Professional Documents
Culture Documents
Abstract: In this paper, a 60 kW proton exchange membrane fuel cell (PEMFC) generation system is modeled in order to design
the system parameters and investigate the static and dynamic characteristics for control purposes. To achieve an overall system
model, the system is divided into five modules: the PEMFC stack (anode and cathode flows, membrane hydration, and stack
voltage and power), cathode air supply (air compressor, supply manifold, cooler, and humidifier), anode fuel supply (hydrogen
valve and humidifier), cathode exhaust exit (exit manifold and water return), and power conditioning (DC/DC and DC/AC)
modules. Using a combination of empirical and physical modeling techniques, the model is developed to set the operation con-
ditions of current, temperature, and cathode and anode gas flows and pressures, which have major impacts on system performance.
The current model is based on a 60 kW PEMFC power plant designed for residential applications and takes account of the elec-
trochemical and thermal aspects of chemical reactions within the stack as well as flows of reactants across the system. The
simulation tests show that the system model can represent the static and dynamic characteristics of a 60 kW PEMFC generation
system, which is mathematically simple for system parameters and control designs.
Key words: Proton exchange membrane fuel cell (PEMFC), Generation system, Dynamic simulation model, Reactant flow
doi:10.1631/jzus.A1000347 Document code: A CLC number: TM911.4
model of the air supply and a variable pressure con- midifier. The stack power is conditioned to match
trol strategy. Caux et al. (2005) considered anode and the user demand.
cathode compartments to control air and hydrogen
flow rates as well as pressure. For the stack voltage Pressure valve Humidifier PEMFC
Fig. 2 PEMFC stack model in MATLAB/SIMULINK where psat is vapor saturation pressure, a function of
WO2 ,ca,in , WN2 ,ca,in , and Wv,ca,in: oxygen, nitrogen, and temperature.
vapor mass flow in the cathode, respectively; pH2 ,an : Similar to the anode, the cathode pressure pca is
hydrogen partial pressure in the anode; pO2 ,ca : oxygen the sum of the oxygen, nitrogen, and vapor partial
partial pressure in the cathode; prm: return manifold pres- pressures. The relative humidity ca is determined
sure; ca,out and Rair,ca,out: humidity ratio and gas constant in from the vapor partial pressure and the vapor satura-
cathode outlet, respectively tion pressure. The oxygen in the air combines with
the electrons in the external circuit and the protons
characteristics. The model is applicable to PEMFC
flowing through the membrane, thus producing wa-
stacks of various configurations and operating con-
ter. Based on the mass conservation law, three state
ditions and has been validated by some experimental
equations of oxygen, nitrogen, and water in the
results (Zhang et al., 2005). The modules of anode
cathode are
and cathode flows aim to calculate internal stack
pressures and humidity. The anode operates in a sub-
dmO2 ,ca nI st
saturated condition, i.e., an<1, to prevent anode = WO2 ,ca,in WO2 ,ca,out M O2 , (5)
flooding in a 60 kW designed system. Thus, it is dt 2F
assumed that the anode purge is zero (Pukrushpan et dmN2 ,ca
= WN2 ,ca,in WN2 ,ca,out , (6)
al., 2002). The total anode pressure is the sum of dt
hydrogen and vapor partial pressures: dmv,ca nI st
= Wv,ca,in Wv,ca,out + M v + Wv,m , (7)
dt 2F
pan = pH2 ,an + pv,an
mH2 ,an RH2 Tst mv,an Rv Tst (1) where M O2 and Mv are the oxygen and vapor molar
= + ,
Van Van masses, respectively.
The total flow rate of the exhaust air from
where RH2 and Rv are the gas constants of hydrogen cathode Wca,out is determined using the simplified
orifice equation. Then the flow rates of oxygen, ni-
and vapor, respectively. Tst is the stack temperature
trogen, and vapor at the cathode exit WO2 ,ca,out ,
and Van is the anode volume. mH2 ,an and mv,an are the
masses of hydrogen and vapor in the anode, and can WN2 ,ca,out , and Wv,ca,out are all calculated.
be described based on the mass conservation law: The membrane hydration module is built for
calculating water flow across the membrane (Ge et
dmH2 ,an nI st al., 1999; Dutta et al., 2001; Pukrushpan et al.,
= WH2 ,an,in M H2 , (2)
dt 2F 2002), which occurs through three distinct phe-
dmv,an nomena: electro-osmotic drag, back-diffusion, and
= Wv,an,in Wv,m , (3) migration resulting from the pressure difference.
dt
Combining three water transports and approximating
where WH2 ,an,in is the hydrogen mass of the intake the water concentration gradient and pressure change
478 Zhang et al. / J Zhejiang Univ-Sci A (Appl Phys & Eng) 2011 12(6):475-482
in the membrane to be linear over the membrane and humidifier, where air varies in pressure, tem-
thickness, the water flow rate across the membrane is perature, and flow. A nozzle flow equation is used to
calculated from calculate the manifold outlet flow Wsm,out for a small
pressure difference between the manifold and the
I c cw,an stack cathode:
Wv,m = M v A n nd st Dw w,ca
AF tm
(8) Wsm,out = ksm,out ( psm pca ), (9)
kp pca pan
cf ,
tm where psm is the manifold pressure and ksm,out is the
nozzle constant.
where A is the cell active area, tm is the membrane Air cooler and humidifier modules describe the
thickness, kp is the hydraulic coefficient, is the air flow state according to thermodynamics and mass
water viscosity, cf is the sulfonic density, is the conservation principles.
water content in the membrane, nd is the electro-
osmotic drag coefficient, and Dw is the water diffu- 3.3 Anode fuel supply
sion coefficient. cw,an and cw,ca are the water concen- The anode fuel supply model includes the hy-
trations at the anode and cathode surfaces respec- drogen valve and humidifier. The pressure drop
tively, which are all calculated empirically from the across the hydrogen valve is reduced to the proportion
humidity. of the flow rate WH2 ,valve square. The hydrogen hu-
3.2 Cathode air supply midifier module is the same as that for the cathode air
The cathode air supply model includes an air supply. There is no purge on the anode side (Fig. 4).
compressor, supply manifold, cooler, and humidifier.
Tst
In the compressor module, the exit air temperature pH ,valve WH ,an,in
2
p hm,an,in
2
Tcp,out and the required power Pcp are both calculated Hydrogen Hydrogen
W v,an,in
WH ,valve valve humidifier
using thermodynamic equations (Fig. 3). The air 2
WH ,hm,in
mass flow rate Wcp is selected as a variable directly 2 W v,hm,an,inj
for the simple system model and control. The supply
manifold module represents the lumped volume as- Fig. 4 Anode fuel supply model in MATLAB/SIMULINK
sociated with pipes and connections from the com- WH2 ,valve : hydrogen mass flow through the second
pressor to the stack cathode, including the air cooler valve; pH2 ,valve : hydrogen pressure after the first valve;
phm,an,in: hydrogen pressure after the second valve; WH 2 ,hm,in :
Pcp pca
Tcp,out hydrogen mass flow in humidifier
Wcp Air
p sm Air
compressor
supply 3.4 Cathode exhaust exit
Wcp manifold
Wsm,out The cathode exhaust exit model includes the
WO ,hm
2
Wair,cool p atm exit manifold and water return (Fig. 5). The exit
WN ,hm Air W v,cool Air Tatm
2
manifold model represents pipes and connections as
W v,hm humidifier p air,cool cooler atm
for the air supply manifold, but its outlet flow is
Tst calculated based on the non-linear relationship for a
W v,hm,ca,inj
large pressure difference between the stack working
in the pressurization condition and the manifold exit.
Fig. 3 Cathode air supply model in MATLAB/
Air-water separation is the main process occurring in
SIMULINK
the cathode exhaust. It is assumed that 90% of water
WO2 ,hm , WN2 ,hm , and Wv,hm: oxygen, nitrogen, and vapor mass
can be separated by the ideal process. The water used
flows out of humidifier; Wair,cool and Wv,cool: air and vapor
flows into humidifier; pair,cool: air pressure out of cooler; patm, in a 60 kW PEMFC generation system can be de-
Tatm, and atm: pressure, temperature, and humidity in atmos- termined by comparing the water recovered from the
phere; Wv,hm,ca,inj: vapor mass flow injected into humidifier exhaust with that needed for reactant humidification.
Zhang et al. / J Zhejiang Univ-Sci A (Appl Phys & Eng) 2011 12(6):475-482 479
3.5 Power conditioning the water injection for air humidification Wv,hm,ca,inj,
the hydrogen flow rate through the second pressure
The power conditioning model includes DC/DC
and DC/AC (Fig. 6). The net power output of the relief valve WH2 ,valve , and the water injection for
system Pnet is that generated by the stack, Pst, minus hydrogen humidification Wv,hm,an,inj are all input
the parasitic load consumed, consisting mainly of the variables of the system model, that is, the adjustable
power consumed by the compressor, Pcp. Wide variables in the system operation. Thus, regardless of
fluctuations in stack voltage would make inverter whether the model is capable of describing the dy-
control difficult, so the low direct voltage of the stack namic characteristics of the 60 kW PEMFC genera-
Vst is boosted to Vdc by DC/DC, and then inverted to tion system, these variables are adjusted deliberately
alternating current voltage Vac and power Pout for user and the system simulation is tested.
load by DC/AC. Since the system model is complex in its
structure and performance, and there are a lot of
adjustable variables and monitoring parameters, the
4 Generation system simulation test dynamics of the anode and cathode sides of the sys-
tem model were tested separately to provide more
The 60 kW PEMFC generation system model detailed results.
and simulation test interface is shown in Fig. 7. The Assuming the cathode side is steady, the simu-
system current Ist is instantaneously drawn from the lation test on the anode side was conducted with
load representing the user demand. The system cur-
rent Ist, the air flow rate at the compressor exit Wcp,
Vst Pdc
Pout
Wca,out
ppr,ca
Pnet Vdc
DC/DC DC/AC
Rair,ca,out converter inverter Vac
Cathode
exit dc
ca,out Wv,rm,ca
manifold Water
return d ac u
Tst
Fig. 5 Cathode exhaust exit model in MATLAB/ Fig. 6 Power conditioning model in MATLAB/
SIMULINK SIMULINK
Wv,rm,ca: vapor mass flow that can be returned; ppr,ca: pressure Pdc: power of DC/DC; dc and ac: conversion efficiencies
of the cathode return manifold of DC/DC and DC/AC; d: duty ratio; u: modulation factor
p O ,ca Wv,rm,ca
pca ca 2
Wsm,out
Stack Vst
psm
Wcp Air Air supply Membrane voltage Pst DC/DC
compressor Tcp,out manifold Wv,m hydration and converter
power
an pan Vdc Pdc
Pcp
WH ,valve WH ,an,in Anode
phm,an,in 2
Hydrogen Hydrogen pH ,an
2
inverter
Vac Pout
Wv,hm,an,inj
variable inputs WH 2 ,valve , Wv,hm,an,inj, and Ist. Simula- 3. According to the principle of PEMFC, the
water mass in the anode is affected by the water
tion results are shown in Fig. 8.
injection into the hydrogen. The partial pressure of
1. At t=6 and 12 s, the anode pressure pan, hu-
the water changes. The anode humidity, pressure,
midity an and stack voltage Vst all drop off dra-
and stack output all change eventually. At t=10 and
matically due to a step increase of Ist, while the stack
16 s, there are sudden increases in pan, an, Vst, and Pst
power Pst increases dramatically. In contrast, at
with the increase in Wv,hm,an,inj (Fig. 8). But at t=22 s,
t=18 s, pan and Vst increase suddenly with the sharp
there is no obvious dynamic variation with the sud-
decline in Ist, an is maintained at a saturation state,
den decrease in Wv,hm,an,inj. This situation is due to
and Pst drops off. These situations are mainly due to
long term water super saturation in the anode.
more hydrogen being consumed in the electro-
The simulation results above show that the
chemical reaction as Ist increases and more water
model describes well the dynamic characteristics and
flow to the cathode across the membrane. The hy-
the effects of variables on the stack output perform-
drogen and water then both decrease in the stack
ance in the anode side.
anode, leading to decreases in both pan and an. When Next, the simulation test of the cathode side was
Ist decreases, the opposite process occurs. The rela- conducted with variable inputs Ist, Wcp, and
tionships between Vst, Pst, and Ist conform to the stack Wv,hm,ca,inj, assuming the system anode side was
output polarization characteristics. steady. The simulation results are shown in Fig. 9.
2. The dynamic response in the hydrogen supply 1. At t=2 s, Ist increases suddenly, followed by a
is shown in Fig. 8. WH2 ,valve increases suddenly at gentle rise in cathode humidity ca. ca increases and
t=8 and 14 s, and then pan increases. This is because decreases with the step increase of Ist at t=12 s and
the hydrogen mass increases in the anode and the the decrease at t=22 s. All these situations are due to
hydrogen partial pressure increases. At t=20 s, the increase and decrease of the water generated in
the cathode and the flow across the membrane from
WH2 ,valve decreases suddenly, and then the hydrogen
anode to cathode when the electrochemical reaction
mass and partial pressure in the anode both decrease. rate increases and decreases. Observing the effect of
The increase in pan clearly slows down. Ist on the cathode pressure pca, at t=2 s, pca stops
pan (105, Pa) WH ,valve (103, kg/s)
600 2
(a) (b)
400
Ist (A)
1
200
0 0
0 6 12 18 24 0 8 14 20 24
2
4.0
0.04
(c) (d)
Wv,hm,an,inj (kg/s)
3.5
0.02 3.0
0 2.5
0 10 16 22 24 0 6 8 10 12 14 16 18 20 22 24
1.0 250
(e) (f)
0.8
Vst (V)
an
200
0.6
0.4
150
0 6 8 10 12 14 16 18 20 22 24 0 6 8 10 12 14 16 18 20 22 24
8 Time
(g) Time (s) (s)
Pst (kW)
4
2
0 6 8 10 12 14 16 18 20 22 24
Time (s)
400 1.0
(a) (b)
Wcp (kg/s)
300 0.8
Ist (A)
200 0.6
100 0.4
0 2 12 22 28 0 4 14 24 28
8
0.04 (c) (d)
6
0.02
4
0 2
0 6 16 26 28 0 2 4 6 12 14 16 22 24 26 28
1.0 220
(e) (f)
Vst (V)
200
ca
0.5
180
0 160
0 2 4 6 12 14 16 22 24 26 28 0 2 4 6 12 14 16 22 24 26 28
80
(g) Time (s)
Pst (kW)
60
40
20
0 2 4 6 12 14 16 22 24 26 28
Time (s)
increasing with the sudden increase of Ist, because the cathode flow. According to the principle of
more oxygen is consumed in the electrochemical PEMFC, Vst should increase with pca, but there are no
reaction in the cathode, and then the oxygen partial obvious changes in Vst and Pst at t=14 s. We can see
pressure decreases. But at t=12 and 22 s, there is no that ca then begins to decrease, which affects the
clear change in pca. This can be explained as follows. response of Vst.
The cathode pressure pca is the sum of the partial 3. At t=6, 16, and 26 s, with the sudden decrease
pressures of nitrogen, oxygen, and water. The elec- or increase in Wv,hm,ca,inj, there is a sudden decrease or
trochemical reaction rate increases with the current, increase in ca. These changes are due to the decrease
which causes the oxygen partial pressure to decrease and the increase of the water mass, or water partial
as more oxygen is consumed, and the water partial pressure in the cathode flow. The relationships be-
pressure to increase as more water is generated and tween Vst, Pst, and ca conform to the stack operation
flows from the anode to the cathode. The water par- principle.
tial pressure is also affected by the pressure differ- Based on these simulation results, the dynamic
ence between the anode and cathode. As a result, characteristics of the cathode side are more complex
there is a total effect of Ist on pca, which is not simple than those of the anode side. This is because of more
as pca increases with Ist or decreases with Ist. The equipment, a longer supply manifold, the presence of
relationships between Vst, Pst, and Ist are observed an exhaust exit, and the use of air, not oxygen, as the
especially at t=2, 12, and 22 s, which conform to the oxidant on the cathode side. The responses of the
stack output polarization characteristics. cathode pressure, humidity, and the stack voltage and
2. As shown in Fig. 9, pca increases or decreases power to all the adjustable parameters in the simula-
suddenly at t=4, 14, and 24 s with the step increase or tion test were in agreement with the system operating
decrease in Wcp, due to the sudden increase or de- rules. The system model describes well the dynamic
crease in oxygen mass, or oxygen partial pressure, in characteristics of the cathode side.
482 Zhang et al. / J Zhejiang Univ-Sci A (Appl Phys & Eng) 2011 12(6):475-482
5 Conclusions Dutta, S., Shimpalee, S., Van Zee, J.W., 2001. Numerical
prediction of mass-exchange between cathode and anode
A 60 kW PEMFC generation system is currently channels in a PEM fuel cell. International Journal of
at the stage of design and simulation research. Sys- Heat and Mass Transfer, 44(11):2029-2042. [doi:10.
1016/S0017-9310(00)00257-X]
tem simulation tests show that the system model built Ge, S.H., Yi, B.L., Xu, H.F., 1999. Model of water transport
in a MATLAB/SIMULINK environment can ap- for proton-exchange membrane fuel cell (PEMFC).
proximate the dynamic responses of the generation Journal of Chemical Industry and Engineering, 50(1):
system. The simulation system is consistent with 39-48 (in Chinese).
published data and operational experience of the Hatti, M., Tioursi, M., 2009. Dynamic neural network con-
system performance and characteristics, and can be troller model of PEM fuel cell system. International
used for system dynamic analyses and design. Journal of Hydrogen Energy, 34(11):5015-5021. [doi:
10.1016/j.ijhydene.2008.12.094]
Moreover, the system simulation tests show that the Pukrushpan, J.T., Stefanopoulou, A.G., Huei, P., 2002. Mod-
dynamic behavior of a PEMFC generation system is eling and Control for PEM Fuel Cell Stack System. Pro-
very complex, so a comprehensive control design is ceedings of the American Control Conference, Anchor-
very important for achieving ideal operation and high age, Alaska, USA, p.3117-3122. [doi:10.1109/ACC.
efficiency. 2002.1025268]
Rodatz, S., Paganelli, G., Guzzella, L., 2003. Optimizing Air
References Supply Control of a PEM Fuel Cell System. Proceedings
Caux, S., Lachaize, J., Fadel, M., Shott, P., Nicod, L., 2005. of the American Control Conference, Denver, USA,
Modeling and control of a fuel cell system and storage p.2043-2048. [doi:10.1109/ACC.2003.1243375]
elements in transport applications. Journal of Process Yang, Y.P., Liu, Z.W., Wang, F.C., 2008. An application of
Control, 15(4):481-491. [doi:10.1016/j.jprocont.2004.08. indirect model reference adaptive control to a low-power
002] proton exchange membrane fuel cell. Journal of Power
Chrenko, D., Pra, M.C., Hissel, D., Geweke, M., 2008. Sources, 179(2):618-630. [doi:10.1016/j.jpowsour.2008.
Macroscopic modeling of a PEFC system based on 01.053]
equivalent circuits of fuel and oxidant supply. Journal of Zhang, Y.Y., Cao, G.Y., Zhu, X.J., 2005. Development of a
Fuel Cell Science and Technology, 5(1):011015. [doi:10.
generalised steady-state voltage model for PEMFC.
1115/1.2786471]
Chrenko, D., Pra, M.C., Hissel, D., Bouscayrol, A., 2009. Journal of Power Sources, 29(6):402-405 (in Chinese).
Inversion-based control of a proton exchange membrane Zhang, Y.Y., Yu, Q.C., Cao, G.Y., Zhu, X.J., 2006. Research
fuel cell system using energetic macroscopic represen- on a simulated 60 kW PEMFC cogeneration system for
tation. Journal of Fuel Cell Science and Technology, domestic application. Journal of Zhejiang University-
6(2):024501. [doi:10.1115/1.2971133] Science A, 7(3):450-457. [doi:10.1631/jzus.2006.A0450]