Online humidification diagnosis of a PEMFC using a static DC DC converter

international journal of hydrogen energy 34 (2009) 2718 2723 Available at www.sciencedirect.com journal homepage: www.elsevier.com/locate/he Online humidification diagnosis of a PEMFC using a static DC DC converter M. Hinaje*, I. Sadli, J.-P. Martin, P. Thounthong, S. Raël, B. Davat GREEN Institut National Polytechnique de orraine 2, Avenue de la Forêt de Haye, 54516 Vandœuvre-lès-Nancy, France article info Article history: Received 28 November 2008 Received in revised form 22 January 2009 Accepted 22 January 2009 Available online 23 February 2009 Keywords: Proton exchange membrane fuel cell Boost converter Diagnosis Humidification Membrane resistance abstract This paper deals with the online checking of the humidification of a Proton Exchange Membrane Fuel Cell (PEMFC). Indeed, drying or flooding can decrease the performance of the PEMFC and even lead to its destruction. An online humidification diagnosis can allow a real-time control. A good indicator of the membrane humidification state is its internal resistance. As known, the membrane ionic conductivity increases with the membrane water content. This resistance can be calculated at high frequency by dividing the voltage variation by the current variation. The proposed scheme makes use of measurements of current and voltage ripples coming from the association of a static DC DC converter and the fuel cell. The experiment thus consists in computing the internal resistance in wet and dry conditions. ª 2009 International Association for Hydrogen Energy. Published by Elsevier td. All rights reserved. 1. Introduction The proton exchange membrane fuel cell (PEMFC) offers some advantages to other fuel cells as relative simplicity design and low operating temperature. An appropriate humidity condition not only can improve the performances and efficiency of the fuel cell [1], but can also prevent irreversible degradation of internal composition such as the catalyst or the membrane. Indeed, little or no water leads to membrane drying, which decreases the membrane ionic conductivity [2], and thus increases voltage drop across the membrane. Conversely, too much water causes flooding, the pores in the electrodes are filled with water and the transport of reactant gases to the catalyst site is thus obstructed. To have an adequate hydration of the PEM at each time, a real-time control of humidification state is needed. A good indicator of the humidification state is the membrane resistance [3,4]. This one can be measured by adding a high frequency component to the main fuel cell current. Indeed, it is well known that the small signal electrical behavior of a PEMFC can be represented, as in Fig. 1, by the Randles equivalent scheme [5]. This representation is a common and practical way of modeling an electrochemical cell. It consists of three resistors, r m standing for ohmic resistance and r a, r c standing for anode and cathode charge transfert resistances, respectively, due to the hydrogen oxidation and oxygen reduction. C dl,a and C dl,c are the double-layer capacities at the electrode/membrane interfaces. Finally, Z w,a and Z w,c are the diffusion impedances associated to the gas diffusion in the anode and the cathode, respectively. At high frequency, this scheme can be reduced to the ohmic resistance r m. Since the electronic and contact resistances contribution is negligible compared to the membrane resistance, the ohmic losses are often assumed to result in the membrane resistance. Therefore, this resistance * Corresponding author. Tel.: þ33 383 59 56 64; fax: þ33 383 59 56 53. E-mail address: melika.hinaje@ensem.inpl-nancy.fr (M. Hinaje). 0360-3199/$ see front matter ª 2009 International Association for Hydrogen Energy. Published by Elsevier td. All rights reserved. doi:10.1016/j.ijhydene.2009.01.076

international journal of hydrogen energy 34 (2009) 2718 2723 2719 anode membrane cathode C dl,a C dl,c r a Z w,a r m r c Z w,c Fig. 1 Equivalent circuit of an electrochemical cell. can be deduced at high frequency, dividing the voltage variation by the current variation: r m ¼ Dv Di HF Experimentally, frequency of the alternative current component, which enables r m measurement, is chosen so that the phase shift between voltage and current is null [6]. In this paper, we present a measurement method for r m that does not need any additional device for generating the high frequency current component. Indeed, as fuel cells are low voltage generators, it is most often necessary to use a power electronic converter, in order to increase fuel cell output voltage. The current ripple, naturally produced by this converter, enables r m measurement if the switching frequency is high enough. We obtain that way a real-time control of the fuel cell humidification state [7]. Hence, in this lecture, we propose an additional use of the converter associated with the fuel cell. In the following parts, the test bench is described, then experimental results are detailed and discussed. 2. Fuel cell and converter unit 2.1. Fuel cell experimental setup The considered stack is a 500 W proton exchange membrane fuel cell constituted of 23 cells, active area 100 cm 2. Fig. 2 (1) presents a sketch of the experimental setup. As can be seen in this figure, air is supplied through a humidification unit to cathode, and pure dry hydrogen from cylinders to anode. The humidifier can be shut down by opening the solenoid valve v 1 so that dry air flows towards fuel cell cathode. This will be used to vary the humidification state of the fuel cell. A cooling system keeps the stack temperature constant. Data acquisition cards are used for all necessary control functions such as reference setting (gas flows, pressures, and stack temperature). Fig. 3 shows the impedance spectrum of our 500 W PEMFC at 30 A, under a stack temperature of 55 C and a humidifier temperature of 50 C at atmospheric pressure. This experimentation was carried out by means of usual impedance spectroscopy method. The intersection with the real axis is obtained at 2 khz, it gives the membrane resistance value, in this case r m ¼ 40 mu. 2.2. Boost converter operating principle As it has been written before, fuel cells are low voltage generators, so that it is necessary to connect them to a converter [8]. The DC DC power electronic converter used in the test bench [9,10] is a standard boost chopper. As depicted in Fig. 4, it is composed of an input inductor, a power semiconductor switch denoted S, a diode D and an output capacitor C. State variables of this power electronic system are the input current i FC and the DC link voltage v OUT. Fig. 2 500 W PEMFC (ZSW, Germany) connected to a DC DC converter.

2720 international journal of hydrogen energy 34 (2009) 2718 2723 Im(Z) R m 0.05 0.1 0.15 0.2 Re(Z) -0.02 2 khz -0.04 Fig. 3 PEMFC impedance spectrum (50 mhz 2 khz). Mean current: 30 A. In normal operating mode, the boost chopper has two working sequences. The first is defined by the switch on-state. It is driven by the switch turn-on, and is characterized by the following state equations system: 8 di FC dt ¼ v IN V T þ r,i >< FC dv OUT ¼ i (2) >: OUT dt C where V T is the on-state voltage drop of the switch, and r the series resistance of the inductor. If the electric time constant /r of the inductor is far greater than the switching period T of the converter, this first sequence leads to a linear increase of the inductor current. The second sequence is defined by the diode on-state. It is driven by the switch turn-off, and is characterized by the following state equations system: 8 >< >: di FC dt ¼ v IN v OUT dv OUT ¼ i FC i OUT dt C V D þ r,i FC where V D is the on-state voltage drop of the diode. This sequence results in a linear decrease of the inductor current (provided that /r >> T ). In steady state, the input current is periodic, so that the mean value of di FC /dt is necessarily equal to zero. That way, one can easily calculate the mean output voltage V OUT,by integrating di FC /dt equations of system (2) and (3) over a period of operation. If the ideal conversion (no losses, i.e. V T ¼ 0, V D ¼ 0 and r ¼ 0) is considered, this results in the following function of the mean input voltage V IN : (3) Fig. 5 High frequency fuel cell model used for simulating the fuel cell/boost converter association. V OUT ¼ V IN 1 a a being the duty cycle of the switch (ratio of the switch conduction duration to the switching period T ). As a [0, 1], it can be deduced that V OUT V IN : the boost converter enables voltage increase, which is an advantage when using fuel cells as power sources. The second great advantage of the boost converter is that its input current is a state variable of the system, and thus can be regulated with an usual feedback loop. It is of great interest for fuel cells, as far as fuel cell current has to be carefully controlled. 2.3. Fuel cell/boost chopper association In our low voltage application (V OUT ¼ 42 V), unipolar semiconductor devices are used, in order to reduce conduction and turn-off losses. Therefore, a power MOSFET STE180NE10 (100 V, 180 A, 4.5 mu) and a Schottky diode STPS80H100TV (100 V, 2 40 A, 2.2 mu) were chosen for switch S and diode D, respectively. Since the switching frequency of the converter is 25 khz, the input inductor was calculated in order to obtain about 10% current ripple at the rated operating point of our fuel cell system (500 W, 12.5 V, 40 A). It was made using soft ferrite cores ETD59-3C90. The inductance value is 100 mh, and the series resistance r is 30 mu. The output capacitance is 30 mf, which is computed from the capacitor rms current (about 22 A for our application) at rated operating point. To this end, we used three aluminium electrolytic components of 10 mf (100 V, 7.3 A) in parallel. The fuel cell current i FC is regulated by a PID controller. The whole system (power electronics structure, parameters of passive elements and semiconductor devices, regulation (4) i FC D i OUT v IN S C v OUT fuel cell boost chopper load Fig. 4 Boost chopper.

international journal of hydrogen energy 34 (2009) 2718 2723 2721 Fig. 8 Relative humidity of inlet air. Fig. 6 Simulated fuel cell current. loops) had been implemented in a circuit simulation software. The fuel cell was represented by a high frequency model (diffusion phenomena are considered through an instantaneous concentration overpotential, included in the activation overpotential), composed of a voltage source standing for open circuit voltage (OCV), a current-dependent resistance associated with activation overpotential, the double-layer capacitor, the membrane resistance, and a series inductor (cf. Fig. 5). Fig. 6 presents the simulated waveform of the fuel cell current obtained in steady state, the fuel cell current reference being set to 30 A, and the load being adjusted so that DC link voltage is 42 V. These results were achieved for the following simulation parameters: OCV ¼ 23 V, activation losses ¼ 8.3 V, C D ¼ 20 mf, r m ¼ 40 mu, l FC ¼ 35 nh. The current ripple is 3.5 A peak-to-peak, which corresponds to the experimental value (cf. Fig. 10). As depicted by. Schindele et al. [11], one can take benefit of the high frequency triangular waveform of input current to measure the fuel cell membrane resistance. However, an accurate measurement requires to set the switching frequency to the value for which the fuel cell exhibits a fully ohmic behavior (2 khz for our stack). The aim of this study is to present an additional use of the fuel cell converter, as a diagnosis device of the membrane humidification state. All we need is a qualitative measurement of the membrane resistance, carried out without modifying the switching frequency for which the converter has been designed. To this end, a higher frequency than the theoretical value required for accurate measurement of membrane resistance can be used, provided that the inductive effect is small enough compared to the resistance. To highlight this, the computed fuel cell voltage linked to the current waveform drawn in Fig. 6, is plotted in Fig. 7, and is to be compared with experimental voltage waveforms presented in Fig. 9. The magnitude of the voltage ripple is equal to 156 mv (instead of 140 mv experimentally obtained, as can be seen in Fig. 10). With such a value, the membrane resistance is estimated to 44 mu, which is 10% higher than the theoretical value. The difference is of course due to the inductive part in fuel cell voltage ripple. Nevertheless, as shown by experimental results presented in the next section, it is low enough compared to the membrane resistance, and it will not hide the increase of membrane resistance. Indeed, it has to be underlined that the diagnosis is based on the evolution of the membrane resistance due to alteration or changes in humidification conditions. 3. Experimental results All experimentations hereafter are carried out under a stack temperature of 55 C and a humidifier temperature of 50 Cat Fig. 7 Simulated fuel cell voltage. Fig. 9 Current and voltage waveforms at t [ 10 s and t [ 60 s.

2722 international journal of hydrogen energy 34 (2009) 2718 2723 Fig. 12 Mean fuel cell voltage under wet and dry conditions. Fig. 10 Magnitude ripples of the voltage and the current. atmospheric pressure. The fuel cell current is regulated by a PID controller. Its mean value is 30 A. The converter is loaded by a rheostat, set so that DC link voltage is 42 V. The experiment consists in shutting down the humidifier (see Fig. 8) by opening the solenoid valve v 1, so that dry air flows through fuel cell cathode. This state is maintained during 60 s. At each second, fuel cell mean voltage, and current and voltage ripple amplitudes are registered. The current and voltage waveforms, at t ¼ 10 s and t ¼ 60 s (humidifier shut down being taken as time origin), are shown in Fig. 9. As can be seen in this figure, the cutoff duration of the humidifier influences the voltage and current ripples. Indeed, the magnitude of the voltage ripple increases with the drying of the membrane (because of an increase of membrane resistance), whereas, the magnitude of the current ripple decreases, as a consequence of the decrease of the fuel cell mean voltage (cf. Fig. 12). The time evolution of voltage and current amplitudes is drawn in Fig. 10. The membrane resistance, plotted in Fig. 11, results from the ratio of the previous curves. We can notice that the resistance strongly increases when the humidifier is shut down; in wet condition this one is equal to 40 mu, and reached 64 mu after 60 s of dry air feeding. We can also remark that the voltage mean value decreases from 13.5 V to 10 V (see Fig. 12). Note that the experiment does not include flooding conditions, which would also cause Fig. 11 Internal resistance under wet and dry conditions. a voltage drop at constant current, the found way to distinguish between flooding and drying is to have an online checking of the internal resistance. Those results are, of course, well known, but what is new in this work, is the way to obtain an online evolution of the membrane resistance. Therefore, the advantage of our method is that nothing is added to the test bench, we only use the natural ripples due to the boost chopper to have a real-time diagnosis of the PEMFC. 4. Conclusions Water management is one of the most critical problems in PEM fuel cell operation. In this paper, a method for checking the humidification state of the membrane has been developed by exploiting the connection of a boost converter to the fuel cell. The method relies on the estimation of the internal resistance calculated from the voltage and current ripples. As this resistance increases in response to drying, its time evolution allows us to have a real-time control of fuel cell humidification state which can avoid destroying it. In short, an additional use of the boost converter as a diagnosis device of the PEMFC humidification has been highlighted. references [1] Hyun D, Kim J. Study of external humidification method in proton exchange membrane fuel cell. Journal of Power Sources 2004;126:98 103. [2] Okada T, Xie G, Meeg M. Simulation for water management in membranes for polymer electrolyte fuel cells. Electrochimica Acta 1998;43:2141 55. [3] Yan Q, Toghiani H, Wu J. Investigation of water transport through membrane in a PEM fuel cell by water balance experiments. Journal of Power Sources 2005;158:316 25. [4] Zawodzinski TA, Davey J, Valerio J, Gottesfeld S. The water content dependence of electro-osmotic drag in protonconducting polymer electrolytes. Electrochimica Acta 1995; 40:297 302. [5] Rubio MA, Urquia A, Dormido S. Diagnosis of PEM fuel cells through current interruption. Journal of Power Sources 2007;171:670 7. [6] Fouquet N, Doulet C, Nouillant C, Dauphin-Tanguy G, Ould- Bouamama B. Model based PEM fuel cell state-of-health monitoring via ac impedance measurements. Journal of Power Sources 2006;159:905 13.

international journal of hydrogen energy 34 (2009) 2718 2723 2723 [7] Macdonald JR. Impedance spectroscopy emphasing solid materials and system. John Wiley and Sons Editions; 1987. [8] iua G, Zhanga J, Sunb Y. High frequency decoupling strategy for the PEM fuel cell hybrid system. International Journal of hydrogen energy 2008;33:6253 61. [9] Thounthong P, Raël S, Davat B. Test bench of a PEM fuel cell with low voltage static converter. Journal of Power Sources 2006;153:145 50. [10] Sadli I, Thounthong P, Martin J-P, Raël S, Davat B. Behaviour of a PEMFC supplying a low voltage static converter. Journal of Power Sources 2006;156:119 25. [11] Schindele, Späth H, Scholta J. PEM-FC control using power-electronic quantities. In: Twenty-first worldwide international battery, hybrid and fuel cell electric vehicle symposium and exhibition. Monaco; April 2005.