Renewable Energy 36 (2011) 451e458

Contents lists available at ScienceDirect

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

Analytical modeling of PEM fuel cell ieV curve

Shaker Haji*
College of Engineering, Department of Chemical Engineering, University of Bahrain, P.O. Box 32038, Bahrain

a r t i c l e i n f o
Article history: Received 29 July 2008 Accepted 14 July 2010 Available online 17 August 2010 Keywords: PEM fuel cell Analytical model Concentration overpotential Exchange current density Current loss

a b s t r a c t
The performance of a fuel cell is characterized by its ieV curve. In this study, the performance of a bench scale fuel cell stack, run on hydrogen/air, is measured experimentally for different air ow rates and temperatures. The experimental data, obtained from the 40-W proton exchange membrane fuel cell (PEMFC), are used in estimating the parameters of a completely analytical model that describes the ieV curve. The analytical model consists of the three fundamental losses experienced by a fuel cell, namely: activation, ohmic, and concentration losses. The current loss is also considered in the model. While the Tafel constants, ohmic resistance, and the concentration loss constant are estimated through regression, the limiting current density and the current loss are obtained through measurements. The effect of temperature on the fuel cell performance, exchange current density, and current loss is also investigated. Both the exchange current density and the current loss are plotted against temperature on an Arrheniuslike plot and the related parameters are estimated. The theoretical equations derived in the literature, which model fuel cell performance, are found to reasonably t the obtained experimental data. 2010 Elsevier Ltd. All rights reserved.

1. Introduction Fuel cell is an electrochemical device which converts fuel (e.g. hydrogen) directly to electricity without undergoing combustion. Fuel cell technology is suitably integrated into the renewable energy scheme. When fueled with hydrogen that is obtained from renewable energy resources, fuel cell is characterized as zeroemission energy generating technology. Even when fed by other fuels, the fuel cell system releases low emission since no combustion reactions take place and less fuel is consumed e for a desired power output e due to the high fuel cell efciency. A hydrogen operated fuel cell can be integrated with renewable energy technologies, such as wind turbines and photovoltaic cells, through water electrolyzer to construct what is known as a regenerative fuel cell system. When the renewable energy source is available, the excess electricity is used to generate hydrogen from water through the electrolyzer. In the absence of the renewable energy source, the stored hydrogen is used to generate electricity. In this case, hydrogen is used as an energy storage medium. The performance of a fuel cell is shown by a graph of its output voltage (V) versus the drawn current density (i). This graph, which is also referred to as the ieV or polarization curve, is the most

important characteristic of a fuel cell. It is a convenient tool for the design and optimization of fuel cells. For a given temperature, pressure, and set of species concentrations, thermodynamics can determine the maximum possible (ideal) voltage that a fuel cell can theoretically produce. However, the actual voltage output is always less than the ideal voltage due to irreversible losses. These losses increase as more current is drawn from the fuel cell. Therefore, the more current that is generated, the lower the voltage output of the fuel cell is. These irreversible losses (also called overpotentials) are caused by the following factors: kinetics of the electrochemical reactions on the electrodes, internal charge transport resistance, and mass transport limitations in the electrodes. Therefore, the fuel cell voltage output can be described by the following equation:

V Ethermo hact;a hact;c hohmic hconc;a hconc;c

(1)

* Tel.: 973 17 876050; fax: 973 17 680935. E-mail address: haji@eng.uob.bh. 0960-1481/$ e see front matter 2010 Elsevier Ltd. All rights reserved. doi:10.1016/j.renene.2010.07.007

where V is the real output voltage; Ethermo is the thermodynamically predicted voltage; hact,a and hact,c are the activation losses on the anode and cathode, respectively, due to reaction kinetics; hohmic is the ohmic loss due to ionic and electrical resistances; and hconc,a and hconc,c are the concentration losses on the anode and cathode, respectively, due to mass transport. For a hydrogen/air fuel cell, the thermodynamic potential, Ethermo, is calculated from the modied Nernst equation that takes into account both temperature and pressure changes:

452

S. Haji / Renewable Energy 36 (2011) 451e458

Ethermo E0

Db s
nF

T T0

aH2 O RT ln 1=2 nF aH2 $aO2

! (2) c

RT 1 1=a: nF

(5b)

where E0 is the standard-state reversible voltage, Db s is the entropy of reaction (assuming it is independent of temperature), n is the number of electrons transferred in the reaction, F is Faraday constant, T is the absolute temperature, T0 is the reference temperature, R is the universal gas constant, and aH2 O , aH2 , and aO2 are the activities of water, hydrogen, and oxygen, respectively. The activation overpotential, hact, is described by the ButlereVolmer equation. In the case of large hact, resulting from slow electrode reaction kinetics and/or large amount of drawn current, the ButlereVolmer equation can be reduced to the Tafel equation (written here in the natural logarithmic form):

It is worth mentioning that another equation, but a rather empirical one, is often used to describe the concentration overpotential and is of the form:

hconc m$expn$i

(5c)

hact a b lni

(3)

where the intercept a RT =anF lni0 and the slope b RT =anF . The constant i0, which is known as the exchange current density, is an indication of the rates of the forward and backward reactions at equilibrium. The transfer coefcient, a, is a parameter that is related to the electrode kinetics. The voltage loss that arises from the resistance to charge transport obeys Ohms law and, therefore, is called ohmic loss (hohmic). The ohmic loss is due to both ionic and electronic resistances within the fuel cell and it is expressed as:

where m and n are parameters that have to be evaluated experimentally [5]. This equation is suggested by Kim et al. [5] but is widely used by others [6e8,13]. In PEM fuel cell running on hydrogen/air, the activation loss in the anode can be neglected compared to that in the cathode because the oxygen reduction half-reaction has far slower kinetics compared to the hydrogen oxidation half-reaction. In addition, mass transfer limitations of oxygen transport are much more severe compared to that of hydrogen because air rather than pure oxygen is used and because O2 diffuses slower than H2 [9]. For instance, Santarelli et al. [10] reported iL of 9 A/cm2 for the cathode versus 43 A/cm2 for the anode. Therefore, Eq. (1) can be simplied to:

V Ethermo hact;c hohmic hconc;c

(6)

hohmic i$Rohmic i$Rionic Relectronic

(4)

where Rohmic is the total internal cell resistance, Rionic is the resistance to ion transport within the electrolyte (including the electrolyteeelectrode interface), and Relectronic is the resistance to electron transport in the electron conductive parts. All of these resistances are area specic resistances (U cm2). Concentration overpotential is attributed to reactants and products concentration gradients between the ow channels and the catalyst reaction sites. On the cathode side for instance, the oxygen concentration in the bulk (within the ow channel) is usually higher compared to its concentration on the catalyst active sites and this difference increases with the increase of the drawn current. In the limiting case of mass transport, where the reactants concentrations at the catalyst active sites drop to zero, the current produced is called the limiting current density (iL). The limiting current density is the maximum current that can be drawn from the cell. Reactant depletion and product accumulation cause reduction in the actual voltage for two reasons [9]: reduction in the thermodynamic cell voltage (lower Ethermo) and reduction in the rate of reaction (higher hact). The voltage loss that is caused by mass transport limitation is expressed by the following theoretical equation:

In addition, a fourth kind of loss is experienced in fuel cell voltage. This loss is due to internal current leakage, hydrogen crossover, and unwanted side reactions. This loss is represented in terms of current density: iloss (A/cm2). This additional loss has a noticeable negative effect on the hact and hconc. The current loss, iloss, has a pronounced effect at small current density, especially when i 0, where the open circuit voltage is reduced below its predicted thermodynamic voltage. The current loss also has an effect at high current density where the actual limiting current density is reduced by the current loss [9]. Therefore, Eq. (6) can be written in terms of the effects of the pronounced losses as:

V Ethermo ac bc $lni iloss Rohmic $i   iL;c cc $ln iL;c i iloss

(7)

where the parameters ac, bc, cc, and iL,c correspond to the cathode electrode. In this paper, the experimentally obtained ieV data of a 40-W fuel cell is modeled by estimating the parameters in the theoretical equation that describes the voltage output versus the current drawn, namely Eq. (7). The tted parameters are the area specic internal resistance of the cell, Rohmic, and the Tafel intercept, ac, Tafel slope, bc, and concentration loss constant, cc, of the cathode side. In addition, the current loss, iloss, and the limiting current density, iL,c, the exchange current density, i0,c, and the transfer coefcient, ac, for the oxygen reduction are determined. Furthermore, the effect of temperature on i0,c and iloss is studied. 2. Experimental and regression analysis The system used in this study is a Heliocentris hy-Expert Instructor fuel cell system. The system is based on a fuel cell stack with a rated and maximum power outputs of 40 W and 50 W, respectively. The stack consists of 10 cells connected in series with a cell active surface area (A) of 25 cm2. The membrane has a thickness of 27 mm and the electrodes are platinum loaded with 0.3 mg Pt/cm2. The fuel cell stack is operated on pure hydrogen (99.999 vol %) and air without prior humidication. Hydrogen is supplied to the stack at around 1.6 bar and the ow rate is measured by a digital hydrogen ow meter. The hydrogen is supplied in a dead-end mode. However, a purging valve automatically opens at intervals to

hconc c ln

iL iL i

 (5)

where c is the concentration loss constant. The constant c is commonly dened as [1e4]:

RT nF

(5a)

where this equation is derived based on the effect of the concentration loss on the thermodynamic voltage only. However, OHayre et al. [9] derived a theoretical equation for c in which the effect of mass transport limitation on both the thermodynamic voltage and the reaction rate are included:

S. Haji / Renewable Energy 36 (2011) 451e458

453

purge the system in order to prevent the accumulation of inert gases and water vapor within the anode. Air is blown into the cells, in parallel, at atmospheric pressure by two fans attached to the cell stack. For each cell, the air is passed through parallel channels with individual inlets and outlets. The fans provide air for both the electrochemical reaction and for cooling. This puts a limitation on controlling the temperature and the air ow rate independently. The two fans are either controlled by the user or by an internal controller. At full fan speed, the ow rate from both fans, is around 478 slm (standard liter per minute; STP). However, the fan can be run at selected percentages of the maximum fan speed. When the fan is controlled by the internal controller, the fan speed is set automatically according to the stack power output so that adequate cooling is ensured. Therefore, the more current is drawn from the fuel cell stack, the more air is provided to the stack. In addition, during the automated fan operation, enough air is provided to the cells to minimize mass transfer limitations between the ow channels and the electrode in the cathode side. It should be noticed that at any given current, increasing the fan speed e above the automatically set speed e does not increase the output fuel cell voltage. Therefore, it is concluded that the fuel cell does not experience noticeable concentration loss when the air ow rate is controlled by the internal controller. The value of the current loss, iloss, is assumed to be related only to the hydrogen consumption/ow rate measured at open circuit. Therefore, iloss is calculated by multiplying the hydrogen ow rate at open circuit by 2F/A. The apparent limiting current density is obtained experimentally. The current at which the voltage drops rapidly, approaching zero, is designated to be the apparent limiting current density. The software Polymath 5.1 is used for multiple nonlinear regression. The program uses the LevenbergeMarquardt (LM) algorithm e for nding the tted parameter values e which minimizes the sum of squares of the errors. Microsoft Ofce Excel 2007 is used for multiple linear regression. The Regression Analysis tool in Excel performs linear regression analysis by using the least squares method to t a line through a set of data.

3. Results and discussion Fig. 1 shows the ieV curves of the fuel cell stack when operated at 40 1.5  C with different air ow rates. The data are recorded for positive and negative load increments, and also for random load values. For the same operating condition, it is noticed that the fuel cell exhibits better performance (i.e. higher V for a given i) during the negative load increments runs compared to the positive load increments runs. The better performance during the negative load increments runs might be due to the higher membrane water content, which results in less ionic resistance of the membrane. When starting with a high load, the amount of water produced at the cathode is more, compared to that when starting with a low load, resulting in a membrane that is wetter, which in turn leads to better ionic conductivity [11,12]. The same phenomenon is observed by Hamelin et al. [13]. In two of the sets of experiments shown in Fig. 1, the fuel cell is deliberately operated under low air ow rates (i.e. 2.39 and 2.87 slm) in order to study the concentration loss effect and to obtain the related parameters. In the third experiment, the air ow rate is manipulated by a controller that increases the ow rate as more current is drawn. This action insures that the fuel cell is operated in a region where the concentration loss is negligible. In other words, the stack is run under current densities much smaller than the limiting current density. The effects of the activation, ohmic, and concentration losses along with that of the current loss are apparent in the rst two sets of experiments as shown in Fig. 1. At low current densities (i.e. i < 0.02 A/ cm2), where the ohmic and concentration losses are insignicant, the fuel cell behaves similarly, despite the difference in the air ow rates, since the operating temperature is the same. Maintaining the same operating temperature results in similar reaction kinetics and, therefore, comparable rate of reactions, which in turn leads to similar activation losses. At larger current densities (i.e. i > 0.2 A/ cm2), however, the fuel cell performs better with higher air ow rates, which reduce the concentration loss. Attempts to t the data to Eq. (7) using nonlinear regression analysis results in parameters with unrealistic values (e.g.

1.0

0.9

Air Flow: 2.39 slm Air Flow: 2.87 slm

Cell Voltage, V

0.8

Air Flow: Proportional to the Current Model Prediction

0.7

0.6

0.5

0.4 0.00

0.05

0.10

0.15

0.20

0.25

0.30

0.35

0.40

Current Density, A/cm2

Fig. 1. The fuel cell ieV curves obtained for different air ow rates at 40 1.5  C along with the tted curves.

454

S. Haji / Renewable Energy 36 (2011) 451e458

Table 1 The measured, tted, and the calculated ieV model parameters at 40 1.5  C and different air ow rates. Flow rate _ V air (slm) 2.39 2.87 i-dependent Measured parameter iL,c (A/cm2) 0.2581 0.3595 Fitted parametersa ac (V) 0.5457 0.0217 0.5342 0.0169 0.5360 0.0106 bc (V) 0.04449 0.00544 0.04232 0.00435 0.04281 0.00287 Rohmic (U cm2) 0.2843 0.0958 0.4140 0.0631 0.3956 0.0269 cc (V) 0.03851 0.0034 0.03936 0.00372 R2 0.9822 0.9881 0.9942 Calculated parameters

ac
0.1516 0.1594 0.1596

cc (V) (Eq. (5a)) 0.0067 0.0067

cc (V) (Eq. (5b)) 0.0512 0.0491

a The numbers following the tted parameters values represent the 95% condence intervals. The Tafel constants are based on Tafel equation written in the natural logarithmic form.

Rohmic 0.01 U cm2). Therefore, it is decided to t the data to the model (i.e. Eq. (7)) using multiple linear regression analysis. In order to nd the model parameters using multiple linear regression, the value of the limiting current density has to be obtained rst. This introduces a limitation to this data tting method. Since the limiting current density is obtained by restricting the air ow to the cathode, the measured limiting current density is, therefore, related to oxygen half-reaction on the cathode and, hence, given the symbol iL,c. The value of iL,c is found experimentally. The current at which the voltage drops rapidly e approaching zero e is designated to be the apparent limiting current density. Since the limiting current density is reduced by the current loss [9], the actual limiting current density, iL,c, is obtained by adding iloss to the apparent limiting current density. For instance, at air ow rate of 2.39 slm, the apparent current density is found, from Fig. 1, to have an average value of 0.2533 ( 0.0023) A/cm2. Since iloss is measured to be 0.0047 A/cm2, the actual limiting current density, iL,c, is estimated to be 0.2580 A/cm2. The actual limiting current density is the maximum current density that can be drawn from the fuel cell in the absence of current losses. As mentioned in Section 2, the value of iloss is calculated by measuring the hydrogen consumption/ow rate at open circuit. Furthermore, iloss is assumed to be constant for a given operating temperature over the whole range of the drawn current density. The validity of this assumption is investigated in this study and is presented later in this report. The thermodynamic voltage is estimated, for a given operating condition, using Eq. (2). The known parameters in Eq. (2), under the experimental conditions where liquid water product is expected and unit activity coefcients for gasses are assumed, are: E0 1.229 V, Db s 163:23 J=mol K, n 2, F 96,485.34 C/mole of e, T0 298.15 K, R 8.314 J/mol K, aH2 O 1, aH2 1:6, and aO2 0:21. Substituting for the known parameters, Eq. (2) can be written as a function of temperature only as follows:

Ethermo 1:229 8:459 104 T 298:15 1:337 105 $T (8)

The thermodynamic voltage at T 40 is calculated to be 1.2121 V. In the multiple linear regression analysis, V Ethermo is
C

considered the dependent variable whereas ln (i iloss), i, and ln [iL,c/(iL,c (i iloss))] are considered the independent variables. The analytical model e Eq. (7) e is tted to the data and the obtained regression parameters values are listed in Table 1. For the case in which the air ow rate is proportional to the current generated and the concentration loss is not pronounced, the last term in Eq. (7) is cancelled and, therefore, only 3 parameters are tted. By inspecting the values of the tted parameters the following are observed. Increasing the air ow rate results in higher cathode limiting current density. Close values of Tafel intercepts, ac, as well as Tafel slopes, bc, are obtained conrming that the cathode kinetics are similar in the three cases. This is expected since the three sets of experiments are conducted at the same operating temperature. The Tafel intercepts and slopes are found to be around 0.5 V and 0.04 V, respectively. The transfer coefcients, ac, calculated from bc, are also found to have similar values in all the cases: between 0.15 and 0.16. This is also to be expected since the obtained values of ac are for the same reaction, i.e. the oxygen reduction reaction. Larminie and Dicks [2] consider the value of ac to range from 0.1 to 0.5 in most circumstances. The difference in the three values of Rohmic (0.28, 0.41, and 0.40 U cm2) could be due to the uctuation in the moisture content of the membrane e from one experiment to another e which leads to change in its ionic conductivity. A well-designed fuel cell would have Rohmic in the range of 0.05e0.10 U cm2 [9], while values between 0.10 and 0.20 U cm2 are considered typical [3]. However, values as high as 0.24 U cm2 at 70  C [14] and even 0.363 U cm2 at 50  C [5] are reported for Naon 115. Moreover, values between 0.26567 and 0.45108 U cm2 e obtained through nonlinear regression analysis e are reported for different types of gas diffusion layers of a single cell operated at 80  C and 1.5 bar using humidied H2/air as reactants [15]. In our study, the membrane thickness is only 27 mm and, therefore, Rohmic is supposed to be at the lower range. But, since neither the anode nor the cathode gasses are humidied, higher values of Rohmic are obtained. The tted values of cc are also similar (w0.039 V) in the cases where mass transfer limitation is pronounced. This is also expected since, theoretically, c is constant for a given reaction at a given temperature e see Eqs. (5a) and (5b). The estimated values are closer to the theoretical values of c obtained from Eq. (5b) than

Predicted Cell Voltage, V

1.0

0.8

0.6

0.4 0.4

0.6

0.8

1.0

0.4

0.6

0.8

1.0

0.4

0.6

0.8

1.0

Experimental Cell Voltage, V

Fig. 2. The predicted cell voltages compared to the actual experimental cell voltages at 40 1.5  C for different air ow rates: (a) 2.39 slm, (b) 2.87 slm, and (c) current-dependent.

S. Haji / Renewable Energy 36 (2011) 451e458

455

5.5
Calculated Assumed Average Linear Fit

5.0 4.5

iloss , mA/cm2

4.0 3.5 3.0 2.5 2.0 1.5 0.00

0.05

0.10

0.15

0.20

0.25

i, A/cm2
Fig. 3. The calculated current loss, iloss, as a function of current drawn. Conditions: 40 1.5  C and 2.39 slm air ow. The assumed (initial) and average current loss and a linear t to the data are also shown.

to that obtained by Eq. (5a) e see the last two columns of Table 1. This suggests the necessity of considering the effect of the mass transfer limitation on the reaction rate when calculating the concentration loss (see the discussion on Eq. (5) in Section 1). In effect, the estimated values of cc are expected to be even higher than the values calculated by Eq. (5b), since the theoretically derived equation does not account for the accumulation of the products and inert gases, which are water and nitrogen, respectively, in our case. Fig. 2 shows the predicted cell voltages compared to the measured cell voltages for the three discussed cases. The current loss, iloss, in the model is assumed to be constant for a given operating temperature over the whole range of the drawn current density. The validity of this assumption is investigated by plotting the calculated iloss versus the drawn i for 0 i iL,c, as illustrated in Fig. 3. The data are for one of the cases in Fig. 1, specically the case where the air ow rate is 2.39 slm. The current

loss is calculated from the following equation: iloss itheoretical i where the theoretical current density, itheoretical, is the current that will be generated if all the consumed hydrogen is reacted to produce useful external current. The theoretical current density is calculated from the measured hydrogen ow rate. Fig. 3 shows that iloss does not remain at a constant value throughout the experimental run. It starts with a value (4.7 mA/cm2), decreases in the mid-range of i, then increases again towards iL,c. But the overall trend shows slight decrease in the iloss value with an average of 3.54 0.71 mA/cm2. Nevertheless, iloss is considered to be constant and to equal the initial value of iloss for the model regression analysis. The initial value is chosen since iloss values are close to the initial value at the regions where it has more signicant effects, i.e. at current densities close to zero and the limiting current density. It might occur to the mind that iloss would be negligible when operating close to iL. However, this might be the case only when iL is reached due to H2 depletion in

0.95

0.85 314 K 306 K 298 K 294 K

Cell Voltage, V

0.75

0.65

0.55

0.45 0 0.05 0.1 0.15 0.2 0.25 Current Density, A/cm2 0.3 0.35 0.4

Fig. 4. The ieV curves for different operating temperatures (T 1.5 K) and xed air ow rate (14.34 slm) where the concentration loss is not pronounced.

456

S. Haji / Renewable Energy 36 (2011) 451e458

0.95

0.85

Cell Voltage, V

0.75

0.65

0.55

0.45 0.001

0.010

0.100

1.000

i +i loss

, A/cm2

Fig 5. Cell voltage versus log-scaled gross current showing the linear behavior of the curve, according to Tefel equation, at low current densities where the activation loss dominates. T 298 K; Air ow rate 14.34 slm.

which case its concentration is very small and permeation rate is retarded. In this study, however, iL is caused by O2 depletion in the cathode. Therefore, in this study the term (i iloss) is included in the cathodic concentration loss term instead of only using i. The effect of temperature on the polarization curve is also studied and the results are shown in Fig. 4. In these experiments the fuel cell is subjected to positive load increments only and it is run under sufcient air ow rate to minimize the concentration loss effect. As expected, the fuel cell performs better at higher temperatures and this is attributed mainly to faster reaction kinetics, which lead to lower activation overpotential. Additionally, the effect of temperature on the cathode exchange current density, i0,c, and current loss, iloss, is studied. At different temperatures, iloss is measured and i0,c is calculated from the Tafel parameters which are obtained from the i-V curves. The values of the Tafel parameters are estimated using linear regression for the data obtained in the activation overpotential dominated region only, i.e. at very small current densities where the other types of losses are negligible. For example, for the illustrated case in Fig. 1, where sufcient air ow rate is supplied, at i 0.02 A/cm2, the activation loss contribution towards the total losses is around 95.2%. More data are acquired in the low i region to perform the linear regression analysis. Although the analysis region is bounded to a maximum current density of 0.02 A/cm2, gures such as the one shown in Fig. 5 are plotted to insure that the studied region is located in the activation loss dominant region. Fig. 5 shows the polarization curve for one of the runs in Fig. 4 (T 298 K), on a rectangular-log scale plot, where the cell voltage is plotted versus the gross current density (igross i iloss). In the low current density region, the curve shows linear behavior, as expected from

the Tafel equation e Eq. (3). This proves that up to current densities of 0.02 A/cm2, the region is indeed dominated by activation overpotential. Despite that the ohmic loss is negligible in the low current density region, since the area specic resistance of the membrane is known, the regression is conducted between the iRohmic corrected voltage and the gross current density. The dependent variable is considered to be V Ethermo iRohmic while the independent variable is ln (igross). The three estimated values of Rohmic at 40  C are averaged and used for all the runs assuming negligible temperature effect. The average Rohmic is calculated to be 0.3646 U cm2. The regression parameters are listed in Table 2. At 25  C, the i0,c is estimated to be 3.4937 105 A/cm2. This value is higher than that reported by Amphlett et al. [16], which is 1.0764 106 A/cm2, at the same temperature. At 45  C, the estimated i0,c is of the same order of magnitude of what is considered a typical value by Larminie and Dicks [2] for a low temperature and ambient pressure H2/air fuel cell. As indicated from the values given in Table 2, the exchange current density increases as the reaction temperature increases. Similar behavior is shown by Santarelli et al. [10] and Parthasarathy et al. [17]. The increase in temperature enhances the rate of reaction including the forward and reverse rates of reactions at equilibrium and, hence, i0 increases. It is reported that the i0 increases exponentially with temperature [3,9], which is reasonable since the exchange current density resembles the rate of chemical reaction. OHayre [9] describes this effect by the following Arrhenius-like equation:

i0 AeDG

=RT

(9)

where A is a temperature independent constant and DG* is the activation energy. Fig. 6 shows the Arrhenius plot for the reduction

Table 2 The measured, tted, and the calculated Tafel equations related parameters at different temperatures and 14.34 slm air ow rate. Temperature T ( C) Measured parameter iloss (mA/cm2) 24.5 0.5 29.5 0.5 35 0.5 45 1.0 3.444 4.018 4.591 5.452 Fitted parametersa ac (V) 0.6962 0.0162 0.72042 0.0348 0.7036 0.0278 0.6777 0.0310 bc (V) 0.06785 0.003674 0.07482 0.00784 0.07622 0.00637 0.07764 0.00706 R2 0.9815 0.9195 0.9538 0.9674 Calculated parameters

ac
0.0945 0.0871 0.0871 0.0883

i0,c (A/cm2) 3.4937 105 6.5839 105 9.7960 105 1.6166 104

a The numbers following the tted parameters values represent the 95% condence intervals. The Tafel constants are based on Tafel equation written in the natural logarithmic form.

S. Haji / Renewable Energy 36 (2011) 451e458

457

-8.5

-9.0

-9.5

ln (i0,c) = 6844.7(1/T ) + 12.872 R = 0.9585

-10.0

-10.5 3.1E-03

3.2E-03

3.3E-03

3.4E-03

1/T, K-1
Fig. 6. Arrhenius plot for the oxygen reduction reaction on the cathode.

(ii) higher H2 concentration compared to that of O2 on the membrane ends, (iii) having pure H2 as a fuel which eliminates the possibility of any side reaction on the anode, (iv) the known high electrical resistivity of PEMFC membranes in general, and nally (v) the good gas-tight seal between the fuel cell components. The mathematical relations that describe the H2 crossover phenomena, _ as explained by Barbir [3], are as follows: iloss nF N H2 ;leak , C1 =T where _ N H2 ;leak Pm Ac P =d, Pm DH2 SH2 , and DH2 C2 e _ N H2 ;leak is H2 permeation rate through the membrane, Pm is H2 permeability through the membrane, Ac is the membrane crosssectional area, P is the gas pressure, DH2 is H2 diffusivity through the membrane, SH2 is solubility of H2 in the membrane (fairly independent of temperature), and C1 and C2 are the diffusivity related constants. For a given membrane and given operating pressure in the anode, combining the previous relations yields the following relation for iloss as a function of temperature:

ln (i0,c )

iloss C0 eC1 =T

(10)

reaction of oxygen, from which the activation energy of the reaction is calculated through tting a straight line to the data (The coefcient of determination (R2) is also shown in the gure). In future work, the cathode exchange current density should be studied under wider temperature range to conrm the linear relationship between ln (i0,c) and 1/T. From the slope, the activation energy barrier is estimated to be 56.9 kJ/mol for the reduction reaction of oxygen on the Pt-loaded electrode. Parthasarathy et al. [18] also used the Arrhenius plot to determine the activation energy of the oxygen reduction on the cathode when fed by pure oxygen at 5 atm. They reported different activation energies for different electrodes which ranged from 73 to 93 kJ/mol. Gasteiger et al. reported an activation energy of 66 kJ/mol [19]. The variation of current loss with temperature did not receive enough attention in the literature. In the different temperature experiments, the current loss is also measured at the beginning of each experiment. The current loss is measured for the temperature range of 20e45  C and is found to vary from 2.87 to 5.45 mA/cm2. It is clear from Table 2 that iloss increases with temperature. The current loss is typically considered to be 2e3 mA/cm2 [2,3]. Santarelli et al. [10] also reported on the increase of iloss with temperature. In their study, the value of the tted parameter, iloss, is estimated to be between 6 and 12 mA/cm2 for the temperature range of 50e80  C. In our study, the iloss is believed to be completely due to the permeation of H2 through the membrane. This assumption is based on the following ve reasons: (i) the higher permeability of H2 through the membrane compared to that of O2,
1.8

where C0 is a temperature independent constant, assuming SH2 is temperature independent. It should be noted that the above equation is only valid for cases in which the current loss is merely due to the hydrogen permeating through the membrane. Fig. 7 is a plot of ln (iloss) versus 1/T with a tted straight line. The two constants, C0 and C1, can be extracted from the tted equation shown in the gure. 4. Conclusion In this paper, the ieV data obtained from a 40-W fuel cell stack operated at 40  C with pure hydrogen and air are tted to an analytical model, where the voltage output is a function of the thermodynamic voltage and activation, ohmic, and concentration losses. The thermodynamic voltage is calculated based on Nernst equation. The Tafel intercept and slope, cell resistance, and the concentration loss constant are found by multiple linear regression analysis and are estimated to be 0.539 V, 43.2 mV, 0.365 U cm2, and 0.0389 V, respectively (average values). Although nonlinear regression analysis could provide mathematically acceptable tted parameters, the obtained parameters values are not reasonable in fuel cell operating terms. In contrast, linear regression results in better t in terms of the reasonability of the tted parameters; however, iL has to be estimated rst, which introduces a limitation to this approach. Therefore, its very important to refer to the typical values of the polarization curve parameters when performing regression. The effect of temperature on the current loss and the oxygen reduction exchange current density is also studied. In the temperatures range of 25e45  C, it is found that i0,c increases from 3.5 105 to 1.6 104 A/cm2, while iloss increases from 3.44 to 5.45 mA/cm2. Each of i0,c and iloss are tted to an exponential function of temperature and the related constants are estimated. The transfer coefcient of the oxygen reduction, ac, is found to be very small e less than 0.16. In conclusion, it is demonstrated in this study that a theoretically derived equation can reasonably model the ieV curve of a PEM fuel cell operating under steady-state. Acknowledgements

1.6

ln (i loss )

1.4

1.2

ln (iloss) = -2342.9(1/T ) + 9.0877 R = 0.9711

1.0 3.1E-03

The author is grateful for the nancial support of the College of Engineering at the University of Bahrain for equipping the laboratory with fuel cell systems.
3.4E-03

3.2E-03

3.3E-03

1/T, K-1
Fig. 7. The effect of temperature on the current loss: ln(iloss) versus 1/T. The data are tted to a straight line with the shown equation and R2.

References
[1] Prentice G. Electrochemical engineering principles. Houston: Prentice-Hall International; 1991.

458

S. Haji / Renewable Energy 36 (2011) 451e458 [12] Gorgun H, Arcak M, Barbir F. An algorithm for estimation of membrane water content in PEM fuel cells. J Power Sources 2006;157:389e94. [13] Hamelin J, Agbossou K, Laperriere A, Laurencelle F, Bose TK. Dynamic behaviour of a PEM fuel cell stack for stationary applications. Int J Hydrogen Energy 2001;26(6):625e9. [14] Squadrito G, Maggio G, Passalacqua E, Lufrano F, Patti A. An empirical equation for polymer electrolyte fuel cell (PEFC) behavior. J Appl Electrochem 1999;29:1449e55. [15] Xia ZT, Chan SH. Analysis of carbon-lled gas diffusion layer for H2/air polymer electrolyte fuel cells with an improved empirical voltageecurrent model. Int J Hydrogen Energy 2007;32:878e85. [16] Amphlett JC, Baumert RM, Mann RF, Peppley BA, Roberge PR. Performance modeling of the Ballard Mark IV solid polymer electrolyte fuel cell. J Electrochem Soc 1995;142:1e15. [17] Parthasarathy A, Supramaniam S, Appleby AJ, Martin CR. Temperature dependence of the electrode kinetics reduction at the platinum/Naon interfaceda microelectrode investigation. J Electrochem Soc 1992; 139:2530e7. [18] Parthasarathy A, Srinivasan S, Appleby AJ. Electrode kinetics of oxygen reduction at carbon-supported and unsupported platinum microcrystallite/ Naon interfaces. J Electroanal Chem 1992;339:101e21. [19] Gasteiger HA, Gu W, Makharia R, Matthias MF. Catalyst utilization and mass transfer limitations in the polymer electrolyte fuel cells. Orlando, FL: Electrochemical Society Meeting; 2003.

[2] Larminie J, Dicks A. Fuel cell systems explained. 2nd ed. England: Wiley & Sons; 2005. [3] Barbir F. PEM fuel cells. Massachusetts: Elsevier Academic Press; 2005. [4] Hirschenhofer JH, Stauffer DB, Engleman RR, Klett MG. Fuel cell handbook. 4th ed. U.S. Department of Energy; 1998. [5] Kim J, Lee SM, Srinivasan S, Chamberlin CE. Modeling of proton exchange membrane fuel cell with an empirical equation. J Electrochem Soc; 1995:2670e4. [6] Laurencelle F, Chahine R, Hamelin J, Agbossou K, Fournier M, Bose TK, et al. Characterization of a Ballard MK5-E proton exchange membrane fuel cell stack. Fuel Cells 2001;1(1):66e71. [7] Berning T, Djilali N. Three-dimensional computational analysis of transport phenomena in a PEM fuel cellda parametric study. J Power Sources 2003;124 (2):440e52. [8] Al-Baghdadi M. Modelling of proton exchange membrane fuel cell performance based on semi-empirical equations. Renewable Energy 2005;30: 1587e99. [9] OHayre R, Cha S, Colella C, Prinz BP. Fuel cell fundamentals. New York: Wiley & Sons; 2006. [10] Santarelli MG, Torchio MF, Cochis P. Parameters estimation of a PEM fuel cell polarization curve and analysis of their behaviour with temperature. J Power Sources 2006;159:824e35. [11] Springer TE, Zawodzinski TA, Gottesfeld S. Polymer electrolyte fuel cell model. J Electrochem Soc 1991;138(8):2334e42.