International Journal of Hydrogen Energy 32 (2007) 4418 – 4423 www.elsevier.com/locate/ijhydene
Optimization of parametric performance of a PEMFC Süleyman Kaytako˘glu ∗ , Levent Akyalçın ˙ Eylül Campus, 26555 Eski¸sehir, Turkey Department of Chemical Engineering, Anadolu University, Iki Received 26 March 2007; received in revised form 11 June 2007; accepted 23 June 2007 Available online 23 August 2007
Abstract In this study, the Taguchi method was applied to determine optimum working conditions in obtaining maximum power density of a PEMFC. Performance measure analysis was also followed by performing a variance analysis, in order to determine the optimum levels and relative magnitude of the effect of parameters. The optimum working conditions were found to be system pressure, 5 bar; flow rate ratio of H2 and O2 , 1/2; temperature of fuel cell, 75 ◦ C and temperature of humidifiers, 75 ◦ C. Under these conditions, the amount of maximum power density was predicted as 353 mW cm−2 by using experimental results obtained according to Taguchi’s orthogonal array (OA) L9 (34 ). Verification experiment was done for the same optimum conditions and maximum power density was observed as 379.64 mW cm−2 . According to the results of this optimization, it was seen that pressure and humidification temperature were the effective parameters. 䉷 2007 Published by Elsevier Ltd on behalf of the International Association for Hydrogen Energy. Keywords: PEMFC; Taguchi’s method; ANOVA
1. Introduction Proton exchange membrane fuel cell (PEMFC), operating at low temperatures and high power densities, is considered as being one of the most promising technology able to produce efficient and environmentally friendly energy for powering electrical vehicles [1]. The performance of PEMFC, being important and getting more and more attention in recent years, is known to be influenced by many parameters such as operating temperatures both fuel cell and humidifiers, pressure, flow rates and relative humidity of fuel and oxidant gases. In order to improve fuel cell performances, it is essential to understand these parametric effects on the fuel cell operations [2]. Other variables which are out of the operating parameters have also been studied; different membrane thickness and type, different catalyst load, path of the reactants flow morphology of the gas diffusion layer etc. However, these aspects, although very important, are not operating variables, i.e. they cannot be modified during the cell utilization [3]. In the open literature, it can be seen that a lot of studies, like given below, that deal with or discuss the parameters ∗ Corresponding author.
E-mail address:
[email protected] (S. Kaytako˘glu).
affecting the performance of PEMFC. Beattie et al. [4] studied temperature and pressure dependence of oxygen reduction at the platinum and Nafion interfaces. Cappadonia et al. [5] studied the conductance of Nafion 117 membranes as a function of temperature and water content. Sridhar et al. [6] performed humidification studies on PEMFC performances. Wang et al. [2] studied experimentally the effects of different operating parameters on the performance of PEMFC using pure hydrogen on the anode side and air on the cathode side. Hyun and Kim [7] studied experimentally the effect of external humidity on fuel cell performance. Hwang and Hwang [8] performed a parametric study of a double-cell stack of PEMFC. Ferng et al. [9] investigated analytically and experimentally the performance of PEMFC by taking into account the effects of different operating conditions, and the flow characteristics in the gas flow channel and cathode diffusion layers. Ticianelli et al. [10] studied experimentally to improve the performance of PEMFC. Santarelli and Torchio [3] reported a paper showing and discussing the results obtained after an experimental session devoted to characterization of the behavior of a single PEMFC with variation of different operation of both cell and humidifier in saturation and dry conditions, and pressure of reactant gases. The optimization of the operating parameters affecting the performance of PEMFC and obtaining related data are very
0360-3199/$ - see front matter 䉷 2007 Published by Elsevier Ltd on behalf of the International Association for Hydrogen Energy. doi:10.1016/j.ijhydene.2007.06.025
S. Kaytako˘glu, L. Akyalçın / International Journal of Hydrogen Energy 32 (2007) 4418 – 4423
4419
Nomenclature ei n nAi , nBi , nCi . . . n0 nr OA Se
the random error in ith experiment the number of rows in the matrix experiment the replication number for parameter level Ai , Bi , Ci , . . . equivalent sample size the number of repetition for verification experiment or experimental combination orthogonal array the two-standard-deviation confidence limit
important in various applications, and especially for fuel cell producers to validate and improve their models [2]. Therefore doing large numbers of experiments are often needed to understand clearly the effects of the parameters on the performance of PEMFC and to optimize them. It is known very well that performing large numbers of experiments of systematic experimental studies to optimize the performance of PEMFC are costly and time-consuming process. To overcome this challenge, Taguchi’s orthogonal array (OA) analysis, known as experimental design methods, may be used in order to evaluate the respective impacts of those parameters on the performance of PEMFC, and to reduce the number of experiments when many parameters are studied. The main advantage of this method over other statistical experimental design methods is that the parameters affecting an experiment can be investigated as controlling and none controlling. Details about the Taguchi’s OA analysis can be found elsewhere [11–14]. The advantage of the Taguchi method on the conventional experimental design methods, in addition to keeping the experimental cost at a minimum level, is that it minimizes the variation in product response while keeping the mean response on target. Its other advantage is that the optimum working conditions determined from the laboratory work can also be reproduced in the real application of PEMFC. Although there are numerous applications of Taguchi’s experimental design method from chemistry to engineering and from microbiology to agriculture, very little amount of researchers have been interested in optimization of operating parameters of PEMFC by using this technique. In this work, to ensure for the PEMFC proper and optimal operating conditions like the system pressure, flow rate of H2 and O2 , temperature of both fuel cell and humidifier leading to high power to be obtained, Taguchi’s experimental design method has been used. Results obtained both experimentally and theoretically are given and analyzed in this paper.
2. Material and methods A schematic drawing of the experimental apparatus employed in this study is shown in Fig. 1. In each run, pure hydrogen and oxygen were used as reacting gas in anode and cathode, respectively. Humidification of the reacting gases was maintained externally by using stainless steel bottles placed in
S/N Xi Yi
performance characteristics for larger-the-better the fixed effect of the parameter level combination used in ith experiment performance value of ith experiment
Greek letter 2e
the overall mean of performance value error variance
a water bath. Regulating the water bath temperature controls the humidification of the reactant gases. The gas connection between the gas control system and the fuel cell inlets are well insulated to prevent condensation of the water vapor on the way to the fuel cell. A single PEMFC made of stainless steel type 316 on house with active surface area of 5 cm2 was used for all run. This active area was obtained by grooving gas channels having 1 mm depth and 1 mm width in serpentine shape flow field on the center of the each stainless steel 316 end plates. To eliminate the effects of the membrane electrode assembly (MEA) on the performance of PEMFC, a standard MEA consists of a Nafion 115 membrane in combination with platinum loadings of 0.5 mg cm−2 per electrode and gas diffusion layers made of carbon cloth was purchased from E-TEK and used. This MEA is placed between two stainless steel end plates, and squeezed tightly via eight bolts around the active areas. In this study, a fuel cell test station made on house is used. This station includes a computer-based control and data acquisition system through a computer equipped with Labview䉸 -based application software. In this test station, fuel cell and water bath temperatures are controlled by a microprocessor-based temperature/process controller (Meter Evo 04). Control of flow rates of the reacting gases were maintained via mass flow controllers (Brooks 5850S) located before the humidifiers. Pressures of the anode and cathode sides are controlled by back pressure regulators (Brooks 5866). The mass flow rates of reacting gases are set and read through the software. The fuel cell polarization curves are obtained from this program as well by controlling the Electronic Load (Agilent N3304), which measures the voltage vs current response of the fuel cell. Experimental parameters and their levels given in Table 1 are determined in the light of literature and preliminary tests. The L9 (34 ) OA was accepted as the most proper method to determine the experimental plan, for four parameters of each three values given in Table 2 [11–14]. The performance of PEMFC can be affected by some factors known as controllable or uncontrollable (noise sources). In order to observe the effects of uncontrollable factors on this process, each experiment was repeated five times with same conditions. Performance characteristics selected to be the optimization criteria are divided into three categories, the larger-the-better, the smaller-the-better and the nominal-the-best. The first of them was calculated by using [13].
4420
S. Kaytako˘glu, L. Akyalçın / International Journal of Hydrogen Energy 32 (2007) 4418 – 4423
1. Mass and Pressure Controller 2. Electronic Load 3. Water Bath 4. Humidifier 5. Mass Flow Control Valve 6. Pressure Gauge 7. Back Pressure Controller 8. Chiller 9. Condenser 10. Temperature Controller 11.Fuel Cell 12. Purgemeter
6 12
6
10
5 12
4
4
3
5 1
10
+
-
11
2
7
10 O2
H2
N2
9
7
9
8
Fig. 1. Schematic representation of experimental apparatus.
Larger-the-better: S/N = −10 log10
nr 1 1 2 nr Y i=1 i
.
(1)
The values that make S/N maximum are optimum if the purpose of a process is to reach the maximum power density value at a certain cell potential. The experiment corresponding to optimum working conditions in obtaining the maximum power density might not be found in planned experimental plan table of the Taguchi method. In such cases, the performance value for optimum conditions in obtaining the maximum power density can be predicted by using the balanced characteristic of OA. For this purpose, an additive model given below can be used [14]: Yi = + Xi + ei .
(2)
Since Eq. (2) is point estimation, which is calculated by using experimental data in order to determine whether the additive model is adequate or not, the confidence limits for the prediction error must be evaluated [15]. The prediction error is the difference between the observed Yi and the predicted Yi . The confidence limits for the prediction error, Se, is 1 2 1 Se = ±2 + , (3) n0 e nr 2e sum of squares due to error , (4) = degrees of freedom for error 1 1 1 1 1 1 1 1 + + + ···. = + + + − n0 n nAi n nBi n nCi n (5)
2e
If the prediction error is outside of these limits, it is ought to be suspected of the possibility that the additive model is not convenient. Otherwise, it can be considered that the additive model is convenient.
Table 1 Parameters and their values corresponding to their levels studied in experiments Parameters
A B C D
Levels
Pressure of the fuel cell (bar) Flow rate ratios of H2 and O2 Temperature of the fuel cell (◦ C) Temperature of the humidifiers (◦ C)
1
2
3
1 1/1 70 70
3 2/1 75 75
5 1/2 80 80
Table 2 Experimental plan table according to L9 (34 ) Expt. no.
A
B
C
D
1 2 3 4 5 6 7 8 9
1 1 1 2 2 2 3 3 3
1 2 3 1 2 3 1 2 3
1 2 3 2 3 1 3 1 2
1 2 3 3 1 2 2 3 1
A verification experiment is a powerful tool for investigating the presence of interactions among the control parameters. If the predicted response under the optimum conditions in obtaining the maximum power density does not match the observed response, then it implies that the interactions are important. If the predicted response matches the observed response, then it implies that the interactions are probably not important and that the additive model is a good approximation [14]. In the present work, the order of the experiments was obtained by inserting parameters into the columns of OA, L9 (34 ), selected to be the experimental plan given in Table 2. The order of experiments
S. Kaytako˘glu, L. Akyalçın / International Journal of Hydrogen Energy 32 (2007) 4418 – 4423
is randomized in order to avoid noise sources which had not been considered initially and which could take place during an experiment and affect results in a negative way. 3. Results and discussion
4421
five times under same working conditions and MEA. Maximum power density was obtained at 0.4 V at each run. Results, consisting of average of the maximum power densities, were presented in Fig. 2 as potential versus current density and power density versus current density curves. Power densities at 0.4 V cell potential obtained from results were tabulated in Table 3.
3.1. Polarization and power density curves 3.2. Statistical analysis Order of the experiments, in Table 2, were randomized and then performed by using the experimental set-up given in Section 2. Since system conditioning took approximately two hours, polarization scans were started thereafter. In each run, performed galvanostatically, the polarization scans were taken 1.2
OA1 OA2 OA3 OA4 OA5 OA6 OA7 OA8 OA9
Potential, V
1 0.8
The Minitab14䉸 software was used to analyze the collected data. A variance analysis was performed in order to see effective parameters and their confidence levels in obtaining the maximum power density of PEMFC. A statistical analysis of variance (ANOVA) was performed to understand whether the process parameters are statistically significant or not. F -test is a powerful tool to observe which process parameters have Table 4 Result of the variance analysis for the maximum power density value of experiment
0.6 0.4 0.2 0 0
200
400 600 800 Current Density,mA cm-2
1000
1200
A B C D Error Total
Degree of freedom
Sum of squares
Average of squares
F
p
2 2 2 2 36 44
190 882 1745 269 9360 13346 215 603
95 441 873 135 4680 371
257.44 2.35 0.36 12.62
0.000 0.109 0.698 0.000
OA1 OA2 OA3 OA4 OA5 OA6 OA7 OA8 OA9
350 300 250
Gas Flow Rate Ratios
Pressure
200
Temperature Temperature of Fuel Cell of Humidifiers
50.0
150
S/N Ratio
Power Density, mA.cm-2
400
100 50 0 0
200
400
600
800
1000
1200
Current Density. mA.cm-2
48.5 47.0 45.5 44.0 1
Fig. 2. Polarization and power density curves recorded for the conditions given in Table 2.
2
3
1
2
3
1
2
3
1
2
3
Fig. 3. The mean effects plot for S/N ratios.
Table 3 Results obtained at cell potential of 0.4 V for the conditions given in Table 2 Expt. no.
Current (mA)
1 2 3 4 5 6 7 8 9
2188 2205 2240 3250 3300 3635 4200 3548 4034
2230 2380 1750 3259 3345 3314 4487 3660 4263
Power density (mW cm−2 )
2134 2550 1840 2945 3245 3370 4340 3640 4102
1360 1580 1907 3084 2900 3740 4416 3950 4075
2140 2500 2177 2963 2670 3513 4295 3830 4118
175.0 176.4 179.2 260.0 264.0 290.8 336.0 283.8 322.7
178.4 190.4 140.0 260.7 267.6 265.1 359.0 292.8 341.0
170.7 204.0 147.2 235.6 259.6 269.6 347.2 291.2 328.2
Average power density (mW cm−2 ) 108.8 126.4 152.6 246.7 232.0 299.2 353.3 316.0 326.0
171.2 200.0 174.2 237.0 213.6 281.0 343.6 306.4 329.4
160.8 179.4 158.2 248.0 247.4 281.2 347.8 298.1 329.5
4422
S. Kaytako˘glu, L. Akyalçın / International Journal of Hydrogen Energy 32 (2007) 4418 – 4423
Table 5 Result obtained from verification experiment for the A3 , B3 , C2 and D2 condition
1
4874
4878
Power density (mW cm−2 ) 4755
4610
4610
389.9
a significant effect in obtaining the maximum power density. The F -value for each process parameter is simply a ratio of the mean of the squared deviations to the mean of squared error. Generally, the larger the F -value, the greater the effect on obtaining the maximum power density because of change of the process parameter. The optimal combination of process parameters can be predicted together with the performance characteristics and ANOVA analyses. The results of variance analysis for the experiments are given in Table 4. The larger-the-better performance characteristic, Eq. (1), has been taken in obtaining the maximum power density of PEMFC. The order of graphs in Fig. 3 prepared for the experiments is according to the degrees of the influences of parameters on the performance characteristics. The optimal level of a process parameter in obtaining the maximum power density is the level with the highest S/N value calculated by Eq. (1). Fig. 3A shows the variation of performance characteristics with pressure. In order to determine the experimental conditions for the first data point, A for that point is level 1, which is 1 bar for this parameter. The experiments for which A Level (column A) is 1 are experiment nos. 1, 2 and 3. The performance characteristics value of the first data point is thus the average of those obtained from experiment nos. 1, 2 and 3. All the points in Fig. 3A graph and other graphs are obtained by using the same way. The numerical value of the maximum point in each graph is corresponding to the best value for that parameter. These values are seen to be A3 (5 bar), B3 (1/2), C2 (75 ◦ C) and D2 (75 ◦ C) (Table 5). If the experimental plan given in Table 2 is examined carefully together with parameter values given as maximum power density conditions, it can be seen that experiments corresponding to optimum for maximum power density conditions have not been carried out during the experimental work. The results of variance analysis for the experiments are given in Table 4. Thus, it should be noted that the maximum power densities given in Table 6 are predicted results by using Eq. (2) and the observed results obtained from verification experiment for the same conditions. The results given in Table 6 are also in between confidence limits of predictions. In order to test the predicted results, verification experiments were carried out at the same working conditions. The fact that the power densities from verification experiments are within the calculated confidence intervals calculated from Eqs. (3)–(5) (see Table 6) shows that the experimental results are within a ±5% error range (Fig. 4). This case states that there is a good agreement between the predicted values and experimental values, and the interactive effects between the parameters are indeed negligible. It may be concluded that the additive model is adequate for
390.3
Average power density (mW cm−2 )
380.4
368.8
368.8
379.64
Table 6 Optimum working condition, and observed and predicted maximum power densities Parameter
Value
Level
Pressure (bar) Flow rate ratios of H2 to O2 Temperature of fuel cell (◦ C) Temperature of humidifier (◦ C) Observed maximum power density (mW cm−2 ) Predicted maximum power density (mW cm−2 ) Confidence limits of prediction for Maximum power density (mW cm−2 )
5 1/2 75 75 379.64 353 299.17–406.91
3 3 2 2
450
1.2 Polarization Curve Powers Denity
1
400 350 300
0.8
250
0.6
200 150
0.4
100
0.2
Power Density, mW.cm-2
Current (mA)
Potential,V
Expt. no.
50
0 0
200
400 600 800 1000 Current Density,mA.cm-2
1200
0 1400
Fig. 4. Polarization and power density curves obtained from verification experiment.
describing the dependence of the performance of this PEMFC on the various parameters [14]. 4. Conclusion The optimization of the operating parameters affecting the performance of PEMFC and obtaining related data are very important in various applications, and especially for fuel cell producers to validate and improve their models. In the present study, Taguchi method has been used to determine the optimum working conditions for maximum power density of a PEMFC. The OA L9 (34 ) technique is described for experimental design as it reduces the number of experiments required to investigate a set of parameters and to minimize time and cost while performing experiments. Experimental investigations into the parameter effects have allowed to determine the optimum configuration of design parameters for maximum power density.
S. Kaytako˘glu, L. Akyalçın / International Journal of Hydrogen Energy 32 (2007) 4418 – 4423
Results can be summarized as follows: • The effective parameters on maximum power density from PEMFC are pressure, temperature of humidifiers, flow rate ratios of H2 to O2 and temperature of fuel cell. • Optimum conditions within the selected parameter values are 5 bar for pressure, 75 ◦ C for humidifiers temperature, 1/2 for flow rate ratios of H2 to O2 and 75 ◦ C for fuel cell temperature. Under these conditions, maximum power density was obtained to be 379.64 mW cm−2 from verification experiment. • Predicted and obtained maximum power densities are close to each other. It may be concluded that the additive model is adequate for describing the dependency of obtaining maximum power density on various parameters. • Since optimum conditions determined by Taguchi method in a laboratory scale PEMFC is reproducible in large scale PEMFC as well, findings of present study may be very useful for PEMFC stack applications.
Acknowledgement Special thanks go to the Scientific Research Project Commission of the Anadolu University for financial support of the present work under project no: 020218. References [1] Wahdame B, Candusso D, Kauffmann JM. Study of gas pressure and flow rate influences on a 500 W PEM fuel cell, thanks to the experimental design methodology. J Power Sources 2006;156(1):92–9.
4423
[2] Wang L, Husar A, Zhou T, Liu H. A parametric study of PEM fuel cell performances. Int J Hydrogen Energy 2003;28(11):1263–72. [3] Santarelli MG, Torchio MF. Experimental analysis of the effects of the operating variables on the performance of a single PEMFC. Energy Convers Manage 2007;48(1):40–51. [4] Beattie PD, Basura VI, Holdcroft S. Temperature and pressure dependence of O2 reduction at Pt|Nafion䉸 117 and Pt|BAM䉸 407 interfaces. J Electroanal Chem 1999;468(2):180–92. [5] Cappadonia M, Erning JW, Niaki SMS, Stimming U. Conductance of Nafion 117 membranes as a function of temperature and water content. Solid State Ionics 1995;77:65–9. [6] Sridhar P, Perumal R, Rajalakshmi N, Raja M, Dhathathreyan KS. Humidification studies on polymer electrolyte membrane fuel cell. J Power Sources 2001;101(1):72–8. [7] Hyun D, Kim J. Study of external humidification method in proton exchange membrane fuel cell. J Power Sources 2004;126(1–2):98–103. [8] Hwang JJ, Hwang HS. Parametric studies of a double-cell stack of PEMFC using GrafoilTM flow-field plates. J Power Sources 2002;104(1):24–32. [9] Ferng YM, Tzang YC, Pei BS, Sun CC, Su A. Analytical and experimental investigations of a proton exchange membrane fuel cell. Int J Hydrogen Energy 2004;29(1):381–91. [10] Ticianelli EA, Derouin CR, Srinivasan S. Localization of platinum in low catalyst loading electrodes to attain high power densities in SPE fuel cells. J Electroanal Chem 1988;251:275–95. [11] Roy RK. Design of experiments using the Taguchi approach. New York: Wiley; 2001. [12] Logothetis N, Wynn HP. Quality through design. Oxford: Clarendon Press; 1989. [13] Ross JP. Taguchi techniques for quality engineering. New York: McGrawHill; 1988. [14] Phadke MS. Quality engineering using robust design. Englewood Cliffs, NJ: Prentice-Hall; 1989. [15] Levent M, Kaya Ö, Kocakerim M, Yi˘git V, Küçük Ö. Optimization of desulphurization of Artvin–Yusufeli lignite with acidic hydrogen peroxide solutions. Fuel 2007;86(7–8):983–92.
