Experimental thermal analysis on air cooling for closed-cathode Polymer Electrolyte Membrane fuel cells

i n t e r n a t i o n a l j o u r n a l o f h y d r o g e n e n e r g y 4 0 ( 2 0 1 5 ) 1 0 6 0 5 e1 0 6 2 6

Available online at www.sciencedirect.com

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

Experimental thermal analysis on air cooling for closed-cathode Polymer Electrolyte Membrane fuel cells W.A.N.W. Mohamed*, R. Atan Alternative Energy Research Centre, Faculty of Mechanical Engineering, Universiti Teknologi MARA, 40450 Shah Alam, Malaysia

article info

abstract

Article history:

This work explores the strength and limits of using separate air cooling for closed-cathode

Received 1 April 2015

Polymer Electrolyte Membrane (PEM) stacks. Evaluating the thermal behavior of the de-

Received in revised form

signs based on stack temperature profiles alone would lead to inaccuracy as the initial

2 June 2015

temperatures and the stack thermal powers are different. Thus, the thermal behavior of

Accepted 19 June 2015

the cooling modes was qualitatively analyzed via heat transfer analyses. An experimental

Available online 15 July 2015

approach is reported here using three stacks with varied cooling channel geometry and aspect ratio. Two stacks were designed on parallel multi channel (20 and 40 channels)

Keywords:

straight flow configuration. The third stack applied the concept of non-linear laminar flow

PEM fuel cells

trajectory for the cooling channels. The 3-cell stacks were constructed with an active area

Cooling channels

of 240 cm2. The cooling mode applied a cooling fan coupling of positive and negative

Thermal engineering

pressure flows. Air flows were between Reynolds number of 200 and 400 while the hu-

Temperature uniformity

midity varied at 50% and 90%. The analytical methodology converted the first-order tem-

Air cooling

perature profiles into second-order heat transfer profiles. The steady-state parameters studied were temperature uniformity, cooling response, average cooling rate, cooling effectiveness, cooling flux, the heat transfer coefficient and the mean local temperature difference. The width of analysis has successfully identified the dynamic capabilities of the individual cooling plate designs for further practical considerations. Copyright © 2015, Hydrogen Energy Publications, LLC. Published by Elsevier Ltd. All rights reserved.

Introduction Hydrogen has been identified as a transition fuel in complementing the oil supply shortage with a prediction of a large scale introduction for hydrogen fuel cell application by 2020. The green aspects of hydrogen fuel cells, where only heat and water are produced as byproducts, places the technology as an

ideal replacement to conventional power conversion systems [1]. The Polymer Electrolyte Membrane Fuel Cell (PEMFC) is one of the fuel cell types that offer numerous advantages according to its application. It is very flexible with respect to power and capacity needs and proven capable of long service life, good ecological balance and very low self-discharges [2]. It is favored in many applications as it offers high power density, quick-startup and low operating temperatures as well as

* Corresponding author. E-mail address: [email protected] (W.A.N.W. Mohamed). http://dx.doi.org/10.1016/j.ijhydene.2015.06.095 0360-3199/Copyright © 2015, Hydrogen Energy Publications, LLC. Published by Elsevier Ltd. All rights reserved.

10606

i n t e r n a t i o n a l j o u r n a l o f h y d r o g e n e n e r g y 4 0 ( 2 0 1 5 ) 1 0 6 0 5 e1 0 6 2 6

Nomenclature cell active area, cm2 surface area of a single cooling channel, m2 a fuel cell parametric coefficient specific heat of carbon graphite, J/kg K concentration of oxygen at the gas/catalyst interface, mol/cm3 Ecell actual cell potential, volts the thermodynamic potential of the cell in an ENersnt open circuit, volts the thermoneutral voltage, volts Etn energy of water at the cathode exit, W Ew,exit DEair,cathode total energy change of the reactant air exit stream, W F Faraday's number, 96,485 Coulombs/mol free reaction enthalpy at 298 K, 237.3 kJ/mol DG
heat value of the hydrogen, kJ/kg Hfuel hf@Tw,exit saturated liquid enthalpy of water at cathode stream exit temperature, kJ/kg h effective heat transfer coefficient, W/m2
C I load current, ampere J actual current density, A/cm2 Jmax maximum current density, A/cm2 l thickness of the membrane, mm molar mass of oxygen, 32 g/mol MO2 molar mass of water, 18 g/mol Mw total mass of carbon graphite plates, kg mcg m_ air;exit rate of reactant air at the cathode exit, kg/s m_ air;inlet rate of reactant air at the cathode inlet, kg/s rate of oxygen consumed in the reaction, kg/s m_ O2 rate of water formation, kg/s m_ w n number of electrons number of cells in the stack ncell number of cooling channels in the stack nch number of moles of oxygen consumed in the nO2 reaction, mol number of moles of water formed in the reaction, nw mol electrical power, W Pel fan power, W Pfan A As,ch B Ccg cO_ 2

rapid response to varying operational loads [3]. Currently, a PEMFC with a net power density of one kW/L has been achieved [4] which is the result of breakthrough in all aspects of PEMFC engineering. Commercialization and interest of fuel cells have just aggressively started at the turn of the century, and apart from full-scale car prototype developments, PEM fuel cell stacks with power outputs less than 3 kW are much in demand. The popular applications include backup power systems and small-scale or demonstration vehicles, mainly conducted by research institutions and academia. From this initial culture, there is a potential market for small-sized PEMFC stacks with power output ratings of up to 3 kW. The main advantage air cooling systems holds over water-cooled systems is that it is more compact, increasing the overall system size by usually

PH2 PO2 Pth Q_ c;avg Q_ c;transient Q_ gen Q_ stack;avg DQ_ stack q Relectronic Rproton rm DS Tair,in Tair,exit Tx Tavg Ti Tiþ1 Tiþn Ts T
DTcathode Dt P t UT Vact Vcell Vconc Vohm Vrev ε hFC l

supply pressure of hydrogen into the stack, bar partial pressure of oxygen supply into the stack, bar stack generated thermal power, W average cooling rate, W transient stack cooling rate, W generated heat from the reaction, W averaged stack heat change rate, W change of stack heat content for the duration of the time step, W cooling flux, W/m2 resistance to electron flow in a fuel cell, 0.0003 U resistance to proton flow through the electrolyte, U specific resistivity of the membrane to electron flow reaction entropy at 298 K, 163.33 J/mol K coolant air temperature at stack inlet, K coolant air temperature at stack exit, K representative zonal surface temperature, K average cooling plate temperature, K initial temperature reading, K temperature reading at end of single time step, K temperature reading at end of operation stack operating temperature, K reference temperature of air at 298 K reactant air inlet-exit temperature difference, K time step, s total operation time, s temperature uniformity index activation over voltage, volts actual cell voltage, volts mass concentration over voltage, volts ohmic over voltage, volts reversible cell voltage, volts cooling effectiveness energy conversion efficiency, % parameter based on PEM fuel cell membrane humidity

less than 50%, whereas water-cooled systems generally increases the system size by more than 200%. Therefore, for portable and limited space applications, air-cooled fuel cells are very much desirable.

Thermal engineering of PEMFC Fuel cells primarily generate heat from the entropic heat of reactions, the irreversibilities of the electrochemical reactions, ohmic resistances and heat from the condensation of water vapors [5]. The sum of the entropic heat, irreversible reaction heat and ohmic heating is comparable to the power output of a PEM fuel cell. Roughly, they account for 55%, 35% and 10% of the total heat release, respectively [6]. The magnitude of thermal energy is associated with the

i n t e r n a t i o n a l j o u r n a l o f h y d r o g e n e n e r g y 4 0 ( 2 0 1 5 ) 1 0 6 0 5 e1 0 6 2 6

conversion efficiency of the cell which is a function of its polarization behavior. A fuel cell with 50% conversion efficiency generates equal electrical and thermal powers and the heat increases exponentially below this level; constraining the practical efficiency. In Ref. [7], it was shown that heat generation keeps on increasing in a PEMFC even when current density has stabilized which is a clear indicator of increasing irreversibility. The operating conditions for best performance require a balance between temperature, humidity, and reactant flow rates to avoid both flooding and dehydration [8]. The sensitivity of PEM fuel cell stacks to temperature is mainly related to the required moisture levels in the membrane which is kept hydrated from water back-diffusion flux from the cathode to the anode. When the operating current density increases, the effects of temperature on membrane hydration slightly decrease. In general, fuel cell performance improves with increased humidity until flooding conditions appear. Heat buildup within the cells would reduce the moisture content of the membrane and leads to greater electrical and thermal resistances [9]. However, heat is also needed for improved reaction kinetics at the catalyst layers. The effects of heat to the operation of a fuel cell are subjective and complex; heat is needed to improve the reaction kinetics, but too much heat would lead to an increase in energy losses. Therefore, thermal management of PEM fuel cells needs to delicately balance both requirements. Thermal engineering of a fuel cell is an operational requirement and individual studies are possible whenever the selected system domain is intentionally detached from the electrochemical reactions, such as the design of cooling systems and its control modes. In practice, application of active cooling systems is compulsory for fuel cells above the 100 W electrical power output. The most common thermal control mode is linear feedback control whereby the cooling system is initiated when the stack reaches a certain specified temperature and its cooling power adjusted accordingly by controlling the mass flow rate of the coolant in response to any temperature variations. Temperature distribution within the cells is a core issue in thermal engineering. Steady and uniform stack temperature is desirable in fuel cell operation as it promotes reaction homogeneity across the cells. However, internal stack temperature mapping shows the existence of large temperature gradients across the cells [10] as the heat generation magnitude within a cell is localized depending on the reaction kinetics and existing temperature gradients. Poor temperature distribution within a fuel cell eventually leads to a performance penalty. Yu and Jung [11] showed that the nonuniformity of the fuel cell temperature can cause uneven humidity distribution and result in more power loss due to higher electric resistance. As such, active sites and large cells require faster cooling rates to maintain near uniform cell temperatures. Wen and Huang [12] acknowledged the importance of identifying the hot spots as a vital step in formulating the right cooling mode towards effective cooling and obtaining uniform cell temperatures. Kang et al. [13] further showed that improved cell performance is also related to reactant flow field and cooling channel configurations.

10607

In a PEMFC operation, it is quite difficult to simultaneously obtain or control both the stack temperature and temperature uniformity. Practically, cooling system control is based only on the stack temperature which is obtained by the placement of temperature sensors or probes on a designated stack surface. For large stacks, temperature measurements are obtained by more than one sensor and the data are processed by the microcontrollers of the cooling system to identify the representative stack temperature. Thus, the location where the sensors are positioned should be feasible in providing an accurate representation of the stack internal temperatures. A PEMFC stack design is based on two configurations; the open cathode and the closed cathode design. The main difference is on the method of stack cooling operation. Open cathode designs directly take in ambient air into the stack to act as both reactant and coolant. The stack design and operation is more simple but the net power output is limited by the heat rejection rate while the overall performance and durability are restricted by high temperature gradients during stack operation [14]. In contrast, the closed cathode design utilizes separate flows for the reactant and coolant. The reactant air is usually supplied at pressurized conditions to overcome large pressure drops within the stack and the coolant used is usually liquid that runs in separate cooling plates between the cells or even externally. Closed cathode PEMFC stacks are designed for large power outputs and so appropriately require adequate active cooling assistance to maintain the cells at a suitable operating temperature as the heat released are much higher than open cathode stacks. In closed cathode design, the cooling effect is achieved by a separate coolant flow from the reactant air. Water is the conventional coolant type where recent trends in fuel cell thermal engineering have shifted to water mixtures with ethylene glycol or propylene glycol solutions [15]. For a similar mass flow rate and temperature difference, the heat capacity of water would be four times more than air. Faghri and Guo [16] have reviewed this challenge and concluded that air cooling would still be viable for closed cathode stacks up till the range of 10 kW power. The main advantage would be on the reduction of overall system size (less than 50% expansion from base fuel cell system) compared to the implementation of liquid cooling system (system size increases by more than 200%). This feature would lead to a more compact and lightweight system with higher applicability in small or constraint enclosures. Reports in applying air cooling for closed cathode stacks are not as extensive as water-cooled systems, such as experimental validations of 3D stack thermal modeling by Adzakpa et al. [17] and Matian et al. [18], cooling enhancement by applying heat spreaders [12,19] and detail stack thermal behavior analysis [20]. Other notable works include the performance analysis of a portable 500 W stack [21] and numerical study on internal forced convection [22]. Matian et al. [23] provided depth to the study of closed cathode air-cooled PEMFC via CFD modeling and experimental approach. Their focus was on the effects of different cooling channels designs on the cooling performance and temperature distribution across the cell. The cooling channels were straight path rectangular geometry with different aspect ratios, but designed with a similar nominal height of 3.75 mm. In terms of

10608

i n t e r n a t i o n a l j o u r n a l o f h y d r o g e n e n e r g y 4 0 ( 2 0 1 5 ) 1 0 6 0 5 e1 0 6 2 6

temperature distribution, the authors concluded that temperature variations are inevitable, but more uniformity was achieved by increasing the aspect ratio of the cooling channels. The geometrical aspect of cooling channels is an important factor in improving the cooling performance of any cooling domain. However, cooling rates, maximum temperature gradient and temperature distribution does not always relate directly. The cooling plate designs for PEMFC stack are usually serpentine for liquid coolants and straight path for air coolants. A more detailed and useful approach in determining the temperature uniformity index was provided by Baek et al. [24] where various cooling channel designs were simulated. It was discovered that the parallel multi-pass serpentine design gives 60% better temperature uniformity as compared to conventional multi-pass serpentine designs. Non-conventional mini channel designs that have been proposed for PEMFC were based on non-linear flow (also termed as chaotic flow) trajectory to promote a greater interaction between the fluids in motion with the channel surface. So far, the thermal performance of chaotic flow designs such as the C-shape and square-wave mixer was numerically studied by Lasbet et al. [25] for water-cooled PEMFC with results that give positive heat transfer enhancements. However, that study has seen limited expansion of use in an actual fuel cell design. In air-cooled PEMFC systems, ambient air is applied freely as the cooling medium; thus, the cooling environment is highly influenced by the ambient conditions of temperature and humidity. High inlet air temperature would reduce the cooling force, while high humidity assists in improving cooling rates with minimal effect on air temperature increase as it travels along a heated environment. Hence, the cooling effect would be more uniform from the inlet to the outlet due to a more consistent temperature difference with the hot surface. The momentum source is normally positioned at the cooling channel exit position and a negative pressure gradient was applied to pull in the air into the cooling plates. This configuration promotes a highly uniform flow stream into the mini channels as compared to forcing the air through from upstream that will encounter higher inlet resistance [26]. However, the upstream fan configuration would promote edge cooling across the frontal stack surface, yielding a slightly improved 5% cooling effectiveness compared to the downstream configuration at a similar heat load with a penalty of higher temperature gradient and slower cooling response [27]. The effect on stack-fan configurations of blow and suction to the temperature and water distribution as well as towards stack performance of an open cathode fuel cell was reported by Sasmito et al. [28,29]. As an important note, the effect of a combined upstream and downstream fan configuration for a closed cathode design has not been reported in mainstream literature. Very recently, Franco [30] conforms the feasibility of separate air cooling for PEM fuel cells by the approach of optimizing the design of the cooling channels. The optimization procedure was analyzed for a fuel cell with a thermal load of 180 W. The stack thermal model at the cooling channel boundary was based on a finned array model with rectangular fins and insulated tips. The work established specific bipolar

plate designs for separate air cooling that may reduce the parasitic load significantly. However, one of the key parameters is to use highly conductive plate material such as polymer graphite (k ¼ 52 W/m K). Transient thermal behavior of fuel cells is a complex science to model due to the ever-changing boundary conditions of the system such as the electrode surface temperatures. In actual fuel cell operation, dynamic local reactions are common due to factors such as localized mass concentration, membrane humidity changes and perturbations in reactant flow. In a solid body with heat generation and internal cooling, temperature changes will continue to occur exponentially until an asymptotic thermal saturation level, or a steady-state temperature distribution is reached [31]. This saturation level is a result of conductive and convective rate equilibrium which is reached within a specific time period and dependent on the specific temperature gradient within the solid, the convective cooling coefficient development and cooling surface geometries. The common method in transient conduction analysis of a system under cooling is the lumped capacitance method. However, its accuracy depends on the magnitude of temperature gradients within the solid body. The normal approach in thermal analysis methodology of experimental PEM fuel cells is as reported in Ref. [23]; stack temperatures are directly interpreted in various perspectives to determine the impact of the cooling mode. The usual perspectives are average temperature, temperature gradients and maximum temperatures. The maximum temperature can be more conveniently measured and monitored than average temperature or mid-point temperature and also an important monitor for the durability of the fuel cell [11]. In addition, the PEMFC temperature control is usually a supplement for algorithms aimed at the PEMFC electrical power control [32]. Conversion of the physical temperatures into actual stack cooling profiles is not as extensive. In experimental analysis involving a stack with a number of cells, the determination of exact convective cooling properties is difficult as the system involves a set of cooling channels. Variations in cooling air parameters such as inlet velocity, flow rate, and even temperature is expected in the individual cooling channels due to the fan-stack configuration, ambient conditions and local cell reaction sites. Dynamic reaction sites are also expected across the cells; thus, the application of the lumped capacitance method in this study is also too complex to achieve. Therefore, empirical solid-state analysis that tracks and applies the magnitude of stack energy changes was taken as the base analytical methodology for cooling performance evaluation. The work of other researchers in this field was referred to develop and formulate the experiment as well as analytical approach presented in this paper. The bulk (or average) stack temperature profiles at different load currents and dynamic temperature to electrical power variation were identified as useful tools in modeling and experimentally validating steady-state and dynamic heat transfers in a stack [33]. A simplified model to track the stack energy changes was then developed from the lumped thermal mass model by Zhang et al. [34] applying the transient stack temperature profiles based on solid-state thermal analysis. The methodology for internal stack temperature measurements and cooling response analysis follows the method of Adzakpa et al. [17].

i n t e r n a t i o n a l j o u r n a l o f h y d r o g e n e n e r g y 4 0 ( 2 0 1 5 ) 1 0 6 0 5 e1 0 6 2 6

The cooling channel geometry and coolant flow control settings are two engineering aspects capable of further improving the cooling performance of an air-cooled PEMFC. The operational relationship between a certain design of mini channel geometry to differences in fan operation e for example on the application of simultaneous upstream and downstream fans as well as manipulation of air properties in terms of humidity e is explored in this publication. Three cooling plate designs were proposed for closed cathode application; two designs were based on straight flow configuration and the third design was based on non-linear flow trajectory. The thermal results are interpreted mainly from the perspective of solid-state thermal analysis in order to obtain unique profiles characterizing the heating and cooling effects of the stack during operation. The actual stack cooling profile provides a tool to obtain the effective heat transfer coefficient within the cooling plates.

10609

reference temperature of air (fixed at standard reference state of 298 K). PH2 is the supply pressure of the hydrogen reactant into the stack and PO2 is the partial pressure of oxygen supply. Since PEM fuel cells normally consume surrounding air to supply the oxygen, the partial pressure of oxygen is 0.21 atm for an intake at 1 atm. If the air supply was compressed, then the partial pressure of oxygen increases linearly to the total air pressure. Thus, from Eq. (4), a simplified Nernst equation is obtained as        ENernst ¼ 1:229  8:5 104 ðTs  298Þ þ 4:308 105 $Ts $ ln PH2    1 ln PO2 þ 2 (5)

Cell energy losses The main energy loss is contributed by the activation over voltage which is the energy loss due to the activation of electrochemical reactions at the anode and cathode.

Electrical and thermal models of a PEM fuel cell The electrical performance of a PEM fuel cell dictates the generated thermal energy within the stack. The theoretical power curve of a fuel cell can be obtained by establishing electrochemical models based on the Nernst equation and subsequent voltage losses within the stack. Higher voltage losses at a specified current density lead to a higher heat generation. The overall reaction of a PEM fuel cell is

H2 þ ½ O2 / H2O þ electrical energy þ heat energy The electrical energy (Pel) is the desired system output and the stack heat generated, or thermal power (Pth), of a single cell are linked through the actual output cell voltage (Ecell), Pel ¼ ðEcell ÞI

(1)

Pth ¼ ðENernst  Ecell ÞI

(2)

Ecell ¼ ENernst  Vact  Vohmic  Vconc

(3)

where ENernst is the maximum achievable (reversible) voltage of a fuel cell due to thermodynamic potential limits, I is the current output while the activation (Vact), ohmic (Vohm) and mass concentration (Vconc) over voltages are cell energy losses. The Nernst equation refers to the thermodynamic potential which is the potential of the cell obtained in an open circuit thermodynamic balance. The Nernst equation for the hydrogen/oxygen fuel cell is ENernst ¼

    1   DG
DS R$Ts ðTs  T
Þ þ  ln PH2 þ ln PO2 nF 2 nF nF

(4)

DG
is the free reaction enthalpy at 298 K which is 237.3 kJ/mol and DS is the reaction entropy at 298 K which is 163.33 J/mol K [33]. The value of n is the number of moles of electron transferred in the fuel cell's reaction (2 electrons) while F stands for Faradays constant (96,485 Coulombs/mol). Ts is the stack operating temperature which is usually in the range of 50
Ce100
C for low temperature fuel cells. T
is the

  Vact ¼ x1 þ x2 $Ts þ x3 $Ts $ln cO_ 2 þ x4 Ts $lnðIÞ

(6)

where x1,2,3 and 4 are parametric coefficients based on electrochemical, kinetics and thermodynamic laws. The values of this coefficient [35] are x1 ¼ 0.9477, x2 ¼ 0.0033, x3 ¼ 7.5  105 and x4 ¼ 1.915  104. The load current (I) is the operating current of the stack or applied load. The term cO_ 2 is the concentration of oxygen at the gas/catalyst interface (in mol/cm3) and can be expressed as cO_ 2 ¼

PO2 498 Ts

5:08ð106 Þ$e

(7)

The second type of energy loss is Ohmic loss contributed by the resistance to charge flow within the cell and can be represented using Ohm's law.   Vohmic ¼ I Relectronic þ Rproton

(8)

where Relectronic is equal to 0.0003 U (constant parameter). It stands for resistance associated with the fuel cell that resists electron flow. Rproton is the resistance associated with resistance to proton flow through the electrolyte (membrane) and can be calculated by Rproton ¼

rm $l A

(9)

where l is the thickness of the membrane. Then, the value of rm, or specific resistivity of the membrane to electron flow, can be determined from    2  2:5  Ts I 181:6 1 þ 0:03 AI þ 0:062 303 A     rm ¼  l  0:634  3 AI $exp 4:18 Ts 303 Ts

(10)

where A is the active area of the PEM cell in cm2. The parameter l is a parameter based on a value from 14 under ideal condition of 100% PEM humidity to the value of 23 under oversaturated conditions.

10610

i n t e r n a t i o n a l j o u r n a l o f h y d r o g e n e n e r g y 4 0 ( 2 0 1 5 ) 1 0 6 0 5 e1 0 6 2 6

The concentration over voltage can be calculated by Vconc

 J ¼ B$ln 1  Jmax

generated thermal powers to the stack heat gain at a local time period. (11)

where B is a parametric coefficient and normally obtained from the technical data of the fuel cell. The parameter J is the actual current density which is the ratio of operating current against active area of the cell. Jmax is the maximum current density or the ratio of maximum current versus active area of the cell.

Heat generation and thermal analysis models The heat generated within fuel cells is assumed to be the heat generated mainly at the electrochemical reaction sites of the cathodes. The amount of heat generated can be estimated using the simplified relations based on the energy balance of the system and depending on the state of water formed. I Hfuel $ncell ¼ I$Ecell $ncell þ Qgen n$F

(12)

where Hfuel is the heat value of the hydrogen, F is the Faraday's number, ncell is the number of cells, n is the number of electrons in the reaction, I is the total generated current, Ecell is the individual cell voltage and Qgen is the generated heat from the fuel cell stack. If the water exists as vapor at room temperature, then the thermoneutral voltage Etn is 1.254 V and the stack thermal power Pth,stack are dependent on the current produced and cell voltage. Q_ gen ¼ Pth ¼ ð1:254  Ecell ÞI$ncell

ðWattsÞ

(13)

The use of overall energy conversion efficiency can be applied to determine the cut-off voltage operating point that leads to excessive stack heating. The energy conversion efficiency represents the degree of actual cell potential departure from the idealized thermoneutral voltage, Etn, and is defined as hFC ¼

E Etn

(14)

In the experimental, the fuel cell temperature is continuously monitored in short periods, or time steps Dt. The transient temperature plots can be converted into transient stack heat changes (DQ_ stack ) based on the temperature profile of the stack during each time step. mcg Ccg ðTiþ1  Ti Þ DQ_ stack ¼ Dt

ðWattsÞ

(15)

where mcg is the total mass of carbon graphite plates, Ccg is the specific heat of carbon graphite, and (Tiþ1  Ti) is the stack temperature change for a single time step. The transient heat transfer profile is an indicator for the rate of thermal energy change within the measured periodicals; a positive value denotes stack heating effect while negative values mean a cooling effect occurrence. From the analysis of the experimental data, the transient plots were shown to behave in an exponential trend. If the stack thermal power generated is assumed constant, then the actual transient stack cooling profile is the difference between the

Q_ c;transient ¼ Pth  DQ_ stack

ðWattsÞ

(16)

For an objective parametric cooling analysis, the transient stack cooling profile is not suitable, hence requiring an arithmetically averaged heat transfer value over the total experimental time St. The averaged stack cooling rate Q_ c;avg is a representative cooling rate value for direct performance evaluation purposes such as the determination of the cooling effectiveness. mcg Ccg ðTiþn  Ti Þ Q_ c;avg ¼ St

ðWattsÞ

(17)

Evaluation of the cooling effectiveness is performed by direct comparison between the actual cooling rates of the stacks to the thermal power generated during operation. The cooling effectiveness is relative to the simultaneous conditions of generated thermal power of the stack at a particular load and the rate of stack heating or energy gain; a higher heating effect does not directly translate to lower cooling effectiveness. ε¼

Q_ c;avg Pth

(18)

The temperature uniformity index is applied in this work to evaluate the capability of the cooling modes and channel design in promoting temperature uniformity within the cells relative to the average temperature of the cell. It quantitatively measures the deviation of the surface temperature, T, from the average temperature, Tavg, at the heat transfer surface [24]. A low index is highly desirable where the temperature distribution is perfectly uniform when the index is zero. To apply it practically, the plate is divided into sub-sections and a representative temperature of the sections (or zones) is acquired. Z UT ¼



Tx  Tavg dA

A

Z

ð
CÞ

(19)

dA A

Another analytical parameter is the cooling flux, q, which is the ratio between the average cooling rates to the total internal convective surface area of the cooling plate design. The analysis allows a neutral perspective to the overall cooling performance as the cooling channel designs varies in geometry and trajectory, leading to differences in fluidewall interface efficiency. Q_ c;avg q¼P As



W=m2



(20)

In thermal engineering analysis involving a surface and a cooling fluid, the convection effects are represented by the heat transfer coefficient. In this case, the internal convection coefficient is calculated by the bulk cooling effect of the stack rather than in individual channels as the flow distribution and local surface temperatures are different in each channel. It is assumed that bulk cooling rate is equally distributed in the individual cells. Thus, the effective heat transfer coefficient, h,

i n t e r n a t i o n a l j o u r n a l o f h y d r o g e n e n e r g y 4 0 ( 2 0 1 5 ) 1 0 6 0 5 e1 0 6 2 6

which is used to define and represent the internal convection effects in each channel, is defined from the average stack cooling rate. Q_ c;avg ¼ h$

∴h ¼

X

As $DTLM

ðWattsÞ

Q_ c;avg 9 8 = < nch $As;ch $ DT1 DT 2 : DT1 ; ln



(21)

 W=m2 $
C

(22)

DT2

where DT1 ¼ Tair,in  Tstack,avg, DT1 ¼ Tair,exit  Tstack,avg, nch is the number of cooling channels in the stack and As,ch is the surface area of a single cooling channel. A fuel cell stack may dissipate its heat energy by internal as well as external mechanisms. Internal heat removal by the cathode fluid stream is more significant than the anode fluid stream as the exothermic reactions occur at the cathode and the water produced absorbs the generated heat. To evaluate the cooling contributions by the cathode fluid stream, the heat content of the reactant air and liquid water needs to be analyzed. The amount of oxygen consumed in the reaction is proportional to the stack current output,  m_ O2 ¼ MO2

I



nO2 F

ðkg=sÞ

(23)

Then, the exit airstream flow rate can be approximated from the inlet flow rate and the rate of consumption, m_ air;exit ¼ m_ air;inlet  m_ O2

ðkg=sÞ

(24)

The total energy change of the reactant air exit stream is based on its temperature difference, DEair ¼ m_ air;exit $Cp $DT

ðWattsÞ

(25)


where the Cp of air is 1007 J/kg C. The rate of water produced from the generation of a specific current is  m_ w ¼ Mw

I



nw F

ðkg=sÞ

(26)

The energy of the water at the exit, based on the liquid water phase is then Ew;exit ¼ m_ w hf @T;exit

ðWattsÞ

(27)

Stack and cooling plate design The concept of the stack designs was based on the closedcathode configuration. Fig. 1 illustrates the main parts and sub-components of a PEM fuel cell stack system that was applied in the stack development. Ideally, a single-cell assembly would provide the most reliable accuracy on temperature distribution and cooling performance as multi-cell configuration would be highly influenced by the thermal conditions of adjacent cells. However, it is practically impossible to integrate a single-cell stack with the cooling channel designs due to the absence of bipolar plates and then to configure it with a suitable cooling fan. A three-cell assembly

10611

provided the minimal spacing for an effective stack-fan configuration with space for a maximum of two cooling plates. Generally, a three-cell assembly consists of three units of Membrane Electrode Assembly (MEA), two bipolar plates, two mono-polar plates, two current collector plates, and two endplates (refer Fig. 2). The cooling channels are integrated within the bipolar plates; therefore, each stack comprises of two sets of cooling channel arrays. The bipolar plates (flow field plates) are the backbone of a PEMFC system and are made of carbon graphite. The flow field designs for both anode and cathode are 7-pass serpentine and 3-pass serpentine respectively. The channel width is 1 mm and depth 0.5 mm. The bipolar plate designs are based on the closed cathode concept and suitable for stacks higher than 2 kW power requirements. The size of the plate was considered to be very suitable in thermal engineering analysis as it provides: i. a cooling channel run-in length higher than 100 mm, allowing greater cooling interface and thermal gradient development, and ii. a plate size over 200 cm2, providing a wide range of locations for zonal temperature monitoring. The MEA is sandwiched between the bipolar plates, while mono-polar plates are used at the stack ends. The physical characteristics of the MEA are provided in Table 1. The effective reaction sites are relative to the total reactant flow in direct contact with the MEA and the active area is 240 cm2. Plate surfaces in contact with the MEA are needed as a pathway for electron charge transport. These opposing technical requirements, together with reactant flow properties within the channels, need optimal balancing and form one of the greatest challenges in bipolar plate engineering. Establishment of the cooling channel designs was the first step towards the overall development of the PEMFC stacks. For the purpose of accurately evaluating the thermal performance of the cooling channels, the three stacks were developed with a similar concept, similar controlled methodology and of equal parts. Table 2 lists the geometrical specifications of the selected cooling channel designs and Table 3 provides the stack parameters. As the stacks are based on the closed cathode configuration, the cooling channels were integrated within the bipolar plates. Three stacks were developed with rectangular cooling channel design with similar height of 2 mm but with differences in the aspect ratio, flow path and total number of channels. Stack label Ms40 consists of 40 straight path channels, stack label Ms20 has 20 straight path channels, while stack label Ch20 consists of 20 chaotic flow channels (see Fig. 2 for cooling plate design concepts). Design Ms40 has the smallest single channel hydraulic diameter (2.53 mm) and the smallest aspect ratio (1.73) but with the largest surface area (0.0593 m2) contributed by a higher number of channels. Design Ms20 gives the largest hydraulic diameter and aspect ratio at 3.18 mm and 3.9 respectively. The hydraulic diameter and aspect ratio of design Ch20 is only slightly higher than stack Ms40 with significantly reduced surface area (0.0355 m2). Even with the same number of channels to stack Ms20, the hydraulic

10612

i n t e r n a t i o n a l j o u r n a l o f h y d r o g e n e n e r g y 4 0 ( 2 0 1 5 ) 1 0 6 0 5 e1 0 6 2 6

Fig. 1 e Major components in the developed PEMFC stacks and operating system.

Fig. 2 e The cooling channel design concepts developed into 3-cell stack assemblies.

Table 1 e Properties of the MEA. Property Active area (size) Number of layers Catalyst

Value

Notes

240 cm2 5 0.3 mg/cm2 Pt loading 15% PTFE

e GDL-electrode-membrane-electrode-GDL Manufacturer specification

10613

i n t e r n a t i o n a l j o u r n a l o f h y d r o g e n e n e r g y 4 0 ( 2 0 1 5 ) 1 0 6 0 5 e1 0 6 2 6

Table 2 e Specification of cooling channel designs. Design label

No. of channels per plate, nch

Width (mm)

Height (mm)

Effective length (mm)

Aspect ratio

Hydraulic diameter, Dh (mm)

Total effective surface area, SAs (m2)

40 20 20

3.48 7.9 4.0

2.0 2.0 2.0

149.5 149.5 160.0

1.7 3.8 2

2.53 3.18 2.67

0.119 0.107 0.071

Ms40 Ms20 Ch20

stack; hence, an averaged stack temperature needs to be presented from a set of temperature readings across the stack.

Table 3 e PEMFC stack parameters.

1. 2. 3. 4. 5. 6.

Properties/parameters

Values

Material Specific heat, C Density Top/bottom surface areas (total) Side surface areas (total) Thermal conductivity

Carbon graphite 710 J/kg K 2240 kg/m3 57.12 cm2 1018.5 cm2 20 W/m K

diameter and aspect ratio of design Ch20 (2.67 mm and 2 respectively) is significantly lower due to its small channel width to accommodate the bending patterns of the design. Further details of the design, fabrication, assembly and conditioning methodology can be found in Ref. [36]. Fig. 3 displays the parts assembly which is similar for all stacks.

Experiment The experiment was developed towards obtaining the heating and cooling profiles of the stacks which are directly dependant on the stack temperature profile. The stack heat generation rate is linked to its overall polarization behavior; therefore, the stack voltage has to be closely monitored across all current loadings and cooling settings where the overall power curve of the stack is statistically determined and standard deviations of the power curve is analyzed. The stack heating and cooling rates were obtained by the changes in the enthalpy of the

Experimental apparatus Temperature probes are positioned within the cooling channels to obtain a direct surface temperature response to the cooling activity. The positioning of the probes was selected to cover the center region of the cooling plate and also to cover both cooling fluid ends. Three stack zones relative to the hydrogen inlet and outlet ports e top zone (near the hydrogen inlet), middle zone and bottom zone (near the hydrogen exit) e was designated as illustrated in Fig. 4. In total, eight thermocouple wires were used and inserted through the cooling channels at various depths according to zones (25 mm deep for end zones and 50 mm deep for center zones). Using thermal paste, the tip of the thermocouples was ensured to stick to the channel surfaces so as to obtain the measurements of the internal plate temperatures rather than the cooling air temperatures. It should be noted that the insertion of wire probes directly into the cooling air flow path would reduce the effective inlet by 30% (size relativity comparison). The inlet resistance is higher and the air flows within the channels would bend around the wire probes. The exact effect on the local cooling interactions are not investigated within this work, but it is assumed that the channel temperatures are highly influenced by the normal cooling conditions of adjacent cells due to its close proximity. Therefore, this work acknowledges the data acquisition penalty imposed by this measurement method and assumed that the penalty is small compared to the

Fig. 3 e Parts assembly of the 3-cell PEMFC stack.

10614

i n t e r n a t i o n a l j o u r n a l o f h y d r o g e n e n e r g y 4 0 ( 2 0 1 5 ) 1 0 6 0 5 e1 0 6 2 6

Fig. 4 e The configuration of the thermocouple sensors relative to hydrogen inlet-exit points and coolant flow direction: red lines and blue lines indicates the probes are inserted at the first and second bipolar plates respectively; for probes number 1, 3, 4, 6, 7 and 8 the insertion depth is 25 mm while for probes 2 and 5 the depth is 50 mm for mid-plane measurement. (For interpretation of the references to color in this figure legend, the reader is referred to the web version of this article.)

number of channels of the fuel cells. The limited number of thermocouples used was also based on these considerations. Fans are mounted together on a plate that covers the entire cooling channel array, forming a fan array. In this study, two arrays consisting of 2 fans upstream and 4 fans downstream were applied. The input power at maximum speed of 6000 rpm is 4.6 W (refer Table 4 for detail fan specifications). At maximum fan power, the measured air velocity was 2.2 m/s. Generally, the selection of a proper fan model in fuel cell application should consider the induced pressure of the airstream and the power consumption. Commercial open-cathode stacks normally apply a single negative pressure fan configuration. In this work, the fans are positioned vertically that provides horizontal air flow into the stack. The fan operation is the dual-fan setting with both positive and negative pressures applied simultaneously upstream and downstream of the stack. The dual-fan setting assists the flow of airstream by using an upstream fan to blow air towards the stack at close proximity while another downstream fan applies negative pressure to pull in (suction flow) air into the stack.

Operation A fuel cell test station was used in the experiment. It comes complete with a real-time data acquisition system and software for monitoring and storage of the polarization behavior.

The maximum current load applicable is 1.6 kW based on the single unit of electrical loader installed. However, the test station offers limited temperature monitoring capability; therefore, a separate temperature data logger with 10 inputs was used. In Fig. 5, the schematic of the test system is illustrated. Pre-tests indicate that the maximum current applicable before the stack operates in the mass concentration region was 40 A. The stoichiometric ratio of the reactants changes with the applied load current. In order to limit the effects of reactant-assisted cooling, parameter control was exercised by applying non-humidified reactants. The fuel cell operation cycle initialization is initiated by pressing the ‘start’ button that signals the initiation of hydrogen and reactant air flow into the fuel cell stack and the monitor will display real-time voltage, current density and power density profiles. Prior to this, the hydrogen and reactant air pressures should be manually set to the proper levels. A free run-in operation was first applied to allow the fuel cell to reach species equilibrium which can be determined by monitoring the voltage profile. A constant open voltage at this stage indicates that the stack has reached its optimal electrical potential. The open voltage under equilibrium represents the total existing potential difference for the whole stack, accounting for the species concentration throughout the flow fields and the inherent conductivity of the whole assembly. For a particular stack, the open voltage should be within a 5% difference for each experimental run. Values

Table 4 e Cooling fan specifications. Model type EC-6025HH12C

Input power (W)

Voltage (VDC)

Max current (A)

Fan speed (rpm)

Flow at max fan speed (CFM)

Pressure (mmH2O)

Noise (dBA)

4.56

12

0.38

6000

31.81

0.37

<39

i n t e r n a t i o n a l j o u r n a l o f h y d r o g e n e n e r g y 4 0 ( 2 0 1 5 ) 1 0 6 0 5 e1 0 6 2 6

10615

Fig. 5 e The schematic of the 1.6 kW load fuel cell test station.

beyond this tolerance means the stack is having difficulty to fully accommodate the reactant gases within the flow channels; a direct indicator of liquid water entrapment that blocks the passageways from previous operation or humidifying activity. A deterioration of open voltage levels during free run-in is an indicator of gas leakage. As the open voltage stabilizes, the cooling fans were allowed to start by turning on the power regulator. The current load is manually set and introduced when the stack temperature reaches equilibrium with the reactants. Upon initiation, the electrical loader completes the external stack circuitry and acts as a regulated resistance to draw current from the stack. Drops in stack voltage levels are expected where the voltage is expected to stabilize as a new level of species equilibrium is reached within the active area. The power and current density graphs should concurrently display positive increments towards a constant value. The test station automatically updates the voltage values over a standard of 5 s interval. Temperature monitoring of the ten thermocouples was performed by the temperature data logger, also manually set at 5 s intervals. It is important to note that the data logger should be activated and a save file is designated at least 30 s prior to load initialization. This is essential in recording the actual temperature jump of the stack as the load is turned on. Hydrogen purging cycle was set at 30s intervals. Each individual test was performed at approximately 10 min of active loading and cooling as the stack thermal

equilibrium was normally reached within that period. The ‘load on’ button is marked again to disconnect the external circuitry. Upon load off, the stack is allowed to cool with active cooling assistance for 5e10 min. The test cycles were based on the stack design, load setting and fan setting orders. The current loading order is strictly based on the lowest to the highest current value. The effect of air humidity changes to the cooling capability of both channel designs was investigated by using ambient dry air and humidified air as specified in Table 5. The fan setting order was first set towards the dual-fan setting using dry air and then the dual-fan setting using humidified air. For the humidified cycle, the cooling air was externally humidified using a 500 W humidifier unit with de-ionized water. The exit hose was located at approximately 20 mm distance to the blower fan intake plane and a hygrometer was used to identify the relative humidity of the airevapor mixture before entering the stack. Table 6 lists the properties of the reactants where the hydrogen supply comes from a centralized supply line with

Table 5 e Cooling airstream properties. Properties/parameters Inlet temperature (ambient) Relative humidity Dry (ambient) air Humidified air

Values 26
Ce29
C 50% RH 90% RH

10616

i n t e r n a t i o n a l j o u r n a l o f h y d r o g e n e n e r g y 4 0 ( 2 0 1 5 ) 1 0 6 0 5 e1 0 6 2 6

Table 6 e Hydrogen and reactant air supply properties. Properties/parameters

Hydrogen

Reactant air

Inlet pressure (gauge) Back pressure difference Inlet temperature Stoichiometry/flow rate - at 20 A load - at 30 A load

0.6 bare0.7 bar 0.2 bare0.4 bar 32
Ce35
C

0.7 bare0.8 bar 0.2 bare0.4 bar 35
Ce38
C

8.7/1.3 lpm 6.7/1.5 lpm

13.2/0.24 lpm 8.8/0.24 lpm

99.99% purity while the reactant air is supplied by an oil-free compressor.

Uncertainty analysis The uncertainty of the various parameters in the results analysis was determined based on the uncertainty in the measurements of all related independent variables. The main measurement in the experimental and the accuracy of the instrument used is listed in Table 7. The method used in the uncertainty analysis is based on Holman [37] where the uncertainty of a parameter is contributed by the precision of individual measured conditions within the parameter. An example is shown here for the analysis of the generated stack thermal power. Data from the experimental of stack Ch20 at 30 A load and 50% relative humidity is used where the parametric values are I ¼ 30 A, wI/I ¼ 1%, Vcell ¼ 0.6 V, wV/V ¼ 0.5% and ncell ¼ 3. Geometrical and internal properties such as surface area (As), mass of stack (mcg) and specific heat of the plate (Ccg) can be considered as exact values. Table 8 lists the uncertainty analysis of 7 main parameters. Overall, the confidence level of the experimental data is approximately 98%. From Eq. (13), Pth ¼ ð1:254  Vcell ÞI$ncell wPth ¼ Pth

Table 8 e Uncertainties in the main analytical parameters. Parameter Pel Pth Q_ c;avg ε UT q Ew,exit

Error ±0.3% ±0.8% ±1.1% ±0.01% ±1
C ±1.1% ±0.02%

Results and discussion In fuel cell performance analysis, the first level of evaluation is the stack polarization curve which is the plotting of cell potential changes as the applied load (current density) is increased under similar operating conditions. Fig. 6 shows that all the developed stacks register an open voltage of approximately 3.02 V, but as the load is increased, there is a small variation in power density. Stack Ms20 and Ms40 show a slightly higher power density (260 mW/cm2) at the highest applied load as compared to stack Ch20 (240 mW/cm2). The consistency of stack performance which relates to its design and development quality can be proven by performing a standard deviation analysis on the stack voltage of individual stacks as the load is increased as shown in Fig. 7. The voltage difference occurs prominently only at high loads with a

ðWattsÞ

sffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi  2  2ffi vPth vPth wI wV þ vVcell vI

(28)

where vPth/vVcell ¼ I$ncell and vPth/vI ¼ (1.254  Vcell)ncell wPth ¼ Pth

qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi ½ð90Þð0:009Þ2 þ ½ð1:962Þð0:03Þ2 ¼ 0:8%

Fig. 6 e Polarization curves for all stacks from 0 A to 30 A load current.

Table 7 e Parameters and accuracy of test instruments. Instrument

Model specifications

Thermocouple wires Anemometer

K-type Lutron AM-4206M

Pressure gauge Digital timer Digital flow meter DC Voltage sensor DC Current Transducer

Wika PG111.16PM e Bronkhorst F-1108C Phidgets Phidgets

Precision ±1
C ±0.8
C (temperature) ±2% (velocity) ±3% ±0.5% ±0.5% ±0.5% ±1%

Fig. 7 e Standard deviation of voltage measurement for all stacks at different loads.

10617

i n t e r n a t i o n a l j o u r n a l o f h y d r o g e n e n e r g y 4 0 ( 2 0 1 5 ) 1 0 6 0 5 e1 0 6 2 6

Fig. 8 e The energy conversion efficiency and bulk thermal power generated from the stacks.

Fig. 9 e Theoretical stack output at Tstack ¼ 40
C, PH2 ¼ 0.4 bar and Pair ¼ 0.4 bar.

maximum value of 5.5% by stack Ms40 at 30 A. This analysis proves the consistent quality in developing the stacks and the reliability of the polarization curve values. In Fig. 8, the thermal power generated by each stack was plotted using Eq. (13) while the conversion efficiency was based on Eq. (14). Equal electrical and thermal power outputs corresponding to a conversion efficiency of 50% would occur at 20 A load for stacks Ms40 and Ch20 while stack Ms20 was capable of achieving 50% efficiency at a higher load of 29 A. From that point onwards, the thermal power increases exponentially as the efficiency falls below 50%. This condition is one of the main factors limiting a fuel cell operation at 40% conversion efficiency as the stack would generate too much heat and reach a thermally critical and damaging state.

The theoretical performance of the 3-cell stack design is shown in Fig. 9 based on Eq. (1) to Eq. (13). The parameters used are listed in Table 9. Comparing to the electrical power of each stack obtained experimentally (Fig. 6), the actual outputs highly conforms to the theoretical outputs of the three stacks. A minor difference in the electrical load at 50% efficiency was obtained where the theoretical calculation predicts 50% efficiency to occur at 120 mA/cm2 (equal to approximately 30 A load) whereas the actual occurrence is approximately at 25 Ae27 A loads. Fig. 10 provides an example of the local stack temperature changes when electrical load is applied. The profile is based on stack Ch20 at 20 A load and 50% relative humidity. Distinct and localized temperatures were obtained at the various measured zones that relates directly to the reaction activity, current distribution and heat removal rates within the stack [38]. For the stack Ch20, the bottom zones give the highest temperature, followed closely by the hydrogen inlet region of the middle zone. These three zones show a very rapid temperature rise during the first 100s, strongly indicating that these are the areas where electrochemical reactions are most active and local heat generation rates are very high. As the cooling effect stabilizes the internal heat transfer rates, the temperature increase rate reduces significantly. The temperatures of these active sites exceed 40
C for nearly 500s of run time as compared to the other zones. The eight temperature measurements within the stacks were averaged to represent a single stack temperature for a local period. The averaged stack temperature analysis is the intermediate phase in linking first-order temperature analysis to the second-order heat transfer analysis. Fig. 11 displays the stack bulk temperature profile of each stack at loads 20 A and 30 A and varied air relative humidity. The temperature profile fits the lumped capacitance model for bulk system cooling where maximum convection effect within the cooling channels is also governed by the limiting rate of heat conduction within the stack. A sharp temperature increase at the beginning of the operation is followed by a decreasing rate of temperature rise as the internal conduction and convection rates reach a near-equilibrium state. In the time frame of the operation (500 se600 s), all stacks have yet to reach steady-state cooling where the stack temperature increase would be very small. However, this condition is nearly achieved at 20 A loads due to the lower heat generation rate within the stack. It is expected that the steady-state cooling for high loads would be achieved around 1000s as found in Ref. [12]. Different cooling channel designs and cooling modes give slightly different bulk temperature profiles. Stack temperature increase rates are dependent on the stack heat energy gain for a designated time period which is relative to the actual cooling effects within the cooling plate. Fig. 11 proves

Table 9 e Applied parameters in theoretical fuel cell power analysis. Operating stack temperature, Ts (K) 313

PH2 (bar abs)

PO2 (bar abs)

Membrane thickness, l (mm)

Parametric coeff, B

Active area, A (cm2)

No. of cells, ncell

Max. current, Imax (Amps)

Electronic resistance, Rel (U)

l

1.4

0.294

2

0.1

240

3

50

0.0003

16.8

10618

i n t e r n a t i o n a l j o u r n a l o f h y d r o g e n e n e r g y 4 0 ( 2 0 1 5 ) 1 0 6 0 5 e1 0 6 2 6

Fig. 10 e Transient temperature profiles of stack Ch20 at 30 A and 50% RH.

that higher electrical loads would lead to a higher heat generation within the stack; thus, higher stack temperatures are recorded as the cooling load increases. Applying greater air humidity also reduces the stack temperature as the wall heat fluxes are also influenced by improved latent cooling effects leading to higher heat rejection from the stack to the cooling airstream. Evaluating the cooling effectiveness of the three cooling channel designs based on first-order stack temperature profile alone could lead to inaccurate analysis as the initial stack temperatures and the stack thermal powers are different from one another. The 40-channel straight flow cooling channel design (Ms40) registers lower maximum stack temperatures than the other stacks but its initial condition is also the lowest (25
C). Meanwhile, the 20-channel chaotic flow configuration operation was initiated at 28
C stack temperature. The stack heat generation at 30 A for Ms20 and Ch20 is also 5% lower and 2% higher respectively as compared to Ms40. The evaluation of cooling effectiveness is therefore approached from the second-order thermal analysis which accounts for the actual heat transfer rates, actual wall heat fluxes and the overall heat transfer coefficient for each design. The average temperatures in Fig. 11 are translated to second-order thermal analysis to trace the changes in internal stack thermal energy using Eq. (15). Fig. 12 provides an example of the transient stack heat gain profile and the relative cooling rate profile of stack Ch20 at 50% cooling air relative humidity. Steady stack thermal power was assumed in this analysis although slight variations are encountered in practice due to small voltage changes at constant load. At 20 A, the heat generated within the stack is 37.7 W while load 30 A has a heat generation rate of 61.2 W. In general, positive thermal power value for the stack heat changes means the stack is storing heat at the specified time frame (5 s intervals) leading to a rise in stack temperature. This outcome is due to lower cooling effect as compared to the generated thermal energy. For the cooling profile obtained from Eq. (16), the presented trend lines are the result of regression of the actual dynamic profile. The thermal power values of the cooling

profiles are negative that indicates heat is being rejected from the stack at the defined values. The actual heat gain profile indicates an inter-changing heating and cooling cycle within the stack. This dynamic thermal behavior of fuel cell stacks is related to the sensitivity of the thermal profile to small changes in bulk stack temperature at short time intervals. It is expected that the actual cooling rates within the stack is dynamic due to the combined factors of ambient air property changes, localized heat generation which is heterogeneous across the MEA of both electrodes and the non-uniform temperature difference within the cooling plate [12,39]. At the beginning of the cycle, the actual cooling rate is very low due to the small temperature difference that exists between the coolant and plate surface temperature. Over time, the stored heat increases the stack as well as the cooling surface temperatures and the cooling rate increases rapidly. The cooling gradient reduces over time as the plate temperature approaches steady-state condition. As shown in Fig. 12 for the stack Ch20 at 30 A and 50% cooling air relative humidity, the cooling rate increased steadily to 60 W, nearly equal to the rate of heat generated within the stack (61.7 W). Hence, the stack heat energy gain is very low at the corresponding period and highly limiting the rising rate of stack temperature. This cooling condition was also similar to the other stacks as shown in Fig. 14. In thermal engineering analysis, a value for steady-state cooling effectiveness is desirable to measure the overall impact of the cooling mode relative to the heat load applied. From the exponential stack cooling profile, an average cooling rate throughout the operation was calculated using Eq. (17) and presented by the linear horizontal lines in Fig. 12. The average cooling rate parameter is applied for steady-state analysis in the evaluation of the bulk cooling effectiveness derived from Eq. (18). In Fig. 13, the cooling effectiveness for each stack design under different load and cooling modes is presented. Overall, all cooling channel designs are capable of operating above 80% cooling effectiveness. Design Ms40 gives a consistent cooling

i n t e r n a t i o n a l j o u r n a l o f h y d r o g e n e n e r g y 4 0 ( 2 0 1 5 ) 1 0 6 0 5 e1 0 6 2 6

Fig. 11 e Average temperature profile for all stacks obtained from 8 measurement points within the stack assembly at all cooling modes; (a) stack Ms40, (b) stack Ms20 and (c) stack Ch20.

effectiveness trend when the load and air humidity increases. The cooling effectiveness was 88%e92% where higher air humidity slightly reduces the cooling effect by approximately 3%. For design Ms20, a contrasting trend was conceived when the air humidity changes. At low load with lower stack temperatures, the added moisture has a negative effect while an improved cooling effectiveness can be found at a higher load of 30 A. This shows that channel surface temperature is an important factor in the issue of compatibility in using humidified cooling air. The cooling effectiveness is improved only when the stack generates higher thermal energy (30 A). Meanwhile, the Ch20 design shows a similar trend at both loads where the cooling effectiveness increases by 4%e5% when the air humidity is increased. Smaller channel aspect

10619

ratios (Ms40) are more compatible with dry air while moist air has a positive effect on larger channel sizes due to the effect of inlet flow resistance. Overall, the stack cooling rate is not highly influenced by the introduction of humid air as the air flow rate is not significantly large for mass cooling effect. The maximum cooling improvement is approximately 3%. The generated thermal load of the fuel cell stacks increases by nearly 20 W for every increase of 10 A electrical loading. In general, all the cooling channel design effectiveness improves slightly at higher load due to better rates of convection at the stackecoolant interface even without increasing the mass flow rate of the cooling fluid. This is made possible by the existence of larger temperature gradients at the cooling channel interface as the thermal energy generated increases. It is expected that the cooling effectiveness will further increase if the load is increased; however, the bulk stack heat gain would be greater and a more rapid rise in stack temperature will be recorded. Hence, the cooling air flow rate must be increased to improve stack temperature control at higher electrical loads. The major positive effect of introducing humid air is the reduction of the period where effective stack cooling occurs for channels with small aspect ratios as shown in Fig. 14. It shows the stack cooling profile for stacks Ms40 and Ms20 was derived from similar methodology to stack Ch20 (as in Fig. 12). From stack Ms40 cooling profile, a 10% time reduction was found. Stack Ms40, in overall, has a slower cooling response as compared to stack Ms20 due to its smaller aspect ratio. However, when the asymptotic thermal saturation level is reached, having more cooling channels allows a higher degree of cooling magnitude to be obtained due to larger convective area. The effective cooling rate for stack Ms40 continues to increase exponentially, while in contrast, stack Ms20 indicates a more stable cooling rate with immediate effect from load initiation. It is also evident here that cooling stability and near instantaneous response to cooling are the strengths of stack Ms20 which has a bigger aspect ratio. In prior discussions, the internal heat removal from the stack was assumed to be fully contributed by the forced convection mechanisms. The reactant flow streams may also assist in removing heat internally. Operation at low stack temperatures promotes liquid water formation that was visually detected at the cathode exit stream during the experiment. The exit cathode fluid stream is therefore composed of saturated reactant air and liquid water. The cathode exit stream temperature is in the range of 37
C at 20 A load and 39
C at 30 A load. The temperature difference compared to the inlet is in the range of 2
Ce4
C. As water is an excellent carrier of heat, the direct cooling effect of the cathode stream was evaluated. From Eq. (26), the rate of water formation for the reported current load of 30 A is 2.8  106 kgH2 O per seconds. Assuming saturated liquid phase at 39
C, the heat content of the exit water is only 0.5 W (from Eq. (27)). The inlet-exit air temperature difference for the 30 A load experiments was approximately 4
C. Taking the flow rate of inlet reactant air to be 0.24 L/min and the specific heat as 1.007 kJ/kg
C, then the heat energy carried out by the air is 1 W (calculated using Eq. (23) to Eq. (25)). Therefore, the total heat energy carried out by the reactant air exit stream is 1.5 W from the total heat generation

10620

i n t e r n a t i o n a l j o u r n a l o f h y d r o g e n e n e r g y 4 0 ( 2 0 1 5 ) 1 0 6 0 5 e1 0 6 2 6

Fig. 12 e An example of dynamic stack energy change and cooling profile during current loads of 20 A & 30 A (50%RH) for chaotic flow (stack Ch20) cooling configuration.

of 60 W. The steady-state cooling rate evaluated for the 30 A load at 90% RH was 55 W. Thus, only 3% of the cooling rate was actually contributed by the reactant air and water exit streams in this case. The air cooling effect is proven to be dominant in removing the bulk of the heat from the stack with little assistance from the cathode fluid stream.

increasing even after 100 s of operation. This means fuel cell stacks with high loads will continue to have increasing temperature gradients across its internal plates as it is operated for longer periods. The thermal model of Sadiq and Shahad [40] shows the occurrence of intensified thermal stresses to critical parts of a fuel cell when large temperature gradients

Temperature uniformity index The temperature uniformity index, UT, is an essential parameter in thermal management of PEM fuel cells due to the required electrode temperature homogeneity for improved reaction kinetics. A lower index points to a more uniform temperature distribution within the cell. Figs. 15 and 16 provide both the change in temperature uniformity index over time and the maximum value for each stack design, calculated from Eq. (19). Generally, the index is always higher at higher electrical loads due to a larger thermal load that influences the internal surface temperatures. The index profile at low loads would increase for the first 100 s and then quickly reach a constant value. In contrast, higher thermal loads would cause the temperature uniformity index to continue

Fig. 13 e Cooling effectiveness of all designs at 20 A and 30 A loads with variations in cooling air humidity.

Fig. 14 e Stack cooling profile at (a) 50% RH and (b) 90% RH for stacks Ms40 and Ms20 at current loads of 20 A and 30 A.

i n t e r n a t i o n a l j o u r n a l o f h y d r o g e n e n e r g y 4 0 ( 2 0 1 5 ) 1 0 6 0 5 e1 0 6 2 6

10621

exist in a fuel cell. Cooling channels with larger aspect ratio and lower number of channels (Ms20 and Ch20) tend to limit the rate of temperature non-uniformity better. As shown earlier, the cooling effectiveness of Ms40 is generally better than Ms20 and Ch20. However, Fig. 16 shows that stack Ch20 with chaotic flow configuration promotes a more uniform plate temperature, followed by Ms20 and finally Ms40. The analysis confirms with those reported in Ref. [24] where cooling rates and temperature distribution relation could be inversely proportional. Larger aspect ratios allow the cooling air to enter the channels at a higher heat capacity. Heat transfer from the channel surface to the air at the inlet and middle regions will not drastically increase the air temperature as it reaches the exit region. Thus, the temperature difference between the surface and the fluid along the channel length remains suitable for effective heat transfer, leading to a more uniform cooling and plate temperature. In Ms40, the small cooling channel size will cause drastic changes to the air temperature profile with reduced heat transfer rates along the channel length; hence, the inlet and outlet region temperature difference is greater and it influences the local plate temperatures in the respective regions. The maximum temperature uniformity index is also not highly influenced by cooling air humidity except for the chaotic flow channel design that allows the uniformity index to improve by approximately 1.5 when humidity increases, which is quite significant in terms of overall cell temperature distribution. The non-linear but orderly motion of the air within the chaotic flow channel provides an enhanced and evenly distributed heat transfer as the air travels along the channel. Here, the results are based on a constant air flow rate. It is expected that the temperature uniformity can be improved by allowing higher air flow rates into the cooling channels as the air temperature gradient across the channel will be reduced due to the larger heat capacity.

Cooling flux Fig. 15 e Temperature uniformity index profile for individual stack designs at different loads and air relative humidity; (a) stack Ms40, (b) stack Ms20, and (c) stack Ch20.

Fig. 16 e The maximum temperature uniformity index for each stack under different operating modes.

In the previous analysis, the cooling effectiveness of Ch20 was lower across all cooling modes. However, the cooling flux analysis in Fig. 17 based on Eq. (20) indicates that the chaotic flow trajectory with smaller surface area successfully promotes a larger heat flux as compared to the straight flow configurations. Relatively, the high cooling flux of Ch20 fits the

Fig. 17 e Comparison of cooling flux for each stack design.

10622

i n t e r n a t i o n a l j o u r n a l o f h y d r o g e n e n e r g y 4 0 ( 2 0 1 5 ) 1 0 6 0 5 e1 0 6 2 6

temperature uniformity index analysis where the cooling effect is well-distributed along the channel. At lower heat load, the difference in cooling flux is approximately 40% and increases to nearly 50% difference at higher heat load with increased surface temperatures. As such, the lower cooling effectiveness of Ch20 might be due to the non-optimal air flow rate in the experiments reported here to overcome a bigger pressure drop across the channel bending patterns. From Eq. (22) as well as referring to the inlet and exit temperatures of air in Fig. 18 and the average stack temperatures in Fig. 11, the effective heat transfer coefficient, h, was calculated. Evaluation of h in Fig. 19 supports the cooling flux analysis as an indicator of actual cooling capability of the various cooling channel designs. Even though the hydraulic diameter of Ch20 is only 13% greater than design Ms40, the heat transfer coefficient is 25%e30% better in all cases. The total cooling rate and effectiveness limitation are highly due to limited total surface area or changes in flow momentum as the air negotiates the bends within the design. In the case of Ms40 design, it was analyzed that it has a lower cooling flux and heat transfer coefficient. The slower cooling response time was a direct manifestation to these conditions. The high cooling rates and effectiveness of stack Ms40 are mainly due to the larger surface area obtained from the higher number of channels. For Ms20, a very high heat transfer coefficient was recorded. In the design of Ms20, smoother flow in a straight configuration with bigger hydraulic diameter leads to minimal flow losses that influence the local Reynolds number in an actual internal flow environment with a constant momentum source. The larger hydraulic diameter and aspect ratio allow less inlet resistance while the wall surface friction effect usually encountered in tight enclosures is minimized that allows a more developed flow to occur with near optimal flow velocity. Drastic pressure changes due to bends are also eliminated. But, the introduction of larger bulk fluid density as the air is humidified to near

Fig. 19 e The effective heat transfer coefficient of the cooling plates based only on internal air cooling effects.

saturation would reduce the effective heat transfer coefficient by approximately 35% under these flow conditions.

Stack temperature monitoring The cooling system operation of PEM fuel cell stacks is normally controlled by direct feedback of the stack temperature. In practice, small stacks would apply a single temperature sensor or probe on the stack surface while larger stacks would need at least two sensors to improve accuracy. The choice of stack surface and exact position has an influence on the measured temperature value and the ensuing cooling operation to be executed such as cooling fan speed or coolant pump power. As shown earlier in Fig. 10, the temperatures at different locations/zones differ as the electrical load changes. An extended analysis on the temperature profile was conducted (refer Fig. 20) to statistically determine the most suitable positioning of temperature sensors based on its transient

Fig. 18 e The inlet and exit air temperatures obtained from bulk measurement for all the stack designs and cooling operation.

i n t e r n a t i o n a l j o u r n a l o f h y d r o g e n e n e r g y 4 0 ( 2 0 1 5 ) 1 0 6 0 5 e1 0 6 2 6

10623

the stack. However, it should be noted that every PEM fuel cell is inherently unique in its operation; thus, the stress here is on the engineering approach rather than the absoluteness of the recommended positions.

Limitations

Fig. 20 e The statistical analysis on the mean local temperature difference for stack Ch20.

local values and was compared to the average stack temperature at a similar period. Thus, the mean local temperature difference was defined as P DTm ¼

Tn  Tavg Nt

 t

D

(29)

where Tn,t is defined as the temperature at a local position in a specific time period, Tavg,t is the average stack temperature in a specific time period, and Nt is the total number of data sets for the whole designated cooling period. The mean local temperature difference was calculated for all the temperature sensor positions within the setup. Here, analysis on stack Ch20 is presented as the stack was proven to have a lower temperature uniformity index than the other stacks. The analysis is further limited to high electrical loads of 20 A and 30 A due to its significant effect on the temperature distribution as compared to lower loads. The analysis gives a general guide for the most optimal temperature sensor positioning to capture a representative temperature closest to the average stack temperature. Here, the most optimal position would be at sensor no. 2, which is at the top center zone of the stack as the mean local temperature difference is the lowest for both sets of electrical loads. The analysis also suggests that the measured data at this position needs to be offset or corrected by 1
C to obtain the approximate average stack temperature for stack cooling operation reference. Next, the second best indicated position is at sensor position 4 (middle inlet region) with an offset value of 2
C to þ0.3
C. The third best position is at sensor no. 5 (middle center region) with an offset value between 1.2
C to 1.7
C. However, in practice, the middle center region of the stack cannot be reached directly as compared to the top center and middle inlet regions which are exposed surfaces. For cooling system control based on maximum stack temperature, the analysis also provides the locations where the sensors can be positioned. The positive values in the analysis such as at positions 6, 7 and 8 (middle exit, bottom center and bottom exit regions respectively) indicate zones with higher than average temperatures. For critical stack temperature control, the bottom center region should be closely monitored as it is found to be the hottest region within

The results provided various in-depth experimental analyses on the thermal engineering of PEM fuel cell stacks for improved understanding of a cooling plate design performance under specific operation. Adapting separate air cooling technique to PEM fuel cell stack designs is subjected to various limitations. Here, the limitations of the results in view of the practical aspects are discussed in brief. The issues highlighted are the stack performance, the relation of water and thermal management, the effect of higher current density, the parasitic load, the influence of ambient air and stack temperature mapping. The current density and power density of the stacks as reported in Fig. 6 are low compared to commercial stacks. The stacks were built using a locally developed MEA and stacking technology. The theoretical analysis that fits the actual polarization curve of the stacks indicates that the equivalent parametric coefficient of the stacks is 0.1 at 50 A maximum current. Highly efficient stacks would have parametric coefficients in the range of 0.001e0.003 that limits the concentration losses within the stack and allowing the stack to operate at good current densities. High parametric coefficient value is a sign of either poor reactant distribution (a flow field design factor) or dominant water flooding effect. The second factor is the most likely contributor in this case study as the stack operating temperature is less than 60
C. Water and thermal management are critical aspects in the operation of an efficient fuel cell system as the byproducts of fuel cell electrochemical reactions are water and heat. These byproducts should be continuously removed from the cell to maintain isothermal operation for electric power generation. Low stack temperatures would give rise to the water flooding phenomena which reduce the rate of reactant mass supply to the reaction sites and degrading the cell performance [41]. Water management is closely related to thermal management due to water evaporation and condensation, particularly at the cathode electrode. Flooding of the cathode electrodebacking layer even in small amounts would lead to severe performance degradation [42]. Therefore, the operating setting of the stack relative to the air cooling system must be outlined to obtain an optimal stack operating temperature that limits the flooding effect. A delay in initiating the cooling system is an alternative mode of operation. The fuel cell is allowed to initially operate without any cooling assistance to a specific temperature that inhibits liquid water formation. The cooling system is then initiated at this temperature (or slightly higher) with suitable flow rates to maintain the desired operating temperature. The feasibility of the presented results for actual stack designs also depends on whether the air cooling technique is effective enough to remove high density thermal power from a very efficient stack. Typical range of PEM fuel cell current and power densities are 750 mA/cm2 to 2800 mA/cm2 and

10624

i n t e r n a t i o n a l j o u r n a l o f h y d r o g e n e n e r g y 4 0 ( 2 0 1 5 ) 1 0 6 0 5 e1 0 6 2 6

0.4 W/cm2 to 1.3 W/cm2 for operation between 50
C and 120
C [43]. If the thermal power is one order of magnitude higher, then the air cooling technique that was initiated concurrently with the load will be viable if the air flow rates within the channels are increased proportionally. Theoretically, if the current density is 1000 mA/cm2, the thermal power generated by the stacks would be in the range of 450 We500 W (based on calculations using Eq. (1) to Eq. (14) at 240 A maximum current). This is 6e7 times the thermal load reported in this study. To cope with higher cooling demands, the mass flow rate of cooling air needs to be increased in proportion as the Prandtl number of air is near to 1. The use of more powerful fans are needed, but the apparent penalty would be on the parasitic load and sound pollution associated with powerful blowers. As fuel cells are power generators, the net power output is an important consideration. Accounting only the required cooling fan power requirement as the main parasitic load, the percentage of parasitic load is calculated from Parasitic load ¼

Pfans ð%Þ Pel;stack

(30)

In the study, the fan power rating was 12 V and 0.38 A. The 2 fan arrays consisting of 6 fans then consume a maximum of 27.6 W of power. At 20 A and 30 A stack electrical loads, the electrical power generated by the stacks were 40 W and 60 W respectively. This leads to a parasitic load percentage from 45% to 70% of the stack generated power. As the stack current density increases, higher flow rates are required for enhanced cooling rates. The parasitic load and net power output in this study was not a major concern as the focus is in characterizing the cooling effects. However, the parasitic load in real designs and operation should be limited to 30% as higher parasitic loads would severely limit the net power output. This is the major drawback in using air as a cooling agent compared to water due to its lower density and lower specific heat. Thus, energy efficient fans with high flow rates and high induction pressure is a compulsory component in practical stack system design. Air cooling is highly influenced by local ambient conditions where the cooling air inlet temperature is assumed similar to the ambient temperature. The ambient temperature reported in the study was based on a ventilated room environment. In practice, ambient air temperatures can vary significantly depending on the applied location such as in closed enclosures with poor ventilation or outdoors. The mean temperature difference between the air and plate surface (average stack) temperatures is the main governing force for convection heat transfer. Higher mean temperature differences leads to lower required flow rates. In this work, the mean temperature difference for all cases is in the range of 1.8
Ce6
C where improved values were obtained in 90% humidity conditions. Large mean temperature difference can still be adequately achieved in fuel cell stack operations at ambient air conditions lower than 32
C such as encountered in this work, or by operating the stack to a higher temperature before the cooling system is initiated. In extreme conditions where the ambient air is greater than 40
C, the mean temperature difference would reduce and have a less significant effect as a heat transfer driving

force. Instead, the Reynolds number must be increased significantly to have the required cooling effect. When ambient conditions are too hot, technically, the cooling system must be turned on from load initiation and steady-state conditions will take longer to be achieved. Practically, the application of air cooling in ambient temperatures above 40
C is not recommended as it would consume a lot of energy to power the momentum source with minimal cooling effect due to the very small difference in mean temperatures that can be achieved. Based on the study presented, inducing large flow rates would have greater effect for channels with large hydraulic diameters. The relations of cooling rate, average stack temperature and temperature distribution are important issues in fuel cell thermal management and control [44,45]. In this study, the number of temperature sensors inserted into the channels was limited, representing area coverage of 30 cm2 per sensor from the total effective cooling area (240 cm2) of a single cooling plate. It is acknowledged here that the area coverage is still large for an accurate temperature mapping analysis. Sensor area coverage of less than 10 cm2 per sensor would give a greater accuracy. Practical mapping of temperature requires evenly distributed temperature sensors within the stack. The results of cooling effect on segmented stack or cell temperature distribution would be accurate if: (i) a high number of internal sensors are used to obtain a low measurement area coverage, (ii) reaction sites are homogeneous across the cells which depends on unique number of electrochemical and thermofluid factors, (iii) the heat generated is uniform across the cells, and (iv) the coolant flow is evenly distributed. It is important to realize that fuel cell active areas are dynamic at both anode and cathode depending on the load requirement, local reactant concentration and local site environment (pressure and temperature). Temperature mapping under these conditions will lead to the identification of active sites as local temperature will be influenced by local heating and cooling rates. Insertion of sensors through open-ended cooling channels will generate an increased resistance to the air flow along the channels and subsequently reduce the convection mechanism. Temperature mapping would also be most valuable at the anode and cathode surfaces compared to temperature mapping within the cooling plate. In practice, this is far more difficult to achieve unless non-intrusive sensory devices are used such as infrared thermography [46] or miniaturized sensors such as the recently developed micro 3-in-1 sensor [47]. The local temperatures obtained can be used as boundary conditions in an analytical or numerical study to map a detail temperature distribution [48], but dynamic active areas would limit the accuracy of the mapping. Inclusion of dynamic active areas is possible with a detail coupling of electrochemical reaction models, cell local environment model and cooling effect model which is an interesting subject for future publications.

i n t e r n a t i o n a l j o u r n a l o f h y d r o g e n e n e r g y 4 0 ( 2 0 1 5 ) 1 0 6 0 5 e1 0 6 2 6

Conclusion [3]

The work here provides an empirical methodology for the evaluation of cooling performance of a specific PEM fuel cell cooling plate design using air as the cooling medium. A detailed experimental and analytical methodology of acquiring the transient stack heating and cooling profile, temperature uniformity index, average cooling rate, cooling effectiveness, cooling flux and overall heat transfer coefficient from the physical electrical and temperature measurements for a PEM fuel cell were established. Using closed cathode aircooled PEM fuel cell stacks where the reactant air and cooling air are separate streams, ambient air was inducted into the cooling channels assisted by the simultaneous operation of a coupled upstream fan blower and downstream suction fan with experiments using dry (50% RH) and moist (90% RH) air. Analysis shows that different channel aspect ratios give different thermal behavior. Generally, a fuel cell stack would undergo a dynamic cycle of net heat gain and net heat loss during operation which is highly dependent on the flow condition uniformity and the local temperature difference within the stack. The larger cooling potential is shown by smaller channel aspect ratio but with much slower cooling response time as the heat transfer coefficient and cooling flux is lower. The cooling effect is more responsive and stable for larger channels. Further analysis also shows the unique compatibility of the designs with dry and moist air as the cooling medium. The chaotic flow trajectory was shown to have a very high cooling flux and behave effectively in the presence of moist air due to its non-linear motion and is advantageous when plate temperature uniformity is the priority. Optimal temperature sensor locations were statistically proposed from the reported temperature profile which can assist the stack cooling system control to obtain the average stack temperature from local temperature measurements. Finally, limitations of the study were discussed which includes the cooling scenario at higher thermal power density, effect of ambient conditions to the cooling capability, cooling contribution from the cathode fluid stream and the challenges in internal temperature mapping. Future work would expand from this methodology and discussions to obtain more information on the thermal behavior of the designs when operated at different settings.

[4]

[5]

[6]

[7]

[8]

[9]

[10]

[11]

[12]

[13]

[14]

[15]

[16]

[17]

[18]

Acknowledgments This work was supported by the Malaysian Ministry of Higher Education through the FRGS research grant no. 600-RMI/ST/ FRGS 5/3/Fst (76/2010).

[19]

[20]

references

[1] Oosterkamp PF. Critical issues in heat transfer for fuel cell systems. Energy Convers Manag 2006;47:3552e61. [2] Kamarudin SK, Daud WRW, Yaakub Z, Misron Z, Anuar W, Yusuf NNAN. Synthesis and optimization of future hydrogen

[21]

[22]

10625

energy infrastructure planning in Peninsular Malaysia. Int J Hydrogen Energy 2009;34:2077e88. Squadrito G, Barbera O, Giacoppo G, Urbani F, Passalacqua E. Polymer electrolyte fuel cell stack research and development. Int J Hydrogen Energy 2008;33:1941e6. Baschuk JJ, Li X. A comprehensive, consistent and systematic mathematical model of PEM fuel cells. Appl Energy 2009;86:181e93. Zhang G, Kandlikar Satish G. A critical review of cooling technology in polymer electrolyte membrane fuel cell stacks. Int J Hydrogen Energy 2012;37:2412e29. Kandlikar SG, Lu Zijie. Thermal management issues in a PEMFC stack e a brief review of current status. Appl Therm Eng 2009;29:1276e80. Mosdale R, Srinivasan S. Analysis of performance and of water and thermal management in PEM fuel cells. Electrochim Acta 1995;40(4):413e21. Millera M, Bazylaka A. A review of polymer electrolyte membrane fuel cell stack testing. J Power Sources 2011;196:601e13. Mohamed WA Najmi W, Atan R. Analysis of stack heating on the thermal and electrical resistance characteristics of polymer electrolyte membrane fuel cell. Int J Automot Mech Eng 2012;5:648e59. Pharoah JG, Burheim OS. On the temperature distribution in polymer electrolyte fuel cells. J Power Sources 2010;195:5235e45. Yu S, Jung D. Thermal management strategy for a proton exchange membrane fuel cell system with a large active cell area. Renew Energy 2008;33:2540e8. Wen C-Y, Huang G-W. Application of a thermally conductive pyrolytic graphite sheet to thermal management of a PEM fuel cell. J Power Sources 2008;178(1):132e40. Kang S, Mina K, Muellerb F, Brouwerb J. Configuration effects of air, fuel, and coolant inlets on the performance of a proton exchange membrane fuel cell for automotive applications. Int J Hydrogen Energy 2009;34:6749e64. Andisheh Tadbir M, Shahsavari S, Bahrami M, Kjeang E. Thermal management of an air-cooled PEM fuel cell: cell level simulation. In: ASME 2012 10th International Conference on Fuel Cell Science, Engineering and Technology, San Diego, California, USA; July 23e26, 2012. Hwang Jenn-Jiang. Thermal control and performance assessment of a proton exchange membrane fuel cell generator. Appl Energy 2013;108:184e93. Faghri A, Guo Z. Challenges and opportunities of thermal management issues related to fuel cell technology and modeling. Int J Heat Mass Transfer 2005;48(19e20):3891e920. Adzakpa KP, Ramousse J, Dube Y, Akremi H, Agbossou K, Dostie M, et al. Transient air cooling thermal modelling of a PEM fuel cell. J Power Sources 2008;179:164e76. Matian M, Marquis A, Brett D, Brandon N. An experimentally validated heat transfer model for thermal management design in PEM fuel cells. Proc IMechE 224 Part A J Power Energy 2010;224(8):1069e81. Wen CY, Ys Lin, Lu CH, Luo TW. Thermal management of a proton exchange membrane fuel cell stack with pyrolytic graphite sheets and fans combined. Int J Hydrogen Energy 2011;36:6082e9. Tolj I, Abdallah M, Barbir F. Thermal behaviour of an air cooled PEM fuel cell stack. ECS Trans 2008;16(2):535e42. Y-Jun Sohn, G-Gon Park, T-Hyun Yang, Y-Gi Yoon, W-Yong Lee, S-Dae Yim, et al. Operating characteristics of an air-cooling PEMFC for portable applications. J Power Sources 2005;145:604e9. Shahsavari S, Bahrami M, Kjeang E. Computational analysis of heat transfer in air-cooled fuel cells. In: Proc. of ASME 2011, 5th Int. Conf. Energy Sustainability & 9th Fuel Cell

10626

[23]

[24]

[25]

[26]

[27]

[28]

[29]

[30]

[31] [32]

[33]

[34]

[35]

i n t e r n a t i o n a l j o u r n a l o f h y d r o g e n e n e r g y 4 0 ( 2 0 1 5 ) 1 0 6 0 5 e1 0 6 2 6

Science, Engineering and Technology Conference. ES Fuel Cell 2011. Matian M, Marquis A, Brandon N. Model based design and test of cooling plates for an air-cooled PEM fuel cell stack. Int J Hydrogen Energy 2011;36:6051e66. Baek SM, Yu SH, Nam JH, Kim C-J. A numerical study on uniform cooling of large-scale PEMFCs with different coolant flow field designs. Appl Therm Eng 2011;31:1427e34. Lasbet Y, Auvity B, Castelai C, Peerhossaini H. A chaotic heatexchanger for PEMFC cooling applications. J Power Sources 2006;156:114e8. Sasmito AP, Birgersson E, Lum KW, Mujumdar AS. Fan selection and stack design for open-cathode polymer electrolyte fuel cell stacks. Renew Energy 2012;37(1):325e32. Mohamed WA Najmi W, Atan R. Experimental cooling mode variation of an air-cooled PEM fuel cell using second-order thermal analysis. J Mech Eng 2014;10(2):55e78. Sasmito AP, Lumb KW, Birgersson E, Mujumdar AS. Computational study of forced air-convection in opencathode polymer electrolyte fuel cell stacks. J Power Sources 2010;195:5550e63. Sasmito AP, Shamim T, Birgersson E, Mujumdar AS. Numerical investigation of water and temperature distributions for open-cathode Polymer Electrolyte Fuel Cell stack with edge cooling. J Fuel Cell Sci Technol 2013;10(6):61003-1e61003-9. Franco A. Optimum design of bipolar plates for separate air flow cooling system of PEM fuel cells stacks. Heat Mass Transfer 2015. http://dx.doi.org/10.1007/s00231-015-1530-6. Incropera FP, DeWitt DP. Fundamentals of heat and mass transfer. 5th ed. John Wiley & Sons; 2002. p. 240e3. Hua P, Ca G-Y, Zhu X-J, Hu M. Coolant circuit modeling and temperature fuzzy control of proton exchange membrane fuel cells. Int J Hydrogen Energy 2010;35:9110e23. Graf C, Vath A, Nicoloso N. Modeling of the heat transfer in a portable PEFC system within MATLAB-Simulink. J Power Sources 2006;155:52e9. Zhang Y, Ouyang M, Lu Q, Luo J, Li X. A model predicting performance of PEM fuel cell stack thermal system. Appl Therm Eng 2004;24:501e13. Spiegel CS. Designing and building fuel cells. USA: McGraw Hill; 2007.

[36] Mohamed W A Najmi W. Solid-state thermal analysis of aircooled PEM fuel cells with predictive empirical profiling [PhD. thesis]. Universiti Teknologi MARA; 2012. p. 85e100. [37] Holman JP. Experimental methods for engineers. 7th ed. New York: McGraw Hill; 2001. p. 50e60. [38] Wang Y, Wang C-Y. Ultra large-scale simulation of polymer electrolyte fuel cells. J Power Sources 2006;153:130e5. [39] Wu J, Yuan XZ, Wang H, Blanco M, Martin JJ, Zhang J. Diagnostics tools in PEM fuel research: part II physical/ electrical models. Int J Hydrogen Energy 2008;33(6):1747e57. [40] Sadiq al-Baghdadi MAR, Shahad al-Janabi HAK. Prediction of hygro-thermal stress distribution in PEM using a threedimensional multi-phase computational fluid dynamics model. Proc of IMechE 2221 Part A J Power and Energy 2007. [41] Misran E, Hassan NSM, Daud WRW, Majlan EH, Rosli MI. Water transport characteristics of a PEM fuel cell at various operating pressures and temperatures. Int J Hydrogen Energy 2013;38:9401e8. [42] Li Xianguo. Principles of fuel cells. New York: Taylor & Francis; 2006. p. 306. [43] Peighambardoust SJ, Rowshanzamir S, Amjadi M. Review of the proton exchange membranes for fuel cell application. Int J Hydrogen Energy 2010;35:9349e84. [44] Soupremaniem U, Le Person S, Favre-Marinet M, Bultel Y. Tools for designing the cooling system of a proton exchange membrane fuel cell. Appl Therm Eng 2012;40:161e73. [45] Pei H, Shen J, Cai Y, Tu Z, Wan Z, Liu Z, et al. Operation characteristics of air-cooled proton exchange membrane fuel cell stacks under ambient pressure. Appl Therm Eng 2014;63:227e33. [46] Gould BD, Ramamurti R, Osland CR, Swider-Lyons KE. Assessing fuel-cell coolant flow fields with numerical models and infrared thermography. Int J Hydrogen Energy 2014;39:14061e70. [47] Lee C-Y, Fang L-H, Su A, Liu Y-M, Lee C-J. Real time interior water and thermal monitoring of fuel cell with micro 3-in-1 sensor. Renew Energy 2015;74:517e22. [48] Cao T-F, Mu Y-T, Ding J, Lin H, He Y-L, Tao W-Q. Modeling the temperature distribution and performance of a PEM fuel cell with thermal contact resistance. Int J Heat Mass Transfer 2015;87:544e56.