Journal of Power Sources 196 (2011) 3810–3823

Contents lists available at ScienceDirect

Journal of Power Sources

journal homepage: www.elsevier.com/locate/jpowsour

Modeling of the Ballard-Mark-V proton exchange membrane fuel cell with power
converters for applications in autonomous underwater vehicles夽
Chien-Hsing Lee ∗ , Jian-Ting Yang
Department of Systems and Naval Mechatronic Engineering, National Cheng Kung, University, 1, University Rd., Tainan 701, Taiwan, ROC

a r t i c l e i n f o a b s t r a c t

Article history: This paper studies the transient response of the output voltages of a Ballard-Mark-V 35-cell 5 kW pro-
Received 21 November 2010 ton exchange membrane fuel cell (PEMFC) stack with power conversion for applications in autonomous
Received in revised form underwater vehicles (AUVs) under load changes. Four types of pulse-width modulated (PWM) dc–dc
10 December 2010
power converters are employed to connect to the studied fuel cell in series for converting the unregu-
Accepted 10 December 2010
lated fuel cell stack voltage into the desired voltage levels. The fuel cell model in this paper consists of the
Available online 21 December 2010
double-layer charging effect, gases diffusion in the electrodes, and the thermodynamic characteristic;
PWM dc–dc converters are assumed to operate in continuous-conduction mode with a voltage-mode
Keywords:
Fuel cells
control compensator. The models of the study’s fuel cell and PWM dc–dc converters have been imple-
Autonomous underwater vehicles mented in a Matlab/SIMULINKTM environment. The results show that the output voltages of the studied
Voltage-model control PEMFC connected with PWM dc–dc converters during a load change are stable. Moreover, the model
dc–dc converters can predict the transient response of hydrogen/oxygen out flow rates and cathode and anode channel
temperatures/pressures under sudden change in load current.
© 2010 Elsevier B.V. All rights reserved.

1. Introduction byproduct—pure water and is stackable to provide a range of space

options. Its disadvantage is that pure hydrogen and oxygen reac-
Over the past few years, hydrogen production and fuel cell tech- tants must be carried. Thus, fuel storage is a key component for
nologies have grown dramatically. Proton exchange membrane fuel applying the PEMFC to underwater vehicles. For the PEMFC, pure
cells (PEMFC) are one of the most promising fuel cell technologies hydrogen must be stored. Hydrogen can be stored as a gas, liquid,
that can be a candidate for both stationary and automotive appli- or in a hydride. Out of the three options researched, the hydride
cations [1]. The traditional energy source of underwater vehicles method is chosen for its proven safety and reliability advantages
is based on the heat engine process of the Carnot cycle. It causes [4].
not only low conversion efficiency but also pollutions. Recently, The purpose of this paper is to study the performance of the com-
an unmanned untethered submersible AUV-HM1 testbed devel- bined PEMFC and pulse-width modulated (PWM) dc–dc converters
oped by National Taiwan University has been powered by lead-acid subject to a load change in autonomous underwater vehicles
batteries. However, the limited operating range detracts from the (AUVs). Although [7] has already described the PEMFC performance
usages of them. In order to design a long-range underwater vehi- in the presence of transient effects, it did not use power convert-
cle, the larger capacity of the lead-acid battery is needed but this ers to regulate the output voltage of the fuel-cell stack. In addition,
will increase the weight and occupy most of the volume of the [8] has built a nonlinear model of buck converters, and in [9–11]
underwater vehicle. Thus, fuel cells coupled with an alternative have mentioned to the use of boost converters to regulate the out-
reactant storage system may have tried to overcome these limita- put voltage of the fuel-cell stack. Nevertheless, they have not yet
tions while maintaining other benefits of an electrochemical power shown the phenomenon that the load variations may affect the
source [2–6]. output voltage of the power converter.
Generally, the PEMFC has multiple advantages. It offers a quiet
low temperature operating condition with only producing a clean
2. Mathematical model of the PEMFC

This section describes a mathematical model for investigating

夽 This work was supported in part by the National Science Council, Taiwan, under
the mass and energy dynamic interactions as well as the electro-
Grant NSC 99-2221-E-006-227.
∗ Corresponding author. Tel.: +886 6 2757575x63539; fax: +886 6 2747019. chemical reaction within a PEMFC [12,13]. As seen from Fig. 1, the
E-mail addresses: [email protected] (C.-H. Lee), hydrogen enters the anode channel, and diffuses through the anode
[email protected] (J.-T. Yang). diffusion layer toward the active catalyst layer, where the hydrogen

0378-7753/$ – see front matter © 2010 Elsevier B.V. All rights reserved.
doi:10.1016/j.jpowsour.2010.12.049
 C.-H. Lee, J.-T. Yang / Journal of Power Sources 196 (2011) 3810–3823 3811

Fig. 1. Schematic diagram of a PEMFC.

dissociates into protons and electrons. The protons then travel from cell. In addition, the physical properties associated with the control
the anode catalyst to the cathode catalyst through a membrane. In volumes are assumed to be uniform.
the meantime, oxygen enters the cathode channel, part of which
diffuse through the cathode diffusion layer to the active catalyst, 2.1.1. Anode and cathode channels
where the oxygen dissociates and combines with protons and elec- Generally, air or oxygen is fed to the cathodic compartment
trons to form water, while heat and current are generated. The heat while hydrogen is fed to the anodic one and the electrolyte per-
produced during the reduction reaction process will warm up the forms both the functions of transferring H+ from the anode to the
fuel cell body that will affect the mass diffusion within the anode cathode and reactant separation. Note that pure hydrogen and oxy-
diffusion layer, cathode diffusion layer and membrane. Typically, gen are served as the fuel and oxidant gases for underwater appli-
the heat transfers to ambient and gas flow channels through con- cations, respectively. Thus, the first term in (1) becomes the mass
vection and radiation. The resulting gas temperature variation will change rate of the respective gases for the control volumes corre-
change the pressure and flow rate within the channels to affect the sponding to the anode and cathode channels. The second term in
gas diffusion within the diffusion layers. All these dynamic inter- (1) is the mass flow rate through the surface of the control volumes.
actions will have a direct effect on the fuel cell output voltage and For the anode channel, the hydrogen mass change rate is written
current. to be [12,13]:

∂ dmH2
2.1. Relationship between mass and energy transfer using control H2 dV = (3)
∂t dt
volume approach
C.V.

For an arbitrary control volume, a continuity equation in an and the mass flow rate through the surface of the control volumes
integral form is given as [14]: is expressed by

∂ H2 (V · n
 ) dA = WH ,in − WH2 ,out − WH ,used
 dV + (V · n
 ) dA = 0 (1) − 2 2
(4)
∂t C.S. C.S.
C.V.
where WH2 ,in and WH2 ,out are the inlet and outlet mass flow rates,
and an energy equation in integral form is expressed by respectively. They are as functions of the hydrogen source pressure,
 the pressure within the anode control volume and ambient pres-
∂ dQ dW
et dV + h(V · n
 ) dA = + (2) sure. With the macroscopic level dynamic behavior of the fuel cell,
∂t C.S. dt dt
the inlet and outlet mass flow rates can be simplified by using the
C.V.
traditional nozzle flow rate equation which are written to be [12]:
where  is the material density within the control volume, V is the
 is the unit vector normal to the con-
velocity vector of flow fluid, n WH2 ,in = kan,in (pan,s − pan ) (5)
trol surface toward the outside, et is the specific internal energy,
WH2 ,out = kan,out (pan − patm ) (6)
h is the specific enthalpy, Q is the heat added to the control vol-
ume, and W is the work done on the control volume. Since the gas where kan,in and kan,out are the inlet and outlet mass flow rate coef-
flow velocity is usually low under fuel cell operating conditions, the ficients; pan,s , pan and patm are the hydrogen source pressure, the
kinetic and potential energies of the gases can be neglected in (2). pressure within the anode control volume and ambient pressure,
Note that the anode channel control volume is defined to be the respectively. In addition, the consumed mass flow rate of hydrogen
lumped volume from the anode inlet to the anode outlet. Similarly, is
the cathode channel control volume is defined at the cathode side.
Ncell I
The membrane electrode assembly (MEA) and all other metal struc- WH2 ,used = MH2 (7)
2F
tures involved in the fuel cell will form the fuel cell body control
volume. As for considering a fuel cell stack, the control volumes where I denotes the current, Ncell is the number of cells in a
of the anode channel, cathode channel and fuel cell body will be stack; F and MH2 are Faraday’s constant and hydrogen’s mole mass,
the summation of corresponding control volumes of each single respectively. Thus, the hydrogen mass transport rate through the
3812 C.-H. Lee, J.-T. Yang / Journal of Power Sources 196 (2011) 3810–3823

d(mO2 cv,O2 Tca )

anode channel control surface becomes: = kconv,ca (Tbody − Tca ) + kca,in (pca,s − pca )cp,O2 (Tca,in − T0 )
dt
Ncell I
(19)
dmH2 N I −kca,out (pca − patm )cp,O2 (Tca − T0 ) − MO2 cp,O2 (Tca − T0 )
= kan,in (pan,s − pan ) − kan,out (pan − patm ) − MH2 cell (8) 4F
dt 2F
Furthermore, the energy equation of (2) can be applied to the anode 2.1.2. Cell body
in a similar manner. The first term on the left-hand side of (2) repre- The continuity equation for the cell body control volume is only
sents the internal energy change rate of hydrogen within the anode concerned with the species’ conservation of electrochemical reac-
channel control volume which is given as [12,13]: tion. That is, water is formed through the electrochemical reaction
 by consuming hydrogen and oxygen. Since most parts of the cell
∂ d(mH2 cv,H2 Tan )
H2 et dV = (9) body are composed of metal structures, gas diffusion and water
∂t dt
C.V.
produced have little effect on the total mass change of cell body.
Thus, we will have [12,13]:
where cv,H2 is the specific heat of hydrogen at constant volume and 
Tan is the hydrogen temperature within the anode channel control ∂
body dV = 0 (20)
volume. The second term on the left-hand side of (2) is the heat ∂t
transfer rate due to the mass transport through the control surface C.V.
which is given as:
and the mass transport rate through the cell body control surface
− H2 hH2 (V · n
 ) dA = WH ,in hH ,in
2 2
(Membrane Electrode Assembly, MEA) is
C.S.

− WH2 ,out hH2 ,out − WH2 ,used hH2 ,used (10) − body (V · n
 ) dA = WH ,used + WO ,used − WH2 O
2 2
(21)
C.S.

where hH2 ,in and hH2 ,out are the inlet and outlet specific enthalpy Note that the above equation includes the gas diffusion
hydrogen, respectively; hH2 ,used is the specific enthalpy of con- through the electrodes and water formed by electrochemical
sumed hydrogen. Since the temperature of a PEMFC body is reactions.
relatively moderate, the radiation effect can be neglected. Based Now let us consider the energy equation for the cell body control
on Newton’s cooling law for governing the convection heat trans- volume. The internal energy change rate of cell body is given as
fer mechanism, the convection heat transfer from the cell body to [12,13]:
the anode channel can be described as [12,13]: 
∂ dTbody
dQ body et dV = mbody cp,body (22)
= Q̇conv,an = hconv,an Aconv,an T = kconv,an (Tbody − Tan ) (11) ∂t dt
dt C.V.
where hconv ,an Aconv,an = kconv,an ,hconv,an is the convection heat trans-
fer coefficient at the anode; Aconv,an is the surface area of hest and the heat transfer rate due to the mass transport through the
convection at the anode. Note that the work done on the control control surface is given as:
volume (i.e., dW/dt) disappears since the viscous normal stresses
− body hbody (V · n
 ) dA = WH ,used hH ,used
at the inlet, outlet and electrode interface of mass diffusion are C.S. 2 2

neglected. Thus, the energy equation for the anode becomes:

+ WO2 ,used hO2 ,used − WH2 O hH2 O (23)
d(mH2 cv,H2 Tan )
= kconv,an (Tbody − Tan ) + kan,in (pan,s − pan )cp,H2 (Tan,in − T0 ) where cp,body is the average specific heat of the cell body. In
dt
Ncell I addition, several heat transfer effects are involved in the term
−kan,out (pan − patm )cp,H2 (Tan − T0 ) − MH2 cp,H2 (Tan − T0 )
2F dQ/dt, which include the heat added to the cell body through
(12)
the electrochemical reaction as well as the heat transferred to
where Tan,in and T0 are the flow-in hydrogen temperature and the
the ambient and the channels via convection. Thus, we have
reference temperature (i.e., 298.15 K), respectively. cp,H is the aver-
[12,13]:
age specific heat of hydrogen.
For the cathode channel, similar equations can be derived by dQ
= Q̇conv,an + Q̇conv,ca + Q̇conv,amb
replacing the relevant subscripts by “ca” (cathode) and “O2 ” (oxy- dt
gen). The following equations are parallel to those given in (3) and
(4) and (8)–(12): + HR,T H2 (V · n
 ) dA (24)
electrode

∂ dmO2 where Qconv,an is the heat convection between the cell body and the
O2 dV = (13)
∂t dt anode channel, Qconv,ca is the heat convection between the cell body
C.V. and the cathode channel, Qconv,amb is the heat convection between
the cell body and the ambient, and HR,T is the lower heating value
− O2 (V · n
 ) dA = WO ,in − WO2 ,out − WO ,used
2 2
(14)
C.S. of the fuel. All the heat convection terms here are assumed to com-
dmO2 Ncell I ply with Newton’s cooling law. For the cell body, the term dW/dt is
= kca,in (pca,s − pca ) − kca,out (pca − patm ) − MO2 (15) the electric power converted from electrochemical reaction, i.e.:
dt 4F
 dW
∂ d(mO2 cv,O2 Tca ) 
= −Vstack I (25)
O2 et · dV = (16) dt
∂t dt
C.V. 
where Vstack denotes the stack voltage. Thus, the heat added to and
the work done on the cell body control volume is then characterized
− O2 hO2 (V · n
 ) · dA = WO ,in hO ,in
2 2
as:
C.S.
dQ dW
− WO2 ,out hO2 ,out − WO2 ,used hO2 ,used (17) + = kconv,an (Tan − Tbody ) + kconv,ca (Tca − Tbody )
dt dt
dQ Ncell I 
= Q̇conv,ca = kconv,ca (Tbody − Tca ) (18) + kconv,amb (Tamb − Tbody ) + HR,T MH2 − Vstack I (26)
dt 2F
 C.-H. Lee, J.-T. Yang / Journal of Power Sources 196 (2011) 3810–3823 3813

Note that the first three terms at right-hand side are the effects As the partial pressure of water vapor in the cathode is fixed at the
of the convection heat transfer from the surrounding structures, saturation level, the partial pressure of O2 at the interface is
the fourth term on the right-hand side is the heat produced by the sat
H O
electrochemical, and the last term is the converted electric power. O2 = (1 − XH2 O ) 2
(34)
Finally, the energy equation for the cell body will become XH2 O

dTbody Ncell I 
mbody Cp,body = HR,T MH2 − Vstack I + kconv,an (Tan − Tbody ) + kconv,ca (Tca − Tbody ) + kconv,amb (Tamb − Tbody ) (27)
dt 2F
+WH2 ,used cp,H2 (Tan − Tbody ) + WO2 ,used cp,O2 (Tca − Tbody ) − WH2 O cp,H2 O (Tca − Tbody )

2.2. Reactant gases diffusion in the electrodes

Also, the effective cathode water vapor pressure equals the satu-
To simply the mathematical model of the PEMFC, the following rated vapor pressure based on assumption 4. Thus, the effective
assumptions are made in this paper [15–21]: partial pressure of O2 by combining (33) and (34) yields:
 
(1) One-dimensional treatment. sat 1
(2) Ideal and uniformly distributed gases. O2 = H −1 (35)
2O channel exp((RT
XH O body Jcc )/4FPca DH2 O,O2 )
2
(3) Constant pressures in the fuel-cell gas flow channels.
(4) The fuel is humidified H2 and the oxidant is humidified O2 . Using the appropriate critical values and constants for a
Assume the effective anode water vapor pressure is 50% of water/oxygen mixture, (32) and (35) finally becomes:
the saturated vapor pressure while the effective cathode water  
pressure is 100%. sat 1
H2 = 0.5H −1 (36)
(5) The reaction gases are pure hydrogen and oxygen. 2O channel exp((1.653J)/T 1.334 )
XH O 2 body
(6) Parameters for individual cells can be lumped together to rep-
resent a fuel-cell stack.
 
sat is a 1
(7) As mentioned in [20], the saturated vapor pressure H O2 = sat
H −1 (37)
2O 2O channel exp((4.192J)/T 1.334 )
function of the operating temperature of the PEMFC and is writ- XH O body
2
ten to be
where (36) and (37) will be used in the Nernst equation to find the
fuel-cell output voltage.
sat −2 −5
log(H O ) = 2.95 × 10 (Tbody − 273.15) − 9.18 × 10
2
2.3. Output voltage of the fuel-cell stack
(Tbody − 273.15)2 + 1.44 × 10−7 (Tbody − 273.15)3 − 2.18 (28)
The total output voltage of a fuel-cell stack is a function of
In order to calculate the PEMFC output voltage, the effective partial
stack current, cathode pressure, reactant partial pressures, fuel cell
pressures of H2 and O2 need to be determined. In a gas mixture
temperature, membrane humidity, etc. Since the fuel-cell stack is
consisting of N species, the diffusion of component i through the
comprised of multiple fuel cells connected in series, the stack volt-
porous electrodes can be described using Stefan-Maxwell equa-
age Vstack is calculated by multiplying the cell voltage Vcell by a
tions as written to be [21]:
number of cells Ncell in the stack, which is written as:

n
RTbody
∇ Xi = (Xi Nj − Xj Ni ), i = 1, 2, . . . , n (29) Vstack = Ncell · Vcell (38)
PDi,j
j=1 The output voltage of a single fuel cell is defined by the sum of three
In the one-dimensional transport process along the X-axis as shown voltage terms: the thermodynamic potential Enernst , the activation
in Fig. 1, the effective mole fraction of water at the anode catalyst overvoltage Vact , and the ohmic overvoltage Vohm as repeated below
interface is given as: [15–19]:
  Vcell = Enernst − Vact − Vohm (39)
channel
RTbody Jac
XH2 O = XH O exp (30)
2 2FPan DH2 O,H2 where
Since XH2 + XH2 O = 1, the effective partial pressure of H2 at the Enernst = E0 − 0.85 × 10−3 (Tbody − 298.15)
interface is
RTbody
sat
H O
+ n[H2 · (O2 )0.5 ] (40)
H2 = (1 − XH2 O ) 2
(31) 2F
XH2 O

According to assumption 4 mentioned above, the effective anode Vact = −{1 + 2 Tbody + 3 Tbody [n(CO2 )] + 4 Tbody n(I)} (41)
water vapor pressure is 50% of the saturated vapor pressure. Thus,
the effective partial pressure of H2 by combining (30) and (31) 2 = 0.00286 + 0.0002n(Acell ) + 4.3 × 10−5 n(CH2 ) (42)
yields:  
  CO2 = 1.97 × 10−7 O2 exp
498
(43)
sat 1 Tbody
H2 = 0.5H −1 (32)
O 2 channel exp((RT
XH body Jac )/2FPan DH2 O,H2 )
 
2O −77
−7
CH2 = 9.174 × 10 H2 exp (44)
Similar to the analysis for the anode, the effective mole fraction of Tbody
water at the cathode catalyst interface can be found as:
  Vohm = I · Rint (45)
channel
RTbody Jcc rmem tmem
XH2 O = XH O exp (33) Rint = (46)
2 4FPca DH2 O,O2 Acell
3814 C.-H. Lee, J.-T. Yang / Journal of Power Sources 196 (2011) 3810–3823

2 Table 1
181.6[1 + 0.03J + 0.062(Tbody /303) J 2.5 ]
rmem = (47) Parameters of the Ballard-Mark-V PEMFC.
[ − 0.634 − 3J] exp[4.18((Tbody − 303)/Tbody )]
Symbol Parameter Value Unit
Note that the two overvoltage terms in (39) are both negative −5
kan,in Upstream nozzle coefficient 4.2 × 10 kg (s atm)−1
and represent nonideal fuel-cell reaction conditions. From an of a node channel
electrical/power-electronics standpoint, a fuel cell is almost a non- kan,out Downstream nozzle 4.2 × 10−6 kg (s atm)−1
linear current-dependent voltage source. While this functional coefficient of a node channel
relationship varies somewhat with different fuel cells (e.g., solid pan,s Pressure of anode 35 psig
Tan,in Upstream hydrogen 296.5 K
oxide fuel cell, SOFC), [22–24] have outlined more detailed work temperature of anode
on planar SOFC. channel
cv,H2 Specific heat of hydrogen at 10,124.71 J (kg K)−1
2.4. Double-layer charging effect constant volume
cp, 2 Specific heat of hydrogen at 14,209 J (kg K)−1
constant pressure
The double layer charging capacitance may affect the activation Van Volume of the anode channel 0.005 m3
overvoltage term in (39). This results in a first-order model where a kconv,an Heat convection coefficient 2 W K−1
sudden change in cell current is associated with a slower variation between the cell body and
in output voltage [17,19]. An electrical circuit model is generally the anode channel
RH2 Gas constant of hydrogen 4071.1 × 10−5 atm m3 (kg K)−1
used to illustrate this capacitive effect [17]. For a given cell current, MH2 Molecular mass of hydrogen 2 × 10−3 kg mol−1
the activation voltage drop can be considered across an activation kca,in Upstream nozzle coefficient 1.8 × 10−3 kg (s atm)−1
resistance Ract . Since the activation voltage term Vact introduced in of a cathode channel
(41) is negative, the derivation of a positive term Vd is required. The kca,out Downstream nozzle 2.71 × 10−4 kg (s atm)−1
coefficient of a cathode
first-order dynamics of the activation overvoltage are described by
channel
the equations as follows [17,19]: pca,s Pressure of cathode 35 psig
Vact Tca,in Upstream hydrogen 296.5 K
Ract = (48) temperature of cathode
I channel
dVd I Vd cv,O2 Specific heat of oxygen at 662.2 J (kg K)−1
= − (49) constant volume
dt Cdl Ract Cdl
cp,O2 Specific heat of oxygen at 922 J (kg K)−1
If the double-layer charging effect is integrated into the modeling constant pressure

by using Vd instead of Vact , to calculate Vstack , the fuel-cell output Vca Volume of the cathode 0.01 m3
channel
voltage now turns out to be:
kconv,ca Heat convection coefficient 10 W K−1
 between the cell body and
Vstack = Ncell (Enernst − Vd − Vohm ) (50)
the cathode channel
Note that the dynamical model for PEMFC electrochemistry RO2 Gas constant of oxygen 262.68 × 10−5 atm m3 (kg K)−1
MO2 Molecular mass of oxygen 32 × 10−3 kg mol−1
described in this section is based on a generalized empirical set of
kconv,amb Heat convection coefficient 17 W K−1
equations. Thus, by identifying appropriate parameters (i.e., num- between the cell body and
ber of cells, cell area, membrane thickness, etc.), fuel cells with the ambient
varying sizes, structures, and materials can be modeled, and the mbody Cp,body Product of mass and the 3.5 × 104 J K−1
relevant dynamics can be studied [17,19]. Table 1 lists the related average specific heat of the
cell body
parameters used to model the PEMFC in this paper. patm Ambient pressure 1 atm
Tamb Ambient temperature 296.5 K
3. Design of DC/DC power converters cp,H2 O Specific hest of water at 2186 J (kg K)−1
constant pressure
HR,T Lower heating value 1.196 × 108 J
For the designed AUV, there are four desired voltage levels
 Membrane resistivity 12.5
including 5-, 12-, 24- and 48-Vdc in response to changes in the parameter
output load and the PEMFC output voltage. To achieve this, a nega- 1 Coefficients of activation −0.948 –
tive feedback control system is used where the converter output is over voltage
compared with its reference value. The compensator produces the 3 Coefficients of activation −1.93 × 10−4 –
over voltage
control voltage, which is used to adjust the duty cycle of the switch 4 Coefficients of activation 7.6 × 10−5 –
in the converter [25–27]. This section describes the design of the over voltage
compensation networks. Cdl Double layer capacitance 0.035 × 232 F
Assume that all power switches and diodes are ideal elements, tmem Membrane thickness 1.78 × 10−2 cm
E0 Reference potential at unity 1.229 V
and the circuit operates in continuous-conduction mode (CCM).
activity
Additionally, an equivalent series resistance rC associated with the R University gas constant 8.314 J (mol K)−1
capacitor and a series parasitic resistance rL associated with the Ncell Number of cells in the 35 –
inductor including the converter equivalent circuit as shown in fuel-cell stack
Fig. 2 are added to increase the simulation accuracy [28–30]. As Acell Fuel cell active area 232 cm2
F Faraday constant 96,485 C mol−1
seen from Fig. 2, the minimum values of the inductances and capac-
itances in the converter can be obtained from the input voltage,
load size, switching frequency, and output voltage. These values errors. As a result, the stable output voltage can be obtained from
are used as a reference to design the electrical elements of power this process.
converters. Since the converter itself is a part of the control loop, the
design of such control processes requires a knowledge of the small- 3.1. Procedure and criterion of compensator design
signal characteristics of the converter. Based on the gain and phase
margin of the open-loop control-to-output transfer function of the The procedure of compensator design is simply described as
converter, the compensator is then utilized to mitigate steady-state follows:
 C.-H. Lee, J.-T. Yang / Journal of Power Sources 196 (2011) 3810–3823 3815

Fig. 2. Schematic diagram of the designed buck or boost converter.

(1) simulating the frequency response of the open-loop control-to- the converter at the first corner is −40 dB dec−1 because the two
output transfer function of the converter, poles generated from the LC filter fall close to each other. Moreover,
(2) setting the gain crossover frequency for the converters and due to the equivalent series resistance associated with a capaci-
using this frequency to feedback the gain of the compensation tor, a zero occurs there at high frequency, resulting in the slope
circuit, of −20 dB dec−1 . As for the designed boost converter, it not only
(3) selecting the compensator type as shown in Fig. 2 and has the characteristics of the frequency response of the designed
designing the electrical elements of the compensation circuit, buck converter but also has a right-half-plane zero. Based on the
and Bode diagram (i.e., gain and phase plots) for the open-loop control-
(4) simulating small-signal analysis of the frequency response of to-output transfer function, one can select the compensator type
the closed-loop control-to-output transfer function of the con- as shown in Fig. 2, corner frequency, and gain curve. The values
verter. of electrical elements used in the compensation circuit are listed
in Table 2 and the close-loop simulation circuit with the com-
As for the criterion of compensator design, it includes the fol- pensator for the designed buck and boost converters is shown in
lowing. Fig. 2.

(1) The large dc gain is required to decrease the steady state error. 4. SIMULINK block diagrams
(2) The small gain switching frequency is required to suppress the
noise produced during the switching. The mathematical expressions for the fuel-cell system and
(3) The gain crossover frequency of the converter is selected to be power converters were simulated in the Matlab/SIMULINKTM envi-
between 1/6 and 1/10 of the gain switching frequency. ronment. A block diagram for building a dynamic model of PEMFC
(4) The phase and gain margins are suggested to be 45–60◦ and and power converters is given in Fig. 3. In this block diagram, the
6–20 dB, respectively. input quantities are anode and cathode pressures (Pan and Pca ) and
operating temperature of the fuel cell (Top ). Additionally, mass dif-
3.2. Frequency response analysis fusion equations are used to calculate the effective partial pressures
of H2 and O2 . Then, the Nernst equation is employed to determine
As for the designed buck converters, the slope of the frequency the internal potential of the fuel cell. The ohmic voltage drop and
response of the open-loop control-to-output transfer function of activation voltage drop together with the double-layer charging
3816 C.-H. Lee, J.-T. Yang / Journal of Power Sources 196 (2011) 3810–3823

Table 2 and the sawtooth signal is negative, the switch is closed and the
Values of electrical elements in the circuits of the compensators and converters.
output value is zero.
Parameters Converters with the output voltage of

48 V 24 V 12 V 5V 5. Simulations results
Vstack 20–40 V 24–40 V 20–40 V 20–40 V
fs 50 kHz 50 kHz 50 kHz 50 kHz 5.1. Effects of operating conditions on cell performance of PEMFC
L 10 ␮H 37.5 ␮H 20 ␮H 230 ␮H
rL 10 m 10 m 10 m 10 m Fig. 9 shows the effects of using air or pure oxygen as the oxi-
C 1.6 mF 70 ␮F 600 ␮F 100 ␮F
dant on the output voltage and output power for Ballard-Mark-V
rc 10 m 10 m 10 m 10 m
Rload 8.2–1.37 24–3.84 2.4–0.384 25–4 PEMFC. As seen from Fig. 9, the output voltage and output power
R1 37.3 k 14.5 k 3.55 k 37.8 k using pure oxygen as the oxidant are higher than those obtained for
R2 316 k 21.4 k 15 k 69.5 k the air. Due to the low operating temperature of the Ballard-Mark-
R3 1 k 100 100 100 V PEMFC, an increase in the operating temperature may result in a
C1 1.88 nF 4.8 nF 61.7 nF 8 nF
better cell performance because of a higher efficiency of the electro-
C2 168 pF 7 nF 14.6 nF 4.36 nF
C3 15.6 nF – – – chemical reaction. Fig. 10 shows the output voltages of the studied
fuel cell under different operating conditions of temperature and
pressure. As seen from Fig. 10, a linear increase in the output voltage
with increasing the operating temperature and pressure has been
effect are applied to determine the output voltage of a single fuel observed in a high current density range. In general, the operat-
cell. ing temperature of the studied fuel cell is limited to approximately
In Matlab/SIMULINKTM a set of equations can be grouped and 80 ◦ C, and both the anode and the cathode sides operating pres-
are called subsystems. Detailed of these subsystems are given in sures are from 1 to 3 atm. As a result, due to the limitations of
Figs. 4–7. The PEMFC subsystem using the block diagram men- the underwater environment, the PEMFC uses pure hydrogen and
tioned in Fig. 3 is shown in Fig. 4. The subsystem of reactant pure oxygen as fuel in this study. Assuming the pressures of oxy-
gases diffusion in the electrodes given in Fig. 5 employs (28)–(37) gen and hydrogen in the inlet gas are operated at 3 atm with the
and calculates hydrogen/oxygen partial pressures. The subsystem temperature at approximately 70 ◦ C.
of electrochemistry for Ballard-Mark-V PEMFC employs (38)–(48)
using the parameters in Table 2 as shown in Fig. 6. It takes hydro- 5.2. Simulation of PEMFC subsystem
gen/oxygen partial pressures, cell current, and temperatures as
inputs and calculates the cell voltage. A gain at the terminal of the Fig. 11 shows the total output voltage and output power for
subsystem multiplies this output by the number of cells inside one Ballard-Mark-V 35-cell 5-kW PEMFC based on the parameters men-
stack. The subsystem of mass and energy transfer at the anode, tioned in Table 2. As seen from Fig. 11, the sharp voltage drop at low
cathode and cell body for Ballard-Mark-V PEMFC using the control currents is caused by an activation overvoltage, and the linear drop
volume approach employs (12), (19) and (27) with the parameters at intermediate currents is due to an ohmic overvoltage. Moreover,
listed in Table 1 as shown in Fig. 7. the maximum output power occurring at the load current of 239 A
As for the power converter subsystem, it includes the circuits is about 3943 W, and the corresponding output voltage is about
of power converters and compensators with voltage feedback con- 16.5 V.
trol. Based on the electrical loads of the AUV, four converters are Since the loads of the designed AUV as listed in Table 3
designed to obtain 5-, 12-, 24- and 48-Vdc output, and compen- are below 2 kW, the maximum output current of the studied
sators are designed to regulate these dc voltages in the presence of fuel cell is computed to be about 78 A, and its corresponding
transients. In Matlab/SIMULINKTM a model of the compensator cir- output is ranged from 27 V to 40 V. Fig. 12 shows the out-
cuit with voltage feedback control is built based on the references put voltage and output current of Ballard-Mark-V 35-cell 5 kW
[31,32]. To limit the paper length, Fig. 8 only shows the model of PEMFC at a power output between 0.3 kW and 2 kW. As seen
the output voltage of 24V dc–dc converter. As seen from Fig. 8, the from Fig. 12, the PEMFC output voltage is not stable due to
repeating sequence block is used to generate the sawtooth wave- the polarization losses (i.e., activation and ohmic overvoltages).
form, and the relay in discontinuities blockset is used to simulate Thus, PWM dc–dc converters (i.e., three buck converters and one
the comparator. If the difference between the compensating signal boost converter) are connected to the PEMFC output for con-
and the sawtooth signal is positive, the switch is open and the out- verting unregulated dc voltage into four desired voltage levels.
put value is one; if the difference between the compensating signal Note that PWM dc–dc converters are designed to operate under

Fig. 3. Diagram of building the models of a PEMFC and dc–dc power converters.
 C.-H. Lee, J.-T. Yang / Journal of Power Sources 196 (2011) 3810–3823 3817

Fig. 4. SIMULINK block diagram of the Ballard-Mark-V PEMFC subsystem.

overloads and light load. The light-load, full-load, and overload con- time of 8 s. As seen from Fig. 14, the output voltage and output
ditions are defined to be 20%, 100% and 120% of the maximum current of the studied fuel cell are 29.3 V and 63.1 A at full load
rated power of the designed AUV as listed in Table 4, respec- condition, respectively. However, the output voltage and output
tively. current of the studied fuel cell are 32.5 V and 11.4 A at light load
For the purpose of simulation, no loss of energy is assumed for condition, respectively.
converting an input energy of Ballard-Mark-V 35-cell 5-kW PEMFC
into an output power. The output power and output voltage of the 5.3. Simulation of PWM dc–dc power converter subsystem
studied fuel cell during a load change from full load to light load for
all four buses of voltages are shown in Figs. 13 and 14. As seen from As for the designed AUV, four PWM dc–dc converters with the
Fig. 13, the total output power of the studied fuel cell varies from compensation network using voltage-model control are connected
1841 W to 370 W as a load change occurs at 4 s during a simulation to the output of the studied fuel cell for converting unregulated

Fig. 5. SIMULINK block diagram of reactant gases diffusion in the electrodes.

Fig. 6. SIMULINK block diagram of the electrochemistry subsystem of Ballard-Mark-V PEMFC.

3818 C.-H. Lee, J.-T. Yang / Journal of Power Sources 196 (2011) 3810–3823

Fig. 7. SIMULINK block diagram of mass and energy transfer for the Ballard-Mark-V PEMFC.

voltage into four desired voltage levels during a load change. Gen- for 5-, 12-, 24- and 48-V dc–dc converters during a load change are
erally, the converter output voltage will dip at the sudden change of shown in Figs. 15–18, respectively. Since the output of the studied
load from light to heavy or jump at the sudden change of load from fuel cell ranges from 27 V to 40 V, only the 48-V dc–dc converter will
heavy to light. The magnitude and duration of the transient voltage need to use a boost-converter. The rest of converters will use buck
occurring in the output voltages during a load change are designed converters. As seen from Figs. 15–18, the output voltages of four
to be within ±5% and 3 ms in this paper. However, the range of dc–dc converters all have a transient spike during a load change at
acceptable variation in the output voltages usually depends upon 4 s. Nevertheless, their output voltages are all stable right after the
the users’ applications. The output voltages and output currents load change.
 C.-H. Lee, J.-T. Yang / Journal of Power Sources 196 (2011) 3810–3823 3819

Fig. 8. SIMULINK block diagram of the converter with 24 Vdc output voltage.

Fig. 9. Output voltage and output power of the Ballard-Mark-V PEMFC for using air
or pure oxygen as the oxidant. Fig. 11. Output power of the Ballard-Mark-V 35-cell 5-kW PEMFC and its corre-
sponding output voltage.

5.4. Simulation of the supply of the reactant gases

Table 3
Voltage levels and loads for the designed AUV.
In this subsection, a parametric analysis to simulate transient
dynamic effects on the supply of the reactant gases within the fuel Rated voltage Name of devices Maximum rated power

cell system is carried by using the parameters listed in Table 1. DC 5V Bus Embedded control main board 5W
Based on Table 4, the output power of the fuel cell system needs to DC 12V Bus Inertial measurement unit 0.78 W
Global position system 1W
Pressure transducer 0.24 W
RF modem 9W
CTD sensors 0.252 W
Headlamp 100W * 3
DC 24V Bus Ultra short baseline 10 W
Acoustic modem 8W
Inertial navigation system 6W
Doppler velocity log 80 W
Digit video camera 5–10 W
Forward-looking sonar 5–10 W
DC 48V Bus Thruster 350 W * 3
Total about 1841 W

generate about 1841 W when four converters are operated on full

load condition, and it only needs about 370 W when load change
from full load to light load is simultaneously occurred on four con-
verters. The time length for the simulation is 900 s.

Fig. 10. Output voltage of the Ballard-Mark-V PEMFC under different operating Fig. 12. Current–voltage curve of the Ballard-Mark-V 35-cell 5-kW PEMFC at the
conditions. power output between 0.3 and 20 kW.
3820 C.-H. Lee, J.-T. Yang / Journal of Power Sources 196 (2011) 3810–3823

Fig. 14. Current–voltage curve of the Ballard-Mark-V 35-cell 5 kW PEMFC when the
Fig. 13. Output power of the Ballard-Mark-V 35-cell 5 kW PEMFC when the load load changes from full load to light load.
changes from full load to light load.
Table 4
Classification of light load, full load and overload of the designed AUV at each dc
Fig. 19 shows the temperature histories of the stack body, bus.
anode channel and cathode channel. The temperatures of these
Rated voltage Light load (20%) Full load (100%) Overload (120%)
three control volumes exhibit similar trends. During the first 300 s
(i.e., at light load), the temperatures slowly increase. The rea- DC 5V 1W 5W 6.25 W
son is that the heat generated by the electrochemical reaction is DC 12V 60 W 312 W 375 W
DC 24V 24 W 124 W 150 W
more than the heat transferred to the surrounding atmosphere
DC 48V 280 W 1400 W 1680 W
through convection. During the second stage (300–600 s) (i.e., load

Fig. 15. Output voltage and output current of 5 Vdc converter during a load change from 5 to 25 .

Fig. 16. Output voltage and output current of 12 Vdc converter during a load change from 0.46 to 2.4 .
 C.-H. Lee, J.-T. Yang / Journal of Power Sources 196 (2011) 3810–3823 3821

Fig. 17. Output voltage and output current of 24 Vdc converter during a load change from 0.46 to 24 .

Fig. 18. Output voltage and output current of 48 Vdc converter during a load change from 1.46 to 8.2 .

change from light to full load), the output power of the stack rent is lower, which leads to lower fuel/oxygen consumption for
increases due to the load current increase, and correspondingly the electrochemical reaction and higher mass of gases within the
the heat generated by the electrochemical reaction increases and channels. The opposite can be observed for the second stage. It
overcomes the heat loss due to convection. Thus, the tempera- is shown that hydrogen and oxygen decrease due to load current
tures of stack/anode/cathode all increase. During the last stage increase. This is because when the oxygen consumption increases,
(600–900 s) (i.e., load change from full to light load), the temper- the pressure within the cathode channel goes down and more
atures drop, as the output power and hence the heat generated oxygen will enter from the inlet. The inlet mass flow rates of
is reduced. It is worth mentioning that the temperatures of the hydrogen and oxygen are shown in Fig. 22. Correspondingly, their
three control volumes differ significantly. This clearly demonstrates
the advantage of dividing the fuel cell system into three control
volumes.
Fig. 20 shows the pressure variation within the two chan-
nels. When the load current is high (i.e., load change from light
to full load), the fuel consumed by the electrochemical reaction
increases. Correspondingly, the gas pressures go down. On the con-
trary, when the load current is low (i.e., load change from full
to light load), the pressures within the channels become rela-
tively high. Fig. 21 shows the change of mass of gases (hydrogen,
oxygen and nitrogen) within the two channels due to the load
current change. During the first stage and last stage, the load cur- Fig. 19. Temperature change at the anode channel, cathode channel and cell body.
3822 C.-H. Lee, J.-T. Yang / Journal of Power Sources 196 (2011) 3810–3823

Fig. 20. Pressure change at the anode and cathode channel.

Fig. 23. Mass consumption of hydrogen and oxygen.

6. Conclusions

This paper studies the electrical behavior of Ballard-Mark-V 35-

cell 5 kW PEMFC with its output voltage range from 27- to 40-V
based PWM dc–dc power conversion under load changes for appli-
cations in AUVs. The models of the studied fuel cell and PWM dc–dc
converters have been implemented in Matlab/SIMULINKTM envi-
ronment. The PEMFC model includes the double-layer charging
effect, gas diffusion in the electrodes, and the thermodynamic char-
acteristic while the operation of PWM dc–dc converter assumes to
be in CCM with a compensator in voltage-mode control. A series
of simulations show that the output voltages-based dc–dc power
conversion with the designed compensation network is stable as
Fig. 21. Hydrogen and oxygen change within the anode and cathode channels. the load changes from full load to light load. However, we regret
that the results of the simulations have not yet been compared with
actual measurements since the experimental work is in progress.

mass consumptions are shown in Fig. 23. The simulation results

References
presented here clearly illustrates the complicated dynamic inter-
actions between temperature and gas flow, and demonstrates the [1] J. Larminie, A. Dicks, Fuel Cell Systems Explained, 2nd ed., John Wiley & Sons,
necessity of developing a transient model using the control volume Inc., New York, 2003.
[2] Y. Chen, IEEE International Symposium on Underwater Technology, 2004, pp.
approach.
9–13.
[3] A.P. Meyer, Proceedings of OCEANS’93, vol. 2, Victoria, BC, Canada, October
18–21, 1993, pp. II/146-II/151.
[4] I. Yamamoto, T. Aoki, S. Tsukioka, H. Yoshida, T. Hyakudome, T. Sawa, S.
Ishibashi, T. Inada, K. Yokoyama, T. Maeda, S. Ishiguro, H. Hirayama, K.
Hirokawa, A. Hashimoto, N. Hisatome, T. Tani, Proceedings of OCEANS’94, vol.
3, November 9–12, 2004, pp. 1732–1737.
[5] “Autonomous Undersea Vehicle Applications Center” http://auvac.org/.
[6] P. Ray, H. Cook, Autonomous Underwater Vehicle, Senior Design Project,
Stevens Institute of Technology, 2006, April.
[7] W.T. Tsai, Nonlinear circuit analysis and control of a proton exchange mem-
brane fuel cell stack, Master Thesis, National Chung Hsing University, 2004.
[8] V.F. Pires, J. Fernando, A.A. Silva, IEEE Transactions on Education 45 (August (3))
(2002) 253–261.
[9] I. Sadli, P. Thounthong, J.-P. Martin, S. Rael, B. Davat, Journal of Power Sources
156 (May) (2006) 119–125.
[10] C. Wang, M.H. Nehrir, H. Gao, IEEE Transactions on Energy Conversion 21 (June
(2)) (2006) 586–595.
[11] S.K. Mazumder, IEEE Transactions on Aerospace and Electronic Systems 42
(January (1)) (2006) 50–69.
[12] X. Xue, J. Tang, A. Smirnova, R. England, N. Sammes, Journal of Power Sources
113 (2) (2004) 188–204.
[13] P.R. Pathapati, X. Xue, J. Tang, Renewable Energy 30 (May (1)) (2004) 1–22.
[14] F.M. White, Viscous Fluid Flow, 2nd ed., McGraw-Hill, 1991.
[15] C. Wang, M.H. Nehrir, S.R. Shaw, IEEE Transactions on Energy Conversion 20
(June (2)) (2005) 442–451.
[16] J.C. Amphlett, R.M. Baumert, R.F. Mann, B.A. Peppley, P.R. Roberge, Journal of
Electrochemical Society 142 (January (1)) (1995) 1–8.
[17] J.C. Amphlett, R.F. Mann, B.A. Peppley, P.R. Roberge, A. Rodrigues, Journal of
Power Sources 61 (July/August (1–2)) (1996) 183–188.
Fig. 22. Inlet mass flow rate of hydrogen and oxygen. [18] M.J. Khan, M.T. Iqbal, Fuel Cells (August (1)) (2004) 97–104.
 C.-H. Lee, J.-T. Yang / Journal of Power Sources 196 (2011) 3810–3823 3823

[19] M.J. Khan, M.T. Iqbal, Fuel Cells (May (4)) (2005) 463–475. [26] L. Palma, M.H. Todorovic, P. Enjeti, IEEE Transactions on Industrial Electronics
[20] T.V. Nguyen, R.E. White, Journal of Electrochemical Society 140 (8) (1993) 55 (March (3)) (2008) 1247–1255.
2178–2186. [27] B. Lau, Small-signal frequency response theory for ideal DC-to-DC converter
[21] A. Rowe, X. Li, Journal of Power Sources 102 (December (1)) (2001) 82–96. system, Ph.D. Thesis, California Institute of Technology, September 1986.
[22] S.K. Mazumder, S.K. Pradhan, J. Hartvigsen, M.R. von Spakovsky, D.F. Rancruel, [28] A. Khaligh, A. Emadi, IEEE Transactions on Aerospace and Electronic Systems
IEEE Transactions on Energy Conversion 22 (June (2)) (2007) 457–466. 44 (April (2)) (2008) 766–782.
[23] S. Pradhan, S.K. Mazumder, J. Hartvigsen, M. Hollist, ASME Journal of Fuel Cell [29] S. Moisseev, K. Soshin, M. Nakaoka, IEEE Transactions on Aerospace and Elec-
Science and Technology May (4) (2007) 154–166. tronic Systems 41 (January (1)) (2005) 174–180.
[24] S.K. Mazumder, K. Acharya, C.L. Haynes, R. William Jr., M.R. von Spakovsky, D.J. [30] H. Li, F.Z. Peng, IEEE Transactions on Aerospace and Electronic Systems 40
Nelson, D.F. Rancruel, J. Hartvigsen, R.S. Gemmen, IEEE Transactions on Power (January (1)) (2004) 272–283.
Electronics 19 (5) (2004) 1263–1278. [31] J.H. Su, C.M. Wang, J.J. Chen, J.D. Lee, T.L. Chen, IEEE PEDS (2005) 1256–1261.
[25] N. Mohan, Tore M. Undeland, W.P. Robbins, Power Electronics: Converters, [32] S.-Y. Choe, J.-W. Ahn, J.-G. Lee, S.-H. Baek, IEEE Transactions on Energy Conver-
Application, and Design, 3rd ed., Wiley, 2002, October 2002, ISBN: 978-0-471- sion 23 (June (2)) (2008) 669–680.
22693-2.