to the cathode. At
to the membrane electrolyte and subsequent power decline as
well as cell deterioration. The power-to-weight ratio and reduced the same time, oxygen is being fed to the cathode, where a
parasitic load, which are the main advantages of an air-cooled catalyst layer creates oxygen ions. When the hydrogen protons
system, pushes the research tendency to replace water cooling arrive at the cathode side, they bond with these oxygen ions,
with air cooling. This work analyzes the heat transfer creating water and heat as the by-products of the
characteristics, using analytical and Computational Fluid electrochemical reaction. Fig. 1 illustrates the fuel cell
Dynamics (CFD) tools, of a 3 kW PEM fuel cell stack which is assembly and subsequent electrochemical reaction.
equipped with a single cooling channel on each bipolar plate. The
base stack design consisting of 73 bipolar plates refers to an
industrial water-cooled PEM fuel cell stack available at the
Faculty of Mechanical Engineering, University of Technology electric circuit
MARA. From the results of the coolant flow over the base stack
design, extended surfaces (fins) was added at an optimized
geometry to enhance the heat transfer. Both designs were fuel H2 O2
subjected to a heat flux magnitude of 1.6 times greater than
heat
theoretically required, and showed excellent simulated cooling
capability of 100% cooling effectiveness when subjected to flows used fuel
recirculation air and water
at Reynolds number of 800 and above. Addition of extended vapor
cooling surfaces further improves the thermal gradient reduction
within the plate by 30%. Though still requires practical evidence,
proton gas diffusion flow field plate
the simulation analysis has provided the groundwork of air gas diffusion
electrode
electrode (anode) exchange
cooling applicability in replacing water cooling for a 3 kW PEM flow field plate membrane (cathode)
fuel cell stack.
Figure 1. PEM fuel cell components and basic operation

Keywords- PEMFC; fuel cell; cooling; simulation; CFD In order to achieve the maximum electrical power output,
cell temperature needs to be maintained at an optimal level.
I. INTRODUCTION The optimal cell temperature is mainly related to the
Polymer Electrolyte Membrane (PEM) fuel cells are the membrane properties and electrode-catalyst interface
most widely applied fuel cell system within power parameters. Works by Ahn and Choe [4], Sohn et al. [5], as
requirements of up to 100 kW. It has seen tremendous well as Yu and Jung [6] among many others, related the effect
technological growth since 1992, and can be found in of cell temperature to the fuel cell performance. Fuel cells
stationary as well as mobile applications. The few trademark generates higher power densities at elevated temperatures due
projects involving PEM fuel cell are the residential power to better electrochemical reactivity at both electrodes;
supply in Japan [1], the Honda FCX car and the Formula Zero however, the power density would soon decline as membrane
race [2], as well as in Malaysia’s own project on the backup drying occurs which leads to higher resistance to proton flow.
power system for telecommunication towers [3]. Thermal engineering in fuel cell design and operation
Fuel cells works on the principle of electrochemical reaction focuses on maintaining the cell at the allowable temperature
between two reactants. A hydrogen-rich fuel and an oxidant limit, usually referring to the membrane operating
are needed, separated by electrodes, catalyst, gas diffusion temperature. Polymer electrolyte membranes have lower

978-1-4244-8986-2/10/$26.00 ©2010 IEEE 91

working temperatures than liquid-based electrolytes. where As is the surface area exposed to the cooling fluid,
Currently, it is in the range of 50oC to 100oC, with several ΔT is the temperature difference between the surface and the
research papers reporting achievements in membrane design of fluid, and h is the average heat transfer coefficient, determined
120oC operating temperature. from the Nusselt number (laminar flow);
Cooling system selection is based on the amount of cooling h Lc
load which is directly related to the cell conversion efficiency. Nu = = 0.664 Re1 / 2 Pr1 / 3 (2)
k
A cell with 50% conversion efficiency, an achievable and Here, Pr refers to the Prandtl number of the fluid evaluated
normal efficiency for fuel cells, would theoretically generate a at the film temperature, Lc is the charactersistic length of the
similar amount of heat and electrical energy. The industrial surface, and Re is the Reynolds number.
tendency is to apply water cooling for PEM fuel cell stacks of Vm Lc
2 kW rated power and above. Though air-cooled PEM fuel Re = (3)
υ
cells rated at 5 kW power are commercially available since
The mean velocity, Vm, is normally taken as the incoming
2007, the optimum efficiency of 55% was actually rated at a
steady stream velocity, and υ is the kinematic viscosity of the
mere power output of 1.63 kW [7].
fluid (evaluated at film temperature).
The key advantages of air cooling over water cooling are the
Evaluation of the internal forced convection cooling within
reduced power-to-weight ratio and lower parasitic load to
the channels requires treating the channels as fins, with
operate the system. Space requirements and sub-components
various geometrical parameters to be determined for acquiring
are also significantly reduced, allowing a more compact design
the estimated cooling channel efficiency, as well as
with lower cost and operating complexity. The existing and
determination of the logarithmic mean temperature difference
popular cooling channel design for a closed-cathode stack is a
across the channel. The models presented below are simplified
straight channel of square or rectangular geometry, evenly
and can be referred in greater detail in [8].
spaced and positioned between the cathode and anode sides of
the bipolar plates. This type of design would be adequate in Q = h Atηoθ LMTD (4)
providing the required cooling rates for less than 1.5 kW fuel
cell systems. The total cooling rate within the channels, Q, are related to
The main objective of the simulation and analysis presented the fin (channel) performance, ηO, the average heat transfer
here is to analyze the fundamental air cooling characteristics coefficient, , the total surface area of the cooling channels, At,
of a single straight path cooling channel configurations as a and log mean temperature difference, θLMTD.
platform to enhance the cooling capability from its traditional The evaluation of internal convection Reynolds number is
maximum 2 kW application to a slightly greater 3 kW power an important step in ensuring analytical and simulation
output limit. In addition to the standard stack design, an accuracy. Fluid streams with Re < 2300 are laminar. In this
extended cooling surface (fins) was designed based on the study, all the streams has been evaluated accordingly based on
optimal coolant flow stream over the stack. The main its velocity and hydraulic diameter, and the internal flow was
analytical parameters are the average stack temperature and determined to be fully developed laminar flow for all cases.
cooling effectiveness based on computational simulation using The maximum Reynolds number was 1007 at 3.5 m/s nominal
STAR-CCM, Computational Fluid Dynamics (CFD) codes. air delivery velocity.
II. THERMAL MODELS The acquired cooling rates from the simulation are presented
in terms of cooling effectiveness, which is a ratio indicating
The thermal models presented here are applied as a the simulated cooling effect against the given heat load. This
validation technique to the simulation results as well as study considers a 3 kW PEM fuel cell stack as its case study.
determining the fundamental aspects of the simulation physics Theoretically, a good fuel cell design would at least reach 50%
in terms of fluid phase and heat load. The main analysis is conversion efficiency; thus a 3 kW of heat would also be
focused to evaluate the cooling effect analytically and to generated. For the purpose of long-term design needs, heat
compare it with the boundary heat transfer of the simulations. generation for the simulation was based on 30% fuel cell
The heat transfer models are based on Newton’s law of efficiency, yielding a thermal safety factor of 1.6 and a total
cooling. cooling load of 4.9 kW.
In principal, internal and external forced convection
mechanisms occurs simultaneously as the coolant air flows III. FUEL CELL DESIGN AND MODELING
into the cooling channel as well as over the side and top The reference fuel cell design for the work is the GASHUB
surfaces of the stack. Evaluation of external forced convection 3 kW water-cooled PEM fuel cell system, available at the
cooling over the exposed surfaces of the stack requires the Energy Laboratory, Mechanical Engineering Faculty,
determination of the Reynolds number over the surface and Universiti Teknologi MARA, Malaysia. The PEM fuel cell
applying the proper Nusselt number correlation based on the stack dimension is 365 mm length x 149.5 mm width x 239
flow phase. mm height and consisting of 73 bipolar plates. Fig. 2 shows a
Newton’s law of cooling states that the cooling rate, simplified single bipolar plate geometry, integrated with
cooling channels along its width.
Q = h As ΔT (1)

92
 layer mesher and tetrahedral meshes, but polyhedral mesh is
used for the finned stack to improve the mesh quality due to
the existence of curved edges.

Outlet

Figure 2. Bipolar plate geometry without gas flow field (units in mm)

This paper presents a base case study of a single straight

channel of 179 mm length x 149.5 mm width x 1 mm height. Stack Air
Two stack designs were investigated; firstly, a stack with only Boundary Boundary Inlet

the cooling channels, and secondly, a stack enhanced with Figure 4. Fuel cell stack and cooling air region with 73 bipolar plates
extended cooling surfaces (or fins). The fin geometry in terms
of optimal fin height was acquired by analyzing the velocity Table 1 lists the boundary conditions applied to the
vectors and boundary layer thickness over the top of the stack simulation.
from the simulation results of the first case (refer section 5A
TABLE I.
for details of analysis). A total of 36 fins of 5 mm thickness SIMULATED BOUNDARY CONDITIONS
(similar to plate thickness) were constructed; 3 bipolar plates
would have two fins and a 5 mm gap between them. Fig. 3 Cooling Air
shows the stack configuration with the fins.
Properties /
Notes Values
parameters
1. Flow phase Laminar -
2. Flow source Uniform inlet -
condition
3. Inlet specifications Velocity 0.01 m/s
(practical 1 m/s
assumption) 2 m/s
3 m/s
3.5 m/s
4. Outlet specifications Pressure outlet 1 atm
5. Inlet temperature Ambient 30oC
6. Outlet temperature Ambient 30oC
7. Inlet pressure Ambient 1 atm
8. Outlet pressure Ambient 1 atm
9. Thermal Constant 0.02588 W/m.K
conductivity
10. Specific heat, Cp Constant 1005 J/kg.K

Figure 3. Configuration of stack design with fins Stack

IV. BOUNDARY CONDITIONS

Properties /
Notes Values
The actual fuel cell stack consists of 73 separate bipolar parameters
plates. However, in order to simplify the simulation and 1. Material Homogeneous Carbon graphite
reduce the computational power, the stack was sketched as a 2. Specific heat [9] Constant 710 J/kg.K
block imbedded with 73 air-cooling channel arrays. The stack 3. Density [9] Constant 2240 kg/m3
4. Thermal Constant 20 W/m.K
was confined in the air cooling region with dimensions as conductivity [10]
specified in Fig. 4. 5. Cell efficiency At design 30%
The STAR-CCM Computational Fluid Dynamics (CFD) (design rated) safety factor of
was used for the simulation. Two regions of cooling air and 1.6
6. Projected heat Steady and 4900 W
stack was assigned with an interface region defined at its generated uniform (standard)
contact surfaces. The fluid physics models are three- 5360 W (finned)
dimensional analysis with ideal gas assumption, steady-state
condition, laminar flow, segregated fluid temperature and 7. Projected heat flux Steady and 441,195 W/m3
uniform
stationary region. The mesh model typically applies prism

93
 V. RESULTS AND DISCUSSIONS Air
Outlet

A. Determination of Optimal Fin Height

20
Plane Velocity (m/s)

15

10

0
10 20 30 40 45

Plane Height (mm) Air Inlet

Figure 5. Average plane velocities over top surface of the stack

Figure 6. Temperature contour of 3.5 m/s simulation for standard stack design

The optimal fin height was determined from the boundary

layer thickness of the air flow over the top side of the stack at Air
Outlet
the highest inlet stream velocity of 3.5 m/s. Based on the
simulation results of the standard stack, external flow over the
fuel cell stack was visualized by aid of vector streamlines.
Two approaches were applied to validate the boundary layer
thickness obtained. The first approach was by using the ruler
tool for measuring distance to directly approximate the
boundary layer. The second approach taken was by creating
several horizontal planes by the increment of 10 mm from fuel
cell stack surface, and checking the maximum values of area-
averaged velocity for each plane (refer Fig. 5). Both
approaches results in a similar height which is approximately
40 mm from the stack top surface. Hence, an extended fin
array based on an optimal 40 mm height was designed. Air
Inlet

B. Stack Temperature and Cooling Effectiveness

Figure 7. Temperature contour of 3.5 m/s simulation for finned stack design

Fig. 6 and Fig. 7 presents the temperature scalar scene The average stack temperature in Fig. 8 was obtained from
obtained for the standard stack and the finned stack at air the volume averaged analysis of the simulation. Higher
delivery velocity of 3.5 m/s, where the cooling effect is most nominal air velocities introduced into the stack allows the
significant. Cold regions below the required 50oC stack steady-state stack temperature to be maintained at 53oC to
temperature are concentrated at the mid-plane section and 55oC. Cooling air at 1 m/s delivery fails to have the required
extend to approximately a quarter of the stack width. cooling effect on the stack. Delivery at 2 m/s displays a
Expectedly, the hottest regions are the edge surfaces most steady-state stack temperature of 62oC, while the average stack
distant from the coolant inlet, which are surfaces not within temperature is approximately 55oC at higher air delivery
the reactive area of electrical power generation. velocities. However, it should be noted that the generated heat
At mid-plane and along the air flow path, the stack flux boundary condition was set 1.6 times higher than the
temperature difference between the inlet and exit of the anticipated practical heat flux. It is also analyzed that
cooling air is 15oC. However, higher temperature difference is increasing the air velocity from 3 m/s to 3.5 m/s does not have
registered along the vertical axis of the stack. At the inlet, the any significant effect on the stack temperature, thus the
difference from mid-plane to the top and bottom surface is optimum cooling velocity from this analysis was determined
approximately 30oC. Similarly at the exit, the difference is as 3 m/s.
23oC. This shows that variable cooling rates exists along the Theoretically, a PEM fuel cell in isothermal condition
cooling air flow path. allows a uniform electrochemical reaction over the MEA. In
practice, this condition is difficult to achieve for both air and
water cooled stacks [11]. Fig. 9 plots the simulated
temperature distribution of the stack at 15 mm planes from the
inlet and normal to the airflow. Non-uniform temperature
distribution exists for all the cooling velocities. The largest

94
thermal gradient was shown by the standard stack design at 1 only 7% difference at 3.5 m/s. The large difference at low
m/s air velocity. A significant reduction in gradient was velocity can be contributed to inaccurate estimation of the air
observed as the velocity (and Reynolds number) is doubled. At stream velocity flowing over the top surface of the stack.
3 m/s and 3.5 m/s, the thermal gradient is less than 0.1 oC/mm Enhanced flow velocity near the surface could occur at 1 m/s
increments, as analyzed in Fig. 10. It is also noted that the as the flow changes direction from a vertical flow to a
addition of fins at a similar generated heat flux lowers the horizontal flow as it passes the stack edges. The low margin of
thermal gradient by approximately 30% compared to the difference at 3 m/s and 3.5 m/s suggests that the simulation
standard stack design at a similar nominal air velocity. This results are satisfactory in accuracy.
fact proves that extended cooling surfaces are capable of
improving the stack thermal gradient due to a larger available
120
area for heat distribution.

V o l u m e A v e ra g e d S ta c k
Cooling effectiveness is a measured ratio between the 100

T em p e ra tu re ( o C )
simulated boundary heat transfers to the generated heat load of 80
the stack. It is desirable to obtain the maximum cooling 60
effectiveness. Previous works by the authors [12] has 40
analytically concluded that active cooling should provide at 20
least 90% of cooling effect for stacks generating higher than
0
one kW of thermal power. Fig. 11 plots the cooling 1 2 3 3.5 fin 3.5
effectiveness for each simulated cases and proves the
Nomina l Air Delivery Veloc ity (m/s)
capability of the single cooling channel design to effectively
dissipate the cooling load. Simulation on air inlet velocity of Figure 8. Average stack temperature
0.01 m/s acts as a reference case when the stack is assumed to
be naturally cooled by the ambient surrounding, registering 120
only 10% of the desired cooling effect. At inlet air delivery
110
velocities of 2 m/s to 3.5 m/s (Reynolds number from 800 to
S ta c k T e m p e ra tu re ( o C )

100 S S 1.0 m/s

1000), 100% cooling effectiveness was achieved. A slightly S S 2.0 m/s
lower cooling effectiveness of 95% was calculated for 1 m/s 90
S S 3.0 m/s
velocity. 80
S S 3.5 m/s
The results shows that the single straight path cooling 70
FS 3.5 m/s
channel design is suitable for a 3 kW PEM fuel cell stack. 60

Operating at Reynolds number higher than 800 would allow 50

the same design to be applied for stacks with a rated maximum 40

power rating of 4.9 kW. 15 30 45 60 75 90 105 120 135

P lane Distance from S ta c k Inlet (mm)
150

Theoretically, the stack temperature should have been Figure 9. Stack temperature distribution normal to the cooling airflow
similar for all cases with similar cooling effectiveness.
Comparing the average stack temperature and the temperature 0.4
distribution to the cooling effectiveness shows that average
L i n e a r T e m p e r a tu r e I n c r e m e n t

0.35
o v e r D i s ta n c e ( o C /m m )

stack temperature is largely influenced by the existing 0.3

s tack
gradients within the stack. Even though at 2 m/s the cooling 0.25
air
effectiveness is 100%, its average stack temperature is higher 0.2

than those at 3 m/s and 3.5 m/s due to the lower active cooling 0.15
0.1
rates within the cooling channels. However, external forced
0.05
convection effects on the outer surfaces, mainly on the surface 0
directly exposed to the incoming air flow increases the total 1 2 3 3.5 fin 3.5
boundary cooling effect between the stack and coolant.
Nominal Air Delivery Velocity (m/s)
Therefore, the average stack temperature was different from
Figure 10. Linear temperature increment of the stack and cooling air
one another even at similar cooling effectiveness as substantial
difference exists in internal cooling rates that directly
influence the steady-state stack temperature. 120

100
C o o l i n g E ffe c ti v e n e s s , (% )

C. Analytical and Simulation Results Comparison 80

60
Internal and external forced convection calculations were
performed using equations (1) to (4) for the standard stack 40

design at all velocities and presented in Fig. 12. The total 20

cooling effect was lower than the corresponding simulated 0

cases, ranging from 40% difference at 1 m/s and reducing to 0.01 1.0 2.0 3.0
Air Velocity, (m/s)
3.5 fin 3.5

95
 Figure 11. Cooling effectiveness of the simulated designs

REFERENCES
5000
[1] New Energy Foundation, “Residential PEFC Demonstration Project”, at
4000
www.nef.or.jp, accessed on 30th March 2009.
Cooling rate (W)

[2] W.A.N.W. Mohamed., R. Atan., and T. S. Yiap, “Current and possible

3000 Total cooling future applications of hydrogen fuel cells in Malaysia”, Proc. of the Int.
Internal cooling Conf. on Advances in Mechanical Engineering (ICAME) June 2009,
2000 External cooling Malaysia.
[3] W.A.N.W. Mohamed, R. Atan and A.A. Razak, “Micro-channel
1000 optimization for an air-cooled PEMFC by CFD simulation”, Proc. of the
2nd Engineering Conference (ENCON) April 2010, Malaysia.
0 [4] J-W. Ahn and S-Y. Choe, “Coolant controls of a PEM fuel cell system”,
1 2 3 3.5 J. Pow. Sour, vol.179 pp. 252-264, 2008.
Nominal Air Delivery Velocity (m/s) [5] Y-J Sohn, G-G. Park, T-H. Yang, Y-G. Yoon, W-Y. Lee, S-D. Yim, and
C-S. Kim, “Operating characteristics of an air cooling PEM fuel cell for
Figure 12. Cooling effect on the standard design based on theoretical analysis portable applications”, J. Pow. Sour, vol. 145, pp 604-609, 2005.
[6] S. Yu and D. Jung, “Thermal management strategy for a PEMFC system
VI. CONCLUSION with a large active area”, Ren. En., vol. 33 (120), pp. 2540-2548, 2008.
PEM fuel cells generally generate as much thermal power as [7] www.ballard.com. Accessed on 28th January 2008
the corresponding electrical power. The heating effect leads to [8] F. P. Incropera and D. P. DeWitt, Fundamentals of heat and mass
transfer, Fifth Edition. John Wiley & Sons, 2002.
reduced cell efficiency and damage, requiring an effective
[9] S. Begot and J.M. Kauffmann, “Estimation of internal fuel cell
cooling method for maintaining the stack temperature at the temperatures from surface temperature measurements”, J. Pow. Sour.,
desired level. The study has successfully applied CFD to vol. 178, pp.316-322, 2009.
investigate the thermal characteristics of a PEM fuel cell with [10] L. Dumercy, R. Glises, H. Louahlia-Gualous, J.M. Kaufmann, “Thermal
a single cooling channel per plate under different air delivery management of a PEMFC stack by 3D nodal modeling”, J. Pow. Sour.,
vol. 156, pp. 78-84, 2006.
velocities. At the fixed channel geometry, inlet air velocities
[11] A. Faghri and Z. Guo, “Challenges and opportunities of thermal
higher than 2 m/s, translated to Reynolds flow number of 800, management issues related to fuel cell technology and modeling”, Int. J.
is fully capable to achieve 100% of the cooling load. A general Heat and Mass Trans., vol. 48, pp. 3891-3920, 2005.
relation of average stack temperature to the cooling [12] W.A.N.W. Mohamed and R. Atan, “Optimizing air cooling capability of
effectiveness was established. The integration of an extended a PEMFC through case-by-case cooling channel analysis”, Proc. of the
cooling surface on the top surface of the stack also offers Int. Conf. on Advances in Mechanical Engineering (ICAME) June 2009,
Malaysia.
enhanced thermal distribution uniformity within the stack at an
improvement of 30% reduced thermal gradient, which is very
desirable in practical conditions.

96