PEMFC Fault Diagnosis, Modeling, and MitigationAbraham Gebregergis, IEEE Member

Delphi Steering SaginawSaginaw, MI 48601

abraham.g.gebregergis@delphi.com

Pragasen Pillay, IEEE FellowDepartment of Electrical and Computer

Engineering, Concordia UniversityMontreal, Quebec, Canadapillay@encs.concordia.ca

Raghunathan RengaswemyDepartment of Chemical Engineering

Clarkson UniversityPotsdam, NY 13699–5707

raghu@clarkson.edu

Abstract—The paper introduces fault diagnosis and separation,mitigation, and modeling of a proton exchange membrane fuelcell (PEMFC). Experimental tests of a single PEMFC cell wereperformed during this study. Flooding and drying faults wereimplemented to be detected from the cell voltage and impedanceresponse of the cell. The impedance response at low frequencywas used to identify the cause of the fault. The slope ofthe magnitude and/or the negative phase response of the cellimpedance at low frequency were observed to allow separation offault. A cell impedance model based on a resistive capacitive (C-model) and resistive constant-phase-element (CPE-model) circuitsare developed. The CPE-model has a better approximation of thecell impedance. However, the C-model is ease to implement sinceit is well known in most simulation tools (Matlab/Simulink orPspice). A power electronic control is designed and tested tomitigate the faults. Pulsing the cell current at low frequency wasseen to increase the cell power by 8% during drying.

Index Terms—PEMFC, Fuel cell, modeling, Simulation, fault.

I. INTRODUCTION

Fuel cells are a promising energy technology with the ad-vantages of high efficiency and low pollution for transportationand stationary applications. However, fuel cells also havechallenging problems associated with its durability, lifetime,etc. For example, the carbon monoxide (CO) poisoning of thecatalyst, flooding of the cell cathode side, and drying of thecell membrane are issues that deserve attention.

Many studies have been done to address CO poisoning of acatalyst in a fuel cell [1] - [6]. The research shows a decreasein the cell voltage performance of the fuel cell with increasein the content of CO in the fuel supplied to the fuel cell.The poisoning problem is more severe when operating at lowtemperature like PEMFC as compared to a phosphors acidfuel cell ( PAFC). Other studies have discovered a means toimprove CO tolerance of a fuel cell [2] - [6]. An alloy catalystPlatinum-Ruthenium (Pt-Ru) was used instead of platinum atthe anode of a PEM fuel cell to reduce the adsorption of carbonmonoxide on the active catalyst surface of Pt [2]. The reactionat the anode for uncontaminated hydrogen is:

H2 + 2(Pt) −→ 2(Pt − H) (1)

2(Pt − H) −→ 2(Pt) + 2H+ + 2e− (2)

However, if the hydrogen is contaminated with CO, thereaction at the anode could occur in either of the following

two processes:

CO + (Pt) −→ (Pt = CO) (3)

2CO + 2(Pt − H) −→ 2(Pt = CO) + H2 (4)

The Pt-Ru alloy catalyst is used to reduce the CO adsorptionsignificantly on the Pt catalyst at the anode of a PEM fuel cell.This allows a water-gas shift reaction to occur if the fuel ishumidified before it is fed into the anode. The reactions thatoccur in the anode are:

(Pt) + H2O −→ (Pt − OH) + H+ + e− (5)

Ru + H2O −→ (Ru − OH) + H+ + e− (6)

(Pt = CO)+(Ru−OH) → (Pt)+Ru+CO2+H++e− (7)

Another approach to oxidizing the CO adsorbed on thecatalyst surface into carbon dioxide (CO2) is air bleeding intothe anode with the fuel stream [3]. This helps to free someof the active catalyst site on the Pt available for hydrogen(H2). Moreover, a thin catalyst layer was added onto theanode, where a direct oxidation of CO with the O2 occurs,before the fuel reaches the internal catalyst layer in which theH2 is oxidized [4]. Operating the cell at higher temperaturecould also improve the tolerance of CO poisoning [3] and [5].However, the dynamics of the fuel cell will be significantlyaffected, which limits its application for transportation. Inter-esting results of improved CO tolerance of a PEMFC werefound when an advanced power converter system was used thatdraws a pulsing current [6]. A low frequency pulsating currentwas drawn from the fuel cell, which drives (pushes) the fuelcell to operate at high over-potential in the VI curve. Operatingthe PEMFC at high over-potential allows CO oxidation intoCO2, which frees some of the active catalyst surface of Pt toallow a fast electro-oxidation of H2.

Water management is crucially important for healthy oper-ation of a PEMFC. [7] - [10] discuss monitoring a PEMFCduring flooding and drying conditions using an electrochem-ical impedance spectroscopy (EIS), and by measuring thecell resistance and pressure drop. Monitoring liquid watercontent in the porous electrodes by measuring the pressuredrop between the inlet and outlet was proposed in [7]. Acombination of pressure drop at the cathode side and measur-ing the cell resistance is used to reliably indicate separatelybetween drying and flooding faults [8]. [9] uses impedance

978-1-4244-2279-1/08/$25.00 © 2008 IEEE 1

response at separate frequency ranges to distinguish betweenflooding and drying faults. A detailed analysis of the state-of-health (SOH) of a PEMFC using EIS to detect and isolate ispresented in [10]. This study uses Randles cell to model thecell during both flooding and drying.

However, detection and isolation of the faults in a PEMFCoperation is not enough. The act of mitigating these faults iscritically important to improve the performance of the cell andhave longer life time. This paper proposes not only detectionand isolation of the fault source, but also mitigating of the cellfrom flooding, drying and CO poisoning faults. EIS is used todetect and isolate the cause of the fault and a simple boost-buck cascaded DC-DC converter is designed to mitigate thefault. Pulsating the cell current was observed to increase thecell power by 8%.

II. FLOODING AND DRYING OF PEMFC

The membrane of the PEM fuel cell has to be wet fornormal operations. The water inside the membrane transportsthe protons (H+) form the anode side to the cathode sideby osmosis through the membrane. Enough water in themembrane can be achieved by humidifying the incoming fuel(H2) and air (O2). However, any shortcomings that arise dueto the imperfect humidifier sensors can lead to too much ortoo little water being injected into the fuel cell, which in turncauses flooding or drying. Prolonged operation in either ofthese two states decreases the output power of the fuel cell.Furthermore, this can be very harmful, or even fatal, to thefuel cell [7]. For flooding to occur, excess water has to beinjected into the cell and operate at high current density. It is aslow process of liquid water accumulation inside the cathodegas diffusion layer (GDL) of the cell. During this process,the voltage drops slowly as if the current was limited by thediffusion of reactants. Eventually water droplets are formedinside the gas channels that prevents oxygen from reachingthe catalytic sites, thus rapidly reducing the cell voltage tozero. For drying to occur, too little water is injected into thecell and it operates at a low current density. The membrane ofthe cell dries out and the voltage gradually drops. Eventually,the cell dries out and the voltage drops suddenly to zero ina similar manner to a concentration drop. Detection of theflooding and drying while it is during the slow voltage dropstage is important as prolonged operation is extremely harmfulto the fuel cell [7].

III. EXPERIMENTAL SETUP

All measurements were carried out on a single PEMFC cellfed with air and pure hydrogen. The fuel cell experimentaltest setup consists of a 10cm2 single cell, gas humidifiers, lineheaters for the gas inlet lines, moisture trap, cell temperaturecontroller, mass flow controller for the incoming reactant gasesand back pressure controller for the exiting gases as shownin Fig.1. Schematics of the cell connected to an electronicload and a frequency response analyzer (FRA) are shown inFig.2. The following parameters of the cell were maintained

Humidifier

Mass Flow and Back Pressure

Controller

PEMFC

Oscilloscope

Fig. 1. Experimental test setup of the Electrochemical Impedance spectro-scope (EIS) of PEM fuel cell.

Cathode

Anode

H2 Tank Humidified H2

Humidified O2

Humidifier

O2 Tank

Mass flow and back pressure

controller

Controlled Input flow

Controlled output flow

V

Frequency Response Analyzer

A

AC Generator

Electronic Load

Impedance

response data

Display, Monitor and data analyzer

Voltage and current readings

AC Control Signal

PEMFC

Fig. 2. Schematics of the the experimental test of the PEM fuel cell.

to ensure consistency and repeatability of the experiment forall the experiments carried out.

• Pure oxygen (O2) and Hydrogen (H2) flow are set at 100standard cubic centimeter/min (SCCM) during the test.And the fuel and oxidant flow back pressure is kept at20 psi.

• The humidity is maintained at the desired level using thehumidifier by controlling the humidifier temperature.

• The cell current was controlled at the predetermined level,which in this case is 8 A using the electronic load. AFRA is used to collect the impedance response of the cellduring the course of the test. And an oscilloscope recordsthe voltage response in time while the cell is either dryingor flooding.

A. Flooding Procedure

Initially the cell and the humidifier temperatures are keptthe same for normal operation. To create flooding in the cell,the humidifier temperature is maintained at about 40 0C orhigher than the cell temperature. The cell can be operated atvery high current to facilitate the flooding process. This createsa situation where there is a net water gain into the fuel cell,and the cell is eventually flooded. This process leads to a dropthe cell voltage. Finally, the fuel cell reaches a point wherethe voltage falls to a very low level indicating flooding hasoccurred.

2

0 15 30 45 60 75 90 105 120 135-0.3

-0.2

-0.1

0

0.1

0.2

0.3

0.4

0.5

0.6

0.7

Time (min)150

Cel

l Vol

tage

(V)

At 8A cell current

At 14.5A cell current

Fig. 3. Cell voltage response during flooding.

B. Drying Procedure

Drying of the cell can be achieved by creating a temperaturedifference between the humidifier and cell. Similar to theflooding process, at the beginning the temperature differenceis kept to zero for normal operation. The cell temperature isthen maintained at about 40 0C or higher than the humidifierto create drying. Operating the cell at low current will speedup the drying process. This process creates a net water lossfrom the fuel cell. As a result, the cell membrane starts todry with time. Eventually, a fall in the cell voltage is seenover time. Finally, the fuel cell reaches a point where the cellvoltage drops to zero and it can no longer produce current.

IV. EXPERIMENTAL RESULTS AND DISCUSSION

The frequency response analyzer (FRA) is employed toprovide the forcing signal to the AC generator and collectthe impedance response as shown in Fig.2. The amplitudeof the AC signal is set to a maximum amplitude of 30mV,which injects less than 5% AC current into the cell at theselected operating current (DC cell current). The cell currentis maintained at 8A during the experiment using the electronicload. The frequency sweep range of the AC signal is setfrom 0.1 to 15kHz of 10 sweeps/decade. If the frequencysweep decreased below 0.1Hz, the data collected is very noisyand hard to analyse. The imaginary part of the impedanceresponse will be dominated by inductive impedance if thesweep frequency increases above 15kHz due to the inductanceof the connecting wires.

Flooding of the cell was attempted at 8A operating cellcurrent as shown Fig.3. Unfortunately, complete floodingwithout injecting water into the cathode side of the cell was notsuccessful. The cell current was increased to 14.5A after about3600s operating at 8A to force the cell to flood completelyas shown in Fig.3. The fluctuating voltage response startingfrom about 6000s to 8100s shows a sign of complete floodingof the cell. The cell voltage drops to zero and recovers toa higher voltage value, and then slowly decreases where itfinally drops again to zero. During this flooding stage, the backpressure of the unreacted exiting oxygen was seen to drop,which also indicates flooding. It was observed that recording

0.004 0.006 0.008 0.01 0.012 0.014 0.016 0.018 0.02-0.001

0.001

0.002

0.003

0.004

0.005

Z_re(Ω)

-Z_i

m(Ω

)

Impedance Nyquist plot during Flooding

0

0.006

Fig. 4. Impedance response of the cell during flooding.

|Z|(Ω

)Ph

ase(

angl

e)

10-1 100 101 102 103 1040

0.01

0.02

0.03Magnitude response

10-1 100 101 102 103 10411.5

0

-11

-23

-34Phase response

Frequency (Hz)

Frequency (Hz)

Fig. 5. Magnitude and phase responses of the cell during flooding.

VI data of the cell during this complete flooding is hard,since the cell recovers and continues with normal operationfor a short period of time. Very low oxygen content is stilldiffused through the cathode GDL and reaches the catalystregion generating current during this complete flooding stage.The cell resumes its normal operation once it is operated atlow current densities, which increases its output power.

The impedance response of the cell during the floodingprocess is shown in Fig.4. Both the real and imaginary part ofthe cell impedance increases as the liquid water accumulationin the cathode side increases. However, the shape (semicircle)of the impedance response remains almost same with theexception of increased diameter and shifting slightly to theright. The increase in the polarization resistance (bulk chargetransfer resistance) in the cathode side causes the increase inthe the diameter of the cell impedance response. The slightshift of the semicircle to the right is due to the decrease incell temperature, which increased the membrane resistance.An empirical formula [11] was used to calculate the cellmembrane resistance as a function of the temperature and op-erating cell current. The result of the empirical formula and themeasured resistance are almost the same. Maximum magnitudeof the imaginary part of the impedance response of the celloccurs at the same frequency 630Hz. This implies the transientresponse remains the same. Fig.5 shows the magnitude and

3

phase response of the cell impedance. An increase in the cellmagnitude response is seen as the cell floods, but the phaseresponse remains almost the same throughout the flooding. Anegative phase shift is seen between 50Hz and 5kHz frequencyranges, which indicates almost the same transient response ofthe cell (settling time of 0.5ms) between the different floodingstages and the healthy cell.

Figs.6 and 7 show the voltage response of the cell during thedrying process. Complete drying of the cell was seen to occurafter running for about 7800s at 8A operating cell current asshown in Fig.6. The drying resulted in a slow decrease of thecell voltage from 0.58V to 0.475V, and finally rapidly dropsfrom 0.45 to 0.258V, which indicates complete drying. Thecell was also subjected to drying again after recovering tonormal operation. The second drying process took less timeto completely dry the cell, which is about 3600s as shownin Fig.7. The drying resulted in a reduction of 50% of themaximum power output of the cell. A power and VI curvecomparison of the normal and drying operation is shown inFig.8 with maximum power of 4.6W. The actual maximumpower loss is about 2.354W, and the voltage drop at 8A is0.308V.

The impedance response of the cell during the drying stagesare shown in Figs.9 and 10. Fig.9 shows the real vs imaginarypart of the impedance response, and the magnitude and phaseresponse are shown in Fig.10. The plot shows an increasein the magnitude of the real and imaginary responses, anda generation of another semicircle at low frequencies. Anegative slope ( d|Z|

df )of the magnitude response plot for the lowfrequency test , and a negative phase shift starts to appear asthe cell membrane dries, which does not occur for normal andflooding. This helps to identify the drying fault from floodingand the healthy cell. In addition, the dynamic response of thecell starts to become slower (settling time of 0.87s) than theresponse of the healthy cell (settling time about 0.35µs).

A fault in the cell operation can be easily detected bymeasuring the cell voltage. A drop in the cell voltage at a givenoperating current could lead to a fault occurring in the cell.Limiting the cause of the fault to either flooding or drying, themagnitude or phase response of the cell at low frequency can

0 15 30 45 60 75 90 105 120 135 1500.25

0.3

0.35

0.4

0.45

0.5

0.55

0.6

0.65

Cel

l vol

tage

(V)

Time (min)

Cell voltage response during drying process

Fig. 6. Voltage response during initial drying attempt.

0 10 20 30 40 50 60 70-0.1

0

0.1

0.2

0.3

0.4

0.5

0.6

0.7

Time (min)

Cell voltage response during drying process

Cel

l vol

tage

(V)

Injected AC voltage

Fig. 7. Voltage response during second drying attempt.

0 2 4 6 8 10 12 14 16

0

0.2

0.4

0.6

0.8

1

Cell current (A)

Cel

l vol

tage

(V),

Cel

l Pow

er (W

)

NormalDrying

Comparison of VI & power curves between normal and drying

∆Ploss= 2.354 W

Pmax

Pmax

Fig. 8. Power and VI curves for healthy and drying cell.

be used for isolating the cause of the fault. A transient analysison the cell voltage can also be used to isolate the faults. Acombination of both isolation methods (low frequency test andtransient analysis) provide a better decision. A comparisonof the normal, flooding and drying impedance response isshown in Fig.11. The response of the normal and floodingstates contain one semicircle with the exception that a highermagnitude response occurs during flooding, while the responseduring the drying state contains two semicircles, which makesit different from the flooding fault. A simple analysis of themagnitude response at low frequency (0.01 to 1 Hz) results inan adequate signature to identify the fault type.

V. PEMFC MODEL INCLUDING FLOODING, AND DRYING

A fuel cell model that contains the flooding, and dryingfaults is developed as shown in Fig.12. An equivalent circuitis added to an existing PEMFC model to account for theeffect of both the flooding and drying faults. For a healthycell, the parameters R1, R2 and C2 are considered in thePEMFC model to represents the cell equivalent impedance.These three parameters start to change as a result of floodingand drying faults. The increase in the value of the first parallelRC circuit (C2 and R2) is a significant marker of flooding only.The change in these parameters is introduced as a decrease incapacitance C2 and increase in resistance R2 . The parametersC3 and R3 are not affected by flooding, but drying of the cell.

4

0 0.01 0.02 0.03 0.04 0.05 0.06-0.002

0

0.002

0.004

0.006

0.008

0.01

0.012

0.014Impedance Nyquist plot during Drying

Z_re(Ω)

-Z_i

m(Ω

)

Fig. 9. Impedance response of the cell during drying..

10-1 100 101 102 103 1040

0.05

0.1

Frequency (Hz)

Magnitude response

10-1 100 101 102 103 10430

0

-30

-60Phase response

Frequency (Hz)

|Z|(Ω

)Ph

ase(

angl

e)

Fig. 10. Magnitude and phase responses of the cell during drying.

The model of the drying fault considers not only the parameterchanges in C2 and R2, but also R1 and the addition of thesecond parallel RC circuit C3 and R3 as shown in Fig.12.

The first RC parallel circuit (C2 and R2 ) is replaced by aresistor R in parallel to a constant phase element impedance(CPE). A CPE impedance is an equivalent electrical circuitcomponent that models the behavior of a double layer, that isan imperfect capacitor, which is given by eq(8). The CPE isintroduced since the impedance response of the cell is nota semicircle centered at the real axis, but below the realaxis. However, the depressed semicircle impedance responsecan be approximated well by using the CPE impedance. TheCPE impedance value depends on the electrode roughness,distribution of reaction rates, varying thickness or compositionof a coating, and non-uniform current distribution [10] - [12].

1Z

= Y = Q ∗ (jw)n (8)

Where, Z is the CPE impedance, Y is the CPE admittance,Q is the admittance in S.sn at w = 1 rad/s, w is the frequencyin rad/s, and n is between 0 and 1.

The CPE impedance can be approximated by a capacitiveimpedance for n values very close to 1. However, the truecapacitance value can be expressed as eq(9) [12].

C = Q ∗ (wmax)n−1 (9)

0 0.02 0.04 0.06 0.08 0.1 0.12 0.14 0.16 0.180

0.01

0.02

0.03

0.04

0.05

0.06

0.07

0.08

0.09

0.1

Z_re(Ω)

-Z_i

m(Ω

)

Comparison between experimental, flooding and drying

NormalFloodingDrying

Fig. 11. Impedance response of health, flooded and dried cell.

Where, wmax is the frequency at which the imaginarycomponent reaches maximum, which is wmax = 3958rad/s.

The parameters of the C and CPE-model equivalent circuitare given in Tables I(a) and I(b) for both faults at differentlevel of flooding and drying.

A comparison of the impedance response betweenexperimental, C-model, and CPE-model are shown inFigs.13, 14, 15, 16 and 17. Fig.13 shows the plot of impedanceresponse for the healthy, flooded, and dried cell. Figs.14 and 16are the magnitude and phase responses for both models duringflooding and drying. And the impedance response for bothfaults are shown in Figs.15 and 17. These plots show thecomparison between experimental, C-model, and CPE-modelsat different stages of flooding and drying. It can be observedthat the CPE-model responds better in both fault modelingcases. However, the C-model has the advantage of simplecircuit implementation in Matlab/Simulink or Pspice as a tradeoff to a better approximation.

VI. PROPOSED METHOD OF IMPROVING PEMFCPERFORMANCE USING POWER CONVERTER

The PEMFC problems discussed in the previous sectionscan be mitigated by selecting the best suitable operating pointin the VI curve of the fuel cell. For example, low currentoperation has the advantage of mitigating the flooding fault,which is observed during the experiment. Drying and COpoisoning faults of the PEMFC are mitigated by operating athigh current densities. High current density operation results in

3R

3C

Nernst Reversible Voltage

Anodic Activatio

n loss

Cathodic Activatio

n loss

Concentration loss

1R2R

Nernst Reversible

Voltage

Anodic Activation

loss

Cathodic Activation

loss

Concentration loss

+-

+ +

2CEquivalent Circuit of the Cell

Due to Drying fault only

Cel

l Vol

tage

+

Due to CO poisoning only

Activation loss

Open Circuit Voltage

Fig. 12. PEMFC model including flooding and drying.

5

TABLE IPEMFC EQUIVALENT CIRCUIT PARAMETERS OF THE C AND CPE-MODEL.

(a) parameters of the C-modelCell Status R1(Ω) R2(Ω) C2(F ) R3(Ω) C3(F )

Healthy Cell

Normal 0.00418 0.0051 0.0625 0.0 infinityFlooding Cell

Stage I 0.00450 0.00662 0.04502 0.0 infinityStage II 0.00495 0.00770 0.03500 0.0 infinityStage III 0.00550 0.00965 0.03205 0.0 infinityStage IV 0.00635 0.01270 0.02750 0.0 infinityFlooded 0.01750 0.17000 0.00510 0.0 infinity

Drying Cell

Stage I 0.00508 0.008500 0.03850 0.001400 600.0Stage II 0.00535 0.001550 0.03150 0.009025 400.0Stage III 0.00645 0.012600 0.02250 0.002900 300.0Stage III 0.00625 0.018300 0.01350 0.018200 25.54Stage IV 0.00625 0.021250 0.01150 0.028850 15.55

Dried 0.00835 0.036900 0.00865 0.049500 11.55

(b) parameters of the CPE-modelCell R1(Ω) R2(Ω) Q R3(Ω) C3(F ) n

Status

Healthy

Normal 0.00418 0.0051 0.1520 0.0 infinity 0.9Flooding

Stage I 0.00450 0.00662 0.1012 0.0 infinity 0.90Stage II 0.00495 0.00770 0.0902 0.0 infinity 0.90Stage III 0.00550 0.00965 0.0705 0.0 infinity 0.90Stage IV 0.00635 0.01270 0.0575 0.0 infinity 0.90Flooded 0.01750 0.17000 0.0051 0.0 infinity 1.00

Drying

Stage I 0.00508 0.00850 0.05850 0.00140 600.0 0.93Stage II 0.00535 0.00155 0.05810 0.00903 400.0 0.93Stage III 0.00645 0.01260 0.03850 0.00290 300.0 0.93Stage III 0.00625 0.01830 0.01850 0.01820 25.54 0.96Stage IV 0.00625 0.02125 0.01650 0.02885 15.55 0.96

Dried 0.00835 0.03690 0.00955 0.04950 11.55 0.96

0 0.02 0.04 0.06 0.08 0.1 0.12 0.14 0.16 0.180

0.01

0.02

0.03

0.04

0.05

0.06

0.07

0.08

0.09

0.1

Z_re(Ω)

-Z_i

m(Ω

)

Comparison between experimental, and C and CPE-models

C -modelCPE-modelExperimental

Fig. 13. Comparison between experimental, C, and CPE-model for normal,flooding, and drying.

10-1 10 0 101 102 103 10 4 105 1060

0.005

0.01

0.015

0.02

0.025

Frequency (Hz)

|Z|(Ω

)

Magnitude Response

10-1 100 101 102 10310

410

5 106-5

05

1015202530

Frequency (Hz)

-Pha

se (d

egre

e)

Phase Response

ExperimentalC-ModelCPE-Model

ExperimentalC-ModelCPE-Model

Fig. 14. Magnitude and phase responses of experimental, C, and CPE-modelduring flooding.

0.004 0.006 0.008 0.01 0.012 0.014 0.016 0.018 0.02-0.001

0

0.001

0.002

0.003

0.004

0.005

0.006

0.007Impedance Response comparison between experimental, C-model, and CPE-model

ExperimentalC-ModelCPE-Model

-Z_i

m(Ω

)

Z_re(Ω)

Fig. 15. Impedance response of experimental, C, and CPE-model duringflooding.

10-1 100 101 10 2 10 3 10 4 105 10 600.010.020.030.040.050.06

Frequency (Hz)

Magnitude Response

10-1 100 101 10 2 103 10 4 105 106-10

0

10

20

30

40

Frequency (Hz)

-Pha

se (d

egre

e)

Phase Response

|Z|(Ω

)

ExperimentalC-ModelCPE-Model

ExperimentalC-ModelCPE-Model

Fig. 16. Magnitude and phase responses of experimental, C, and CPE-modelduring drying.

more water production at the cathode side, which increases theliquid water in the membrane. During the drying experimentaltest, it was observed that the performance of the cell wasimproved by pulsing the cell current between 8A and 10A asshown in Fig.18. An arbitrary switching periods (Ts) 1min and2min of 50% pulse width increased the average cell power by8%. This operating point is also favorable for the CO at theanode side to be oxidized and allow additional surface areain the Pt catalyst for the incoming hydrogen (fuel). Overallperformance of the PEMFC is improved by employing a powerconverter that controls the fuel cell operation during thesefaults.

Fig.19 shows the power converter connected at the terminalsof the PEM fuel cell. The power converter is made of aboost-cascade-buck converter. The boost converter controls thevoltage and current operating point of the cell, while the buckconverter provides constant output power (voltage) to the load.

The power converter connected to the terminals of a PEMFCis simulated in a Matlab/Simulink program. A current controlis implemented using two hysteresis current control loops anda switch as shown in Fig.20. The hysteresis loops control themagnitude of the cell current, where as the switch controls theswitching frequency and pulse width between the two currentmagnitudes of the cell. A PI controller of the buck converterensures a constant voltage output to the load.

The cell current, cell voltage and the load voltage of thesimulation are shown in Figs.21 and 22. Figs.21(a) and 21(b)show the pulsing current drawn and voltage waveforms of

6

0.01 0.02 0.03 0.04 0.05 0.06 0.07-2

0

2

4

6

8

10

12

14

16

Impedance Response comparison between experimental, C, and CPE-model

-Z_i

m(Ω

) x 1

0 -3

ExperimentalC-ModelCPE-Model

Z_re(Ω)

Fig. 17. Impedance response of experimental, C, and CPE-model duringdrying.

0 50 100 150 200 250 3000

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

1

Cel

l vol

tage

(V),

Cel

l Pow

er (W

)

Average PowerInstantaneous PowerVoltage

Time (s)

(a) Voltage and power waveforms at Ts=120 s

0 50 100 150 200 250 3000

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

1 Average PowerInstantaneous PowerVoltage

Cel

l vol

tage

(V),

Cel

l Pow

er (W

)

Time (s)

(b) Voltage and power waveforms at Ts=60 s

Fig. 18. Improved performance of the drying cell when pulsing the cellcurrent.

Power Converter

Measurement Circuit

Power Supply

Fig. 19. Power electronic converter setup circuit.

Switch

In1Out1

Voltage Control

RL

[Vo]

CCVfc

PEM

FC

Mod

el

A

Vo

Current Control

Ifc

Boost Converter Buck Converter

PWM

PWM

Fig. 20. Fuel cell interconnected to the power converter simulation circuit.

0 1000 2000 3000 4000 5000 60003

3.5

4

4.5

5

5.5

6

Time (ms)C

ell C

urre

nt (A

)(a) Pulsing cell current waveform

0 1000 2000 3000 4000 5000 60000.4

0.45

0.5

0.55

0.6

0.65

Cel

l vol

tage

(V)

Time (ms)

(b) Voltage response of the cell

Fig. 21. Cell current and cell voltage waveforms of a PEMFC.

the cell. The cell current is controlled at 4.25A and 5.25Aby the hysteresis controls and the corresponding cell voltagesare 0.55V and 0.475V. The load voltage at the terminals ofthe buck converter is controlled at 5V as shown in Fig.22.Experimental tests of the power converter shown in Fig.19was done to show proper operation. The power converter isconnected to a DC power source instead of the actual PEMfuel cell. The pulsing current drawn by the power converter isshown in Fig.23. The source current is controlled at an averageof 1.5A and 3.5A as shown in Fig.23(a). A maximum currentripple of ∆i = 0.5A occurs when the average current is 3.5Aas shown in Fig.23(b). The reason for the high current rippleis due to the delay in the reading sensors, limitation of the

7

0 1000 2000 3000 4000 5000 6000 70000

1

2

3

4

5

6

7

8

Time (ms)

Load

Vol

tage

(V)

Fig. 22. Voltage waveform of the buck converter load

0 1 2 3 4 5 60

0.5

1

1.5

2

2.5

3

3.5

4

4.5

Cel

l Cur

rent

(A)

Time (ms)

(a) Current waveform of the boost converter

4.8 4.8005 4.801 4.8015 4.802 4.8025 4.8033

3.1

3.2

3.3

3.4

3.5

3.6

3.7

3.8

3.9

4

Time (ms)

Cel

l Cur

rent

(A)

(b) Ripple current of the boost converter

Fig. 23. Pulsing current waveform of the power converter experimental test.

controller, and the inductor size. The ripple can be significantlyreduced either by increasing the inductor size or increasing theswitching frequency of the boost converter. As a result of thesechanges, the efficiency of the power converter will decrease.

VII. CONCLUSION

In this paper, an experimental setup of a single cell PEMFCfor flooding and drying tests was presented. The impedanceresponse due to the flooding and drying faults was brieflydiscussed. The impedance response at low frequency wasfound to carry the information required for isolating the causeof the fault. A pure capacitance and CPE based modelswere discussed to represent the equivalent circuit of the cellimpedance. The CPE model equivalent circuit showed better

approximation of the cell impedance as compared to the C-model. However, the C-model is simple and well known insimulation.

A power converter circuit was built to mitigate the faultycell. Pulsing the cell current at low frequency (about 0.0085Hz) improves the performance of the cell. The cell voltage wasseen to increase due to the pulsing and higher water productionat higher cell current. Simulation and experimental results ofthe converter were presented to show proper operation. Theconverter controls the cell current and voltage operating points,which mitigates the fault.

REFERENCES

[1] Aida Rodrigues, John C. Amphlett, Ronald F. Mann, Brant A. Peppley,and Pierre R. Roberge, Carbon Monoxide Poisoning of Proton-ExchangeMembrane Fuel cells, IECEC-94 proceedings of the 32nd intersociety, 27july - 1 Aug, 1997, Vol.2, page(s) 768-773.

[2] Z. Qi and A. Kaufman, CO-tolerance of Low Loaded Pt/Ru Anodes forFuel Cells, J. of Power Source, Vol.113, No.1, pp115-123, 2003.

[3] S. Gottesfeld and J. Pafford, A New Approach to the Problem of CarbonMonoxide Poisoning in Fuel cells Operating at Low Temperature, J.Electrochemical Society, Vol.135, p 2561(1988).

[4] Uribe Francisco A., Zawodzinki Thomas A., Valerio Judith A., andGarzon Fernando H., PEMFC Reconfigured Anodes for Enhancing COTolerance with Air Bleed, J. Electrochemical Society, Vol.7, No.10, ppA376-A379, 2004.

[5] Mahesh Murthy, Manual Esayian, Woo-kum Lee, and J.W. Van Zee,The effect of Temperature and Pressure on the Performance of PEMFCExposed to Transient CO Concentrations.,J. Electrochemistry Society, Vol150, No 1, pp.A24-A34, January 2003.

[6] Woojin Choi, Prasad N. Enjeti, and Anthony J. Appleby, An AdvancedPower Converter Topology to Significantly Improve the CO Tolerance ofthe PEM Fuel Cell Power Systems, Industrial Applications Society, IAS2004.

[7] W. He, G. Lin, T.V. Nguyen, ”Diagnostic tool to detect electrode floodingin proton exchange membrane fuel cells”, AIChE J., Vol. 49 (12) (2003),3221.

[8] F. Barbir, H. Gorgun, X. Wang, ”Relationship between pressure drop andcell resistance as a diagnostic tool for PEM fuel cells”,Journal of PowerSources, Vol.141(2005), pp.96101.

[9] W. McYerida , D.A. Harrington , J.M. Le Canut , G. McLeand, ”Char-acterisation of proton exchange membrane fuel cell (PEMFC) failuresvia electrochemical impedance spectroscopy”, Journal of Power Sources,Vol. 161 (2006), pp.264274.

[10] N. Fouquet, C. Doulet, C. Nouillant, G. Dauphin-Tanguy , and B. Ould-Bouamamab, ”Model based PEM fuel cell state-of-health monitoring viaac impedance measurements”, Journal of Power Sources, Vol. 159(2006),pp.905-913.

[11] David J. Hall, R. Gerald Colclaser, ”Transient Modeling and Simulationof a Tubular Solid Oxide Fuel Cell”, IEEE Transactions on EnergyConversion, Vol. 14, No. 3, September 1999.

[12] WH Mulder, JH Sluyters, T Pajkossy, I Nyikos, ”Tafel current at fractalelectrodes. Connection with admittance spectra,” J. ElectroanalyticalChemistry., vol. 285(1990), pp.103-115.

8