Journal of Power Sources 466 (2020) 228311

Contents lists available at ScienceDirect

Journal of Power Sources

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

Improved water management by alternating air flow directions in a proton

exchange membrane fuel cell stack
Lizhong Luo , Bi Huang , Zongyi Cheng , Qifei Jian *
School of Mechanical and Automotive Engineering, South China University of Technology, Guangzhou, 510641, Guangdong, China

H I G H L I G H T S

� The proposed method can significantly increase the stack output voltage.
� Alternating air flow directions causes a redistribution of the water.
� This method can achieve self-humidification without any external humidifiers.

A R T I C L E I N F O A B S T R A C T

Keywords: This work presents the water management improvement of a water-cooled proton exchange membrane fuel cell
Proton exchange membrane fuel cell stack with alternating air flow directions. This method only needs to add two three-way solenoid valves and
Water management several pipelines to the original system. The effects of different alternating modes, different alternating intervals
Self-humidification
and different hydrogen purge periods on the stack performance are experimentally studied. The advantages and
Alternating air flow directions
Voltage
disadvantages of the scheme have also been discussed in detail. Experimental results confirm that the method can
significantly increase the stack voltage. When the operating temperature is 65 � C and the current density is 0.8
A/cm2, the stack voltage can be improved by 19.46%. Alternating air flow directions causes a redistribution of
the water inside the stack. This can make the water distribution more uniform and achieve self-humidification.
The proposed solution can also increase air pressure. However, alternating air flow directions will make the stack
voltage fluctuate significantly. The alternating interval should be a trade-off between the fluctuation amplitude
and the fluctuation frequency. The amplitude of the stack voltage fluctuation cannot be reduced by adjusting the
purge period. This research provides a new reference for the water management improvement.

1. Introduction Some researchers explore modifying material properties to improve

water management in PEMFCs. Cha et al. [6] compared the performance
In recent years, the commercialization of proton exchange mem­ of membranes with short-side-chain and long-side-chain on
brane fuel cells (PEMFCs) is accelerating. There are still some challenges self-humidification PEMFCs and found that the membrane with
that researchers need to solve, such as higher performance [1], lower short-side-chain has higher power density due to higher water retention.
cost, smaller system size, durability [2] and stability. Water manage­ Li et al. [7] shown that the thin membrane is easy to be hydrated under
ment is closely related to these challenges [3]. Proper water manage­ low humidity condition. Jang et al. [8] deposited a functional
ment ensures that the membrane is sufficiently hydrated to maintain carbon-covering layer onto the cathode catalyst layer to retain gener­
high ionic conductivity. It requires to avoid flooding caused by too much ated water in the membrane electrode assembly. Ferreira et al. [9] found
water. In addition, it also requires to prevent the membrane from drying that gas diffusion layer hydrophobic treatment facilitates proper hu­
out. Water management failure can cause flooding or drying out, midification of the cell and can significantly improve the cell perfor­
resulting in voltage decay [4]. Therefore, more researches on water mance. They also pointed out that microporous layers can create a
management are needed. Material performance improvement, structural pressure barrier for water produced at the cathode and force water to
design and optimized operating conditions are the main methods for travel to the anode side. A fuel cell containing a cathode microporous
improving water management [5]. layer with large pores can effectively provide water pathways, thereby

* Corresponding author.
E-mail address: [email protected] (Q. Jian).

https://doi.org/10.1016/j.jpowsour.2020.228311
Received 13 January 2020; Received in revised form 7 April 2020; Accepted 4 May 2020
Available online 15 May 2020
0378-7753/© 2020 Elsevier B.V. All rights reserved.
L. Luo et al. Journal of Power Sources 466 (2020) 228311

enabling the fuel cell to obtain superior performance [10]. However, cell.
thick microporous layer will reduce the performance of fuel cells [11]. Although there are a lot of discussion about water management,
Hasanpour et al. [12] investigated the impact of woven gas diffusion there are some issues that need to be addressed when improving water
layers at different fluorinate ethylene propylene loading on water management and fuel cell performance. First, developing new materials
transport, distribution and location of breakthrough. They found that or modifying material properties requires significant costs. Second, the
increasing fluorinate ethylene propylene loading promotes the propa­ external humidification can improve performance but require humidi­
gation of water inside gas diffusion layers and improves the thermal fiers, such as bubbling humidifiers [22], enthalpy wheel humidifiers
conductivity of gas diffusion layers, but the electrical conductivity de­ [23] and membrane humidifiers [30]. This increases the size, weight
creases. Kim et al. [13] attached a secondary porous layer to a bi-polar and cost of the PEMFC system. Third, the commonly used method of
plate to form a bi-porous layer flow-field. They proved that the bi-porous self-humidification by exhaust gas recirculation has the potential to
layer flow-field enables liquid water redistribution to improve water reduce the PEMFC performance. To solve these problems, a low-cost
management. Although modifying material parameters can significantly method that can achieve self-humidification and improve performance
improve the performance of PEMFCs, applying them to large-scale is proposed in this study.
commercialization requires more investment and a longer time. In this In this study, a low cost air supply operation strategy is proposed to
case, make reasonable adjustments to the existing system structure and improve water management. The proposed method is based on the
select appropriate operating strategies are convenient and effective concept of modifying the PEMFC system physical structure and opti­
methods. mizing the operating conditions. This method only needs to add two
Flow field structure design has a great impact on water management three-way solenoid valves and several pipelines to the original system. It
[14]. Wu et al. [15] investigated the characterization of water man­ can achieve self-humidification and obtain a more uniform water dis­
agement in metal foam flow field using in-operando neutron radiog­ tribution. It can also increase air pressure under the same air flow rate
raphy and found that the fuel cell with the metal foam flow-field exhibits condition. In this way, the performance of the PEMFC system is
a ~50% increase in voltage compared to the serpentine flow field enhanced. The proposed method is simple and effective for water
design. This is related to water being “stored” in the pores of the metal management improvement.
foam. Xu et al. [16] investigated wave-like gas channels and found that
the geometric parameters have an obvious impact on the water accu­ 2. Experimental
mulation process. Liu et al. [17] designed a slope turn in the serpentine
flow channel and investigated the water behavior in it by 2.1. Description of alternating air flow directions
volume-of-fluid method. They found that the presence of the designed
slope turn can enhance water removal from the membrane electrode The main purpose of alternating air flow directions is to achieve a
assembly surface. He et al. [18] machined a series of three-dimensional more uniform water distribution and self-humidification. This can
flow fields with trapezoid-shaped baffles and proved that the designed improve the water management of PEMFC stacks, thereby increasing the
flow fields can improve water management. They also pointed that the output voltage. Two identical three-way solenoid valves (JSF2637003,
flow fields with baffles will increase parasitic power. For existing PEMFC TCS Precision Technology Co., Ltd, China) with three ports were used to
systems, it is very convenient and effective to set reasonable operating alternate the air flow direction, as shown in Fig. 1(a). These solenoid
conditions for water management improvement. Shao et al. [19] shown valves are available in existing products, which can reduce the design
that the co-flow configuration will result in a low water content near the cost. When the three-way solenoid valve is not energized, Port C and
air inlet area and the counter-flow configuration can avoid this problem. Port D (Port G and Port H) are connected. When the solenoid valve is
Wang et al. [20] shown that the counter-flow configuration greatly in­ energized, Port C and Port E (Port F and Port H) are connected. Only one
creases the current density by 21.21% compared with the co-flow three-way solenoid valve needs to be powered during the system oper­
configuration. Ko et al. [21] studied the changes in water distribution ation. First, the three-way solenoid valve on the right in Fig. 1(a) is
under different pressure conditions and shown that cathode side pres­ energized and the three-way solenoid valve on the left in Fig. 1(a) is not
surization can increase water saturation in the membrane. Kuhn et al. energized. Then, Port C and Port E are connected. Port D and Port F are
[22] mentioned that the humidification system with a water bubbling closed and the air cannot pass. Therefore, the air flow path is
device also requires additional heaters to ensure the gas has correct dew G→H→A→B→C→E (Red solid line in Fig. 1(a)). The remaining air is
point. Casalegno et al. [23] suggested that the optimal enthalpy wheel is discharged from Port E into the surroundings. At this time, the air flow
characterized by peculiar geometries and revolution speed. This requires direction inside the PEMFC stack is from A to B. When switching the air
a special additional integrated design of the PEMFC system. These flow direction, the three-way solenoid valve on the left in Fig. 1(a) is
external humidification methods will increase the size, weight and cost energized and the three-way solenoid valve on the right in Fig. 1(a) is
of the PEMFC system. Optimizing operating conditions to achieve not energized. Then, Port F and Port H are connected. Port E and Port G
self-humidification can avoid additional equipment and no parasitic loss are closed and the air cannot pass. Therefore, the air flow path becomes
is consumed [24]. Rodosik et al. [25] investigated a PEMFC system with D→C→B→A→H→F (Black dotted line in Fig. 1(a)). The remaining air is
cathode exhaust gases recirculation and found that high to medium discharged from Port F into the surroundings. At this time, the air flow
recirculation ratios are required to achieve stable and uniform cell direction inside the PEMFC stack is from B to A. A controller (V51,
voltage. They also pointed out that cathode gas recirculation will reduce Shenzhen Yahboom Technology Co., Ltd. China) with a time control
oxygen concentration and result in performance degradation. Xu et al. accuracy of 0.01 s was used to control these solenoid valves. The control
[26] used a recirculation valve to mix the dry air with the moisty exhaust strategy is very simple. Only need to power these three-way solenoid
gas at the inlet to achieve self-humidification. They found that the valves in turn. Alternating air flow directions can be achieved by
exhaust gas recirculation ratio has a great impact on system perfor­ switching the air flow direction at regular intervals.
mances. Zhao et al. [27] achieved self-humidification by the exhaust gas The alternating interval is the time interval for switching the air flow
recirculation but the system performance was degraded. Kim et al. [28] direction. The trial and error method is used to determine the alternating
experimentally studied a 5-cell PEMFC stack with an active area of 250 interval. The criterion for the alternating interval is that under the same
cm2 under the cathode exhaust gas recirculation condition and found current density condition, the higher and more stable the output voltage,
that the voltage of the gas recirculation was 6.75% less than that of the the more reasonable the alternating interval is.
reference case. Ramiar et al. [29] conducted a numerical study to The water and reaction gases distribution inside the PEMFC is related
investigate the effect of the pulsation air flow on water management and to the two-phase flow in the flow channel and the transient in the two-
proved that the method is effective for improving water removal from phase flow usually takes 1–5 min to reach a new steady operation state

2
L. Luo et al. Journal of Power Sources 466 (2020) 228311

Fig. 1. Test PEMFC stack. (a) Description of alternating air flow directions, (b) air flow direction switching occurs at the end of the hydrogen purge operation,
Alternate 1, (c) air flow direction switching is in the middle of the purge interval, Alternate 2.

[31]. If the alternating interval is greater than the time required for the droplets and flowing with the air. This enables a more uniform water
two-phase flow to stabilize, the distribution of the water and reaction distribution. So the alternating interval does not exceed the time that the
gases will show fluctuating variations during the whole operation. When two-phase flow reaches a new steady state. In this study, the PEMFC
the water distribution is stable, larger water droplets will accumulate stack runs in a dead-ended anode mode with a purge period of 30.50 s.
near the flow channel outlet. Then, the air will bring a lot of water back Considering that the cathode air flow direction changes and the anode
to flow channels due to the air flow direction reverses. This will cause hydrogen purge operation may interact, the minimum alternating in­
more water to accumulate in the flow path, resulting in flooding. The terval is chosen to be 15.25 s. Therefore, the alternating interval for tests
water droplets in flow channels are relatively small before the two-phase are 15.25 s, 30.50 s, 45.75 s, 61.00 s, 76.25 s and 91.50 s.
flow reaching stability. This is a reasonable assumption, because in To study the relationship between cathode air flow direction changes
general, the water droplets in flow channels gradually increase. Small and anode hydrogen purge operation, there are two modes to set the
droplets are more easily crushed by reverse airflow, forming smaller alternating interval. As shown in Fig. 1(b), the air flow direction

3
L. Luo et al. Journal of Power Sources 466 (2020) 228311

switching occurs at the end of the hydrogen purge operation. At this The PEMFC stack used in the experiment is a water-cooled stack with
moment, both the hydrogen and the air change to a new state, that is, the a cell active area of 80 cm2. It contains 20 single cells, as shown in Fig. 1
hydrogen stops discharging and the air flow direction is reversed. For (a). A membrane (DuPont Nafion® HP) with a catalyst loading at anode
the convenience of analysis, this mode is named Alternate 1. Another of 0.27 mg Pt/cm2 and at cathode of 0.53 mg Pt/cm2 was used. Multiple
mode is to switch the air flow direction during the purge interval, as serpentine flow fields at the anode and the cathode side were designed
shown in Fig. 1(c). Air flow direction switching is in the middle of the on graphite bipolar plates. At the bend of the cathode serpentine
purge interval, which is 15 s after the end of each purge operation. This channel, grooves were provided for inserting thermocouples (GG-K-30,
mode is named Alternate 2. Omega Engineering, INC. U.S.A.), as shown in Fig. 2(c). Thus, thermo­
couples were disposed on the cathode gas diffusion layer surface.
2.2. Experimental set-up Inserted thermocouples do not block air flow. Five cells were selected for
inserting thermocouples. These five cells were Cell 2, Cell 6, Cell 10, Cell
The original system refers to the PEMFC stack system in which the air 14 and Cell 18 (The cell number near the air inlet side is 1). Sixteen
flow direction does not change during operations, as shown in Fig. 2(a). thermocouples were placed in each cell and a total of eighty thermo­
The modified system refers to the PEMFC stack system in which the air couples were placed in the PEMFC stack. Parallel flow field was used for
flow direction changes alternately during operations, as shown in Fig. 2 the coolant (deionized water). The membrane electrode assembly pa­
(b). rameters and channel dimensions are listed in Table 1.

Fig. 2. Experimental set-up. (a) The original system, (b) the modified system, (c) cathode flow field and thermocouples.

4
L. Luo et al. Journal of Power Sources 466 (2020) 228311

Dry hydrogen having a purity of 99.999% and a pressure of 160 kPa 30.50 s. The air alternate mode was Alternate 1. Second, the effect of the
(absolute pressure) was supplied to the anode. A solenoid valve (2V025- alternating interval on the stack voltage was tested under the current
08, ZCKM. China) was located at the anode outlet. The purge time and density condition of 0.8 A/cm2. The reason for choosing the current
purge interval was 0.50 s and 30.00 s, respectively. Air at an average density of 0.8 A/cm2 is that it is the rated operating current density of
temperature of 33.2 � C and an average relative humidity of 60% was the PEMFC stack. Third, the effect of different purge periods on the stack
provided by an air compressor (DT550W-30L, DAERTUO, China). The voltage was tested under the condition of alternating air flow directions.
air flow rate was 46.4 SLPM (Standard Liter Per Minute) (the air stoi­ This was also tested at the current density of 0.8 A/cm2. The purpose of
chiometry was 2.18 at 0.8 A/cm2). The manufacturer of the air mass this test is to explore the effect of the purge period on the stability of the
flowmeter (CAFS5008B, Consensic Inc. China) provided errors in the stack voltage under the condition of alternating air flow directions. In
device as �1.5% of the full scale which was 100 SLPM. Therefore the this test, the purge time remained constant at 0.50 s. The purge period
error in air flow rate reading was �1.5 SLPM. Pressure sensors (LST-131, was 10.50 s, 15.50 s, 20.50 s, 25.50 s and 30.50 s, respectively. The
Qingdao LST Instrument Co., Ltd. China) and relative humidity modules detailed test conditions are listed in Table 2. The results obtained can be
(HM1500LF, Measurement Specialties, Inc. MEAS France) were pro­ used to analyze the effect of alternating air flow directions on the water
vided at the cathode inlet and outlet for measuring the pressure and management of the PEMFC stack.
relative humidity of the air. The manufacturers of these devices pro­ The detailed test steps are as follow: The first step was to connect air
vided an accuracy of 1 kPa and �3% for the pressure and the relative lines of the fuel cell system and set the controller according to the
humidity, respectively. Coolant was supplied by a water pump selected air alternate mode. After that, the cooling water was heated to
(QJM81C, Changzhou Wanhe Mechanical & Electrical Co., Ltd. China). the required temperature and passed into the PEMFC stack to maintain
The coolant flow rate was set to 2 L/min. This PEMFC stack achieves the operating temperature. Then, the air and the hydrogen were sup­
different operating temperatures by adjusting different coolant inlet plied for the start-up of the PEMFC stack. Note that reaction gases were
temperatures. In this study, the cooling water inlet temperature is not humidified before entering the stack. Next, the PEMFC stack was
considered as the operating temperature of the PEMFC stack. An elec­ operated at a current load set by the electronic load. At the same time,
tronic load (JT6331A, Nanjing Jartul Electronics Co., Ltd. China) was voltage and pressure data were recorded. We set different test conditions
used to make the PEMFC stack operate under galvanostatic mode. The according to Table 2 and repeated the above steps to complete all the
current was measured by the electronic load with an accuracy of 10 mA. tests. In addition, the experiment was repeated at 45 � C and proved to be
Temperature and cell voltage data were read by a data acquisition in­ repeatable.
strument (GM10, Yokogawa Electric Corporation, Japan). The manu­
facturer of this unit provided an accuracy of 0.2 � C and 1 mV for the 3. Results and discussion
temperature and the voltage, respectively. Temperature and cell voltage
were recorded at 2 Hz. The uncertainty in the power is �0.3%. The 3.1. Performance characterizations
resulting uncertainty in the power can be calculated as [32].
� qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi Fig. 3 shows the stack voltage variation with the current density at
δpower POWER ¼ � ðδI =IÞ2 þ ðδV =VÞ2 (1) different temperatures. The lines with solid symbols are the polarization
curve of the original system with constant air flow direction. The lines
with hollow symbols are the polarization curve of the modified system
where I is current, A; V is voltage, V.
with alternating air flow direction. Solid lines refer to the stack voltage.
Dotted lines refer to the stack power.
2.3. Experimental procedure Under the same current density conditions, the voltage and power of
the modified system are higher than those of the original system, as
There are three sets of tests in total. All the tests were performed on shown in Fig. 3. As the current density increases, the performance dif­
the same PEMFC stack. First, the polarization curve under different ference between the modified system and the original system becomes
temperature conditions were tested to verify the effectiveness of the larger and larger. Under the rated operating current density condition of
proposed scheme (alternating air flow directions). The stack ran steadily 0.8A/cm2, when the stack operating temperature is 45 � C, the stack
for 3 min under each current density condition and the average voltage voltage is increased from 11.719 V to 12.113 V, an increase of 3.36%;
of the last minute was taken as the points on the polarization curves. The when the stack operating temperature is 55 � C, the stack voltage is
optimal operating temperature to establish a stable behavior of the increased from 11.817 V to 12.431 V, an increase of 5.20%; when the
PEMFC stack without reaction gases humidification was 55 � C according stack operating temperature is 65 � C, the stack voltage is increased from
to the previous experimental results. The previous experimental results 9.566 V to 11.428 V, an increase of 19.46%. The polarization curve
come from a separated test campaign and its purpose is to select the results clearly demonstrate that the proposed method of alternating air
appropriate test temperature for subsequent experiments. Higher tem­ flow directions can significantly increase the output voltage of the
peratures will make the cell inside the stack tend to dry out. Therefore, PEMFC stack. Especially for higher operating temperature, the effect of
these test temperatures were 45 � C, 55 � C and 65 � C, respectively. At this
time, the alternating interval was set the same as the purge period,
Table 2
Test conditions.
Table 1
Experiment Test group 1 Test group 2 Test group 3
Membrane electrode assembly parameters and channel dimensions.
Polarization Different Different purge
Parameters Value Parameters Value curves alternating periods
(mm) (mm) intervals

Thickness of anode gas 0.235 Thickness of anode 0.045 Coolant inlet 45, 55, 65 55 55
diffusion layer microporous Layer temperatures (� C)
Thickness of cathode gas 0.235 Thickness of cathode 0.045 Current density 0–1.1 0.8 0.8
diffusion layer microporous Layer conditions (A/cm2)
Thickness of membrane 0.020 Anode channel width 1 Air alternate mode Alternate 1 Alternate 1 Alternate 1
Anode channel height 0.3 Anode channel rib width 0.9 Alternate 2
Cathode channel width 1 Cathode channel height 0.5 Alternating interval 30.50 15.25–91.50 30.50
Cathode channel rib width 1 Coolant channel width 2 (s)
Coolant channel height 0.5 Coolant channel rib width 2 Purge period (s) 30.50 30.50 10.50–30.50

5
L. Luo et al. Journal of Power Sources 466 (2020) 228311

Fig. 3. Polarization curve.

improving performance is more obvious. smaller water content in the membrane and an increase in the mem­
The proposed method of alternating air flow directions can achieve brane resistance, which reduces the output voltage. For the current
self-humidification without any external humidifiers, thereby improving density of 0.8 A/cm2, when the operating temperature was 45 � C and 55
the output voltage of the PEMFC stack. As shown in Fig. 3, for the �
C, there was obvious liquid water discharge at the air outlet in the
original system without humidification and constant air flow direction, original system. When the operating temperature was 65 � C, less liquid
the PEMFC stack achieved the best performance at the operating tem­ water was discharged from the air outlet in the original system. These
perature of 55 � C (red line with solid triangles in Fig. 3). When the phenomena in the experiment can also prove that there is less moisture
operating temperature of the PEMFC stack is increased to 65 � C, the inside the PEMFC stack at higher temperatures. However, when two
stack voltage is greatly reduced (blue line with solid circles in Fig. 3). three-way solenoid valves were added to the cathode side of the original
This phenomenon is related to the membrane hydration. Under normal system, the performance of the stack at 65 � C was greatly improved, as
operating conditions, the membrane has to absorb enough water to shown in Fig. 3. The higher the current density, the more the output
maintain high ionic conductivity [33]. In this study, the air and the voltage is increased. Because the higher the current density, the more
hydrogen were not humidified before entering the PEMFC stack. water is generated by the electrochemical reaction. Therefore, alter­
Therefore, the hydration of the membrane mainly depends on the water nating air flow directions can increase the moisture inside the PEMFC
generated by the electrochemical reaction. Without humidification, the stack and make the membrane better hydrated. It achieves
fuel cell performance depends strongly on the operating temperature self-humidification. As shown in Fig. 3, the PEMFC stack with alter­
[34]. According to the relationship between temperature and the water nating air flow direction can achieve better performance at higher
vapor saturation pressure, water saturation pressure is higher at higher temperatures. When the operating temperature of the PEMFC stack is 45
temperatures. The low water saturation pressure at low temperature �
C, the maximum power of the stack increased from 792.3 W to 845.1 W,
makes the water condensation easier [35]. Higher operating tempera­ an increase of 6.66%. When the operating temperature of the PEMFC
ture will decrease water absorption in the membrane. This results in a stack is 55 � C, the maximum power of the stack increased from 830.6 W

6
L. Luo et al. Journal of Power Sources 466 (2020) 228311

to 903.6 W, an increase of 8.79%. When the operating temperature of stack with the air. For the situation where there is no alternating air flow
the PEMFC stack is 65 � C, the maximum power of the stack increased directions, the water on the cathode side is distributed along the cathode
from 664.9 W to 862.6 W, an increase of 29.73%. The effect of flow path from the inlet to the outlet and the content is increasing. The
self-humidification becomes more pronounced at higher operating relative humidity increases continuously along the flow direction [19].
temperatures. When the air flow direction is reversed, the downstream of the original
Alternating air flow directions causes a redistribution of the water cathode flow path becomes upstream. The water accumulated in the
inside the PEMFC stack. In general, the water in the cathode flow downstream and near the outlet is no longer discharged out of the
channel will gradually accumulate downstream and near the outlet due PEMFC stack, but flows in the flow channel in the original opposite flow
to the flow of the air. The water can be discharged out of the PEMFC direction. In other words, the reversal of the air flow direction causes the

Fig. 4. Variation of pressure and pressure drop. (a) Pressure variation, (b) cathode pressure drop variation.

7
L. Luo et al. Journal of Power Sources 466 (2020) 228311

water that is originally discharged to flow back into the stack, increasing cell potential due to the Nernst equation and on the exchange current
the water content inside the PEMFC stack. This corresponds to the density. Exchange current density is proportional to surface concentra­
self-humidification behavior of the PEMFC stack. The proposed method tion, which in turn is directly proportional to pressure. Increasing the
makes the cathode side flow field has no fixed upstream and down­ pressure can accelerate the diffusion of the reactant gas, thereby accel­
stream region. Therefore, the distribution of water does not appear to erating the electrochemical reaction inside the PEMFC. In addition,
increase or decrease monotonously along the flow channel. Thereby a increasing the air pressure can make the internal gas concentration
more uniform water distribution is achieved. Therefore, alternating air distribution more uniform [37]. These can make the voltage increase.
flow directions not only achieves self-humidification without any The inlet and outlet pressures of the original system with constant air
external humidifiers, but also improves the uniformity of the water flow direction were 20 kPa and 0 kPa, respectively. The inlet and outlet
distribution, thereby increasing the output voltage of the PEMFC stack. pressures of the modified system with constant air flow direction were
Liquid water redistribution can improve liquid water management [13]. 39 kPa and 19 kPa, respectively. In Fig. 5, the average voltage of the
The proposed method of alternating air flow directions requires the original system with constant air flow direction is 11.834 V and the
addition of two three-way solenoid valves to the original system, which average voltage of the modified system with constant air flow direction
can increase the pressure of the cathode side under the same air flow rate is 12.279 V. The stack voltage results shown in Fig. 5 can prove that the
condition. Fig. 4(a) shows the pressure variation at the cathode inlet and proposed method of alternating air flow directions can increase the
outlet positions under different current conditions. The corresponding output voltage by increasing the air pressure. In summary, the proposed
temperature condition is 45 � C. The reason for choosing 45 � C is that at method of alternating air flow directions to improve performance
lower temperature, there can be more moisture in the cathode flow mainly comes from two aspects. One is to achieve self-humidification
channel and it can be more clearly seen that the pressure drop changes without any external humidifiers and the other is to increase the air
under different current conditions. There is no back pressure valve at the pressure. These two aspects are the direct reasons for the performance
air outlet of the original system and excess air is directly discharged into improvement of the proposed method.
the surroundings. Therefore, the pressure at the air outlet of the original Fig. 4(b) shows the variation of the cathode pressure drop under
system is 0 kPa (gauge pressure), as shown by the blue line in Fig. 4(a). different current conditions. The corresponding temperature condition
The modified system has three-way solenoid valves installed at the is 45 � C. Pressure drop can be used to evaluate the water flooding [16].
original air inlet and outlet positions. This makes it impossible for the air The increase in cathode pressure drop is caused by an increase of water
from the stack to be discharged directly into the surroundings, but to in the cathode side flow path. It can be found from Fig. 4(b) that the
pass through a three-way solenoid valve before being discharged into cathode pressure drop starts to gradually increase when the current
the surroundings. In the modified system, the three-way solenoid valve reaches 32 A (current density is 0.4 A/cm2). During the experiment,
can function as a back pressure valve. This increased the air outlet obvious liquid water was observed to be discharged from the air outlet
pressure of the modified system to 14 kPa (gauge pressure). Therefore, when the stack current reached 32 A. When the current increased to 64 A
the air outlet pressure of the modified system is higher than that of the (current density is 0.8 A/cm2), the cathode pressure drop no longer
original system, as shown by the red line in Fig. 4(a). Increasing back increased significantly and remained basically unchanged. When the
pressure is good for cell performance [36]. An increase in pressure of current is higher, the cathode pressure drop under the alternating air
reaction gases can cause an increase in the potential according to the flow direction condition is less than that under the constant air flow
Nernst Equation. In addition, water diffusion from the cathode to the direction condition. The variation in the cathode pressure drop can
anode is increased by pressurizing the cathode side. Therefore, the in­ indirectly reflect the accumulation of water in the flow channel. It can be
crease in air pressure is also one of the reasons for the higher perfor­ inferred that the alternating air flow directions can slow down the
mance of the modified system. Fig. 5 shows the stack voltage variation of accumulation of water in the flow channel downstream and make the
the original and modified systems with constant air flow direction. The water distribution more uniform. Therefore, the cathode pressure drop
cooling water inlet temperature was 55 � C. Each test was run for 900 s at is reduced under higher current conditions. This shows that the alter­
the current density of 0.8 A/cm2. The stack voltage variation within the nating air flow directions will not cause more water to accumulate inside
last 500 s was used for analysis. Under the condition of constant air flow, the PEMFC stack. This can prevent flooding.
water near the air outlet is not taken back to the inside of the PEMFC
stack. Therefore, the direct cause of the voltage increase in Fig. 5 is the
pressure increase. An increase in cell operating pressure results in higher

Fig. 5. Stack voltage variation under the current density condition of 0.8 A/cm2 and the cooling water inlet temperature of 55 � C. (a) Constant air flow direction-
Original system, (b) constant air flow direction-Modified system.

8
L. Luo et al. Journal of Power Sources 466 (2020) 228311

3.2. Effect of alternating air flow directions on the stack voltage method of alternating air flow directions will increase the fluctuation of
the stack voltage. In Fig. 6, the positions indicated by the red arrows are
To study the effect of alternating air flow directions on the stack the stack voltage change caused by the air flow direction switching and
voltage stability, the modified system using different alternating in­ the positions indicated by the blue arrows are the stack voltage change
tervals and alternating modes were tested. The cooling water inlet caused by the hydrogen purge operation. The red and blue arrows
temperature was 55 � C. Each test was run for 900 s at the current density indicate the same position, indicating that the stack voltage change is
of 0.8 A/cm2. The stack voltage variation within the last 500 s was used affected by both the hydrogen purge operation and the air flow direction
for analysis. The test results with different air alternating modes and switching. Comparing Figs. 5(b) and 6(a), it can be found that the
different alternating intervals are shown in Fig. 6. fluctuation of the stack voltage becomes larger because of alternating air
The fluctuation of the stack voltage is mainly caused by the hydrogen flow directions. Comparing Figs. 5(b) and 6(b), the same conclusion can
purge operation and the air flow direction switching. The proposed be obtained. It can be found from Fig. 6 that the shorter the alternating

Fig. 6. Stack voltage variation under the current density condition of 0.8 A/cm2 and the cooling water inlet temperature of 55 � C. (a) Stack voltage variation under
different alternating intervals - Alternate 2, (b) stack voltage variation under different alternating intervals - Alternate 1.

9
L. Luo et al. Journal of Power Sources 466 (2020) 228311

interval, the more frequent the stack voltage fluctuation. caused by the air flow direction switching is larger than the voltage
The stack voltage data shown in Figs. 5 and 6 is quantitatively range caused by the hydrogen purge operation. The maximum voltage
recorded in Table 3. Voltage range refers to the amplitude of the stack range caused by the air flow direction switching can reach 1.200 V,
voltage change caused by the hydrogen purge operation or the air flow accounting for 9.52% of the average stack voltage. This is a large voltage
direction switching operation. The last column in Table 3 lists the factors fluctuation. Comparing the voltage ranges at the same alternating in­
that cause voltage fluctuations. It can be known from Table 3 that the terval in Alternate 1 and 2, it can be found that the average stack voltage
hydrogen purge operation has little effect on the stack voltage. The ranges under these two conditions have little difference. For example,
average voltage range caused by the hydrogen purge operation is only when the alternating interval is 30.50 s, the average voltage range in
0.092 V. For the modified system with constant air flow direction, the Alternate 1 and 2 is 0.615 V and 0.620 V, respectively. The voltage range
average voltage range caused by the hydrogen purge operation increases of Alternate 1 mode is related to both the hydrogen purge operation and
to 0.193 V. However, when the cathode side is operated with alternating the air flow direction switching operation. The voltage range of Alter­
air flow directions, the voltage range rapidly increases. In Alternate 2 nate 2 mode is only related to the air flow switching operation. This
mode (air flow direction switching is in the middle of the purge inter­ shows that compared to the voltage fluctuation caused by the air flow
val), the air flow direction switching and the hydrogen purge operation direction switching, the stack voltage fluctuation caused by the
are not performed simultaneously. The voltage range of Alternate 2 hydrogen purge operation is very small.
mode is only related to the air flow switching operation. These voltage
changes are indicated by the red arrows Fig. 6(a). In Alternate 2 mode,
3.3. Effect of the purge interval on the stack voltage
the average voltage range is 0.620 V–0.834 V. It has an increasing
tendency with the increase of the alternating interval. In Alternate 1
Fig. 7 shows the stack voltage variation under different purge pe­
mode (air flow direction switching occurs at the end of the hydrogen
riods. The corresponding current density condition was 0.8 A/cm2 and
purge operation), the air flow direction switching and the hydrogen
the operating temperature was 55 � C. The alternate mode of this test is
purge operation are performed simultaneously. The voltage range of
Alternate 1. The purpose of this test is to explore the effect of the purge
Alternate 1 mode is related to both the hydrogen purge operation and
period on the stack voltage stability under the condition of alternating
the air flow direction switching operation. These voltage changes are
air flow directions. At the same time, it is also investigated whether the
indicated by the red and the blue arrows arrow at the same time in Fig. 6
amplitude of voltage fluctuation can be reduced by adjusting the purge
(b). In Alternate 1 mode, the average voltage range is 0.530 V–0.990 V.
period.
It also has a tendency to increase as the alternating interval increases.
It can be found from Fig. 7 that despite the change in the purge
Therefore, to reduce the fluctuation size of the stack voltage, a larger
period, the stack voltage still fluctuates greatly. The voltage variation
value should not be selected for the alternating interval. In addition, the
shown in Fig. 7 is similar to the voltage fluctuation in 6. This shows that
average stack voltage of the modified system with different alternating
the amplitude of the stack voltage fluctuation cannot be reduced by
intervals does not change much. Their average stack voltage is between
adjusting the purge period. It can be inferred that although the operation
12.499 V–12.629 V, as shown in Table 3. This is a small average stack
of alternating air flow directions can increase the water content on the
voltage range, 0.130 V. It is inferred that the air alternating interval has
cathode side and the self-humidification is achieved, the amount of the
a small effect on the stack voltage. It mainly affects the fluctuation
added water is not large and it has little effect on the water back
frequency of the stack voltage. Shorter alternating intervals will cause
diffusion. When the anode is in purge mode and the cathode has its
frequent fluctuations in stack voltage. Therefore, to reduce the fluctu­
highest pressure, the pressure difference between the two sides become
ation frequency of the stack voltage, a smaller value should not be
large. It should be noted that the purge time is only 0.5 s. Unlike charge
selected for the alternating interval. Further, the alternating interval
transport through the proton, water transport is not instantaneous. In
should be a trade-off between the fluctuation amplitude and the fluc­
such a short period of time, no large amount of water will be transferred
tuation frequency.
to the anode side by the mean of back diffusion. Wang and Wang [38]
In addition, fatigue is higher also in the membrane due to sudden
shown that it takes several seconds for the membrane to be re-hydrated
changes in the pressure gradients. Therefore, the alternating interval
by the back diffusion of water. This indicates that it takes at least a few
should also consider the fatigue failure of the membrane. Performance
seconds for more water to be transferred to the anode through the effect
and durability are indeed important issues that fuel cells need to address
of back diffusion. In addition, at higher current density conditions, more
and trade-offs need to be made for different applications. In practical
water is transferred from the anode side to the cathode side due to the
applications, a solution that can meet the requirements of practical use is
stronger electro-osmotic diffusion. Hydrogen purge operation can cause
reasonable.
voltage fluctuations [39]. If a large amount of water accumulates inside
The effect of the air flow direction switching on the stack voltage is
the PEMFC stack, the stack voltage will change significantly during the
greater than that of the hydrogen purge operation on the stack voltage.
purge process. However, despite the change in the purge period, the
As can be seen from Table 3, in most cases, the minimum voltage range
stack voltage still fluctuates greatly. The amplitude of the stack voltage

Table 3
Average stack voltage and voltage range under the current density of 0.8 A/cm.2.
Air flow direction Alternating interval (s) Average stack voltage (V) Voltage range (V) Factors

Average Maximum Minimum

Constant-Original – 11.834 0.092 0.102 0.081 Purge

Constant-Modified – 12.279 0.193 0.227 0.168
Alternate 2 30.50 12.603 0.620 0.910 0.362 Air flow direction switching
61.00 12.564 0.798 1.069 0.439
91.50 12.499 0.834 1.071 0.627
Alternate 1 15.25 12.629 0.530 0.764 0.169 Purge
30.50 12.614 0.615 0.839 0.246 þ
45.75 12.615 0.807 0.928 0.528 Air flow direction switching
61.00 12.606 0.636 1.032 0.247
76.25 12.581 0.990 1.091 0.881
91.50 12.609 0.820 1.200 0.538

10
L. Luo et al. Journal of Power Sources 466 (2020) 228311

Fig. 7. Stack voltage variation under different purge periods.

fluctuation cannot be reduced by adjusting the purge period, as shown in scheme have been discussed in detail. Finally, the following research
Fig. 7. These large voltage fluctuations are caused by the air flow conclusions are obtained:
switching operation. That is, the purge operation did not cause large
voltage fluctuations. This means that purge discharges less water. In (1). The proposed method of alternating air flow directions is effec­
addition, no more liquid water was observed at the hydrogen outlet. It is tive in improving the performance of the water-cooled proton
inferred that less water accumulates on the anode side of the PEMFC exchange membrane fuel cell stack. Under the current density
stack. That is, alternating air flow directions does not significantly in­ condition of 0.8 A/cm2, the output voltage of the stack increased
crease the water content in the stack. Therefore, no large amount of by 3.36%, 5.20%, and 19.46% at operating temperatures of 45
water is transferred from the cathode side to the anode side in the form �
C, 55 � C and 65 � C, respectively.
of back diffusion. According to such results, it can be further inferred (2). The proposed method can improve the uniformity of water dis­
that the proposed method of alternating air flow directions improves tribution inside the stack and achieve self-humidification. It is an
performance by improving the water distribution inside the PEMFC improvement in water management. Due to the addition of three-
stack. It does not cause more water to accumulate inside the PEMFC way solenoid valves at the air inlet and outlet positions, the air
stack. The proposed method increases the water content in the place pressure inside the stack is also increased.
where the water content is originally low (the upstream area of the flow (3). Alternating air flow directions will make the stack voltage fluc­
channel in the original system). This is an improvement in water tuate significantly. The alternating interval should be a trade-off
management. between the fluctuation amplitude and the fluctuation frequency.
Comparing the stack voltage variation curves in Figs. 6 and 7, it can The amplitude of the stack voltage fluctuation cannot be reduced
be found that in Alternate 1 mode, the frequency and amplitude of the by adjusting the purge period.
stack voltage fluctuation are relatively small when the purge period and
the alternate interval are the same at 30.50 s. Such operating conditions This research provides a new reference for the improvement of fuel
should be optimal for alternating air flow directions. cell water management and has guided significance for practical
applications.
4. Conclusion
Declaration of competing interest
An air supply system with alternating air flow directions is proposed
to improve water management and increase the output voltage of a The authors declare that they have no known competing financial
water-cooled proton exchange membrane fuel cell stack. The effects of interests or personal relationships that could have appeared to influence
different alternating modes, different alternating intervals and different the work reported in this paper.
hydrogen purge periods on the performance of the stack were experi­
mentally studied. The advantages and disadvantages of the proposed

11
L. Luo et al. Journal of Power Sources 466 (2020) 228311

CRediT authorship contribution statement Acknowledgements

Lizhong Luo: Conceptualization, Methodology, Investigation, This work was supported by the National Natural Science Foundation
Writing - original draft. Bi Huang: Investigation, Writing - review & of China (No. 21776095), the Guangzhou Science and Technology
editing. Zongyi Cheng: Writing - review & editing. Qifei Jian: Re­ Project (No. 201804020048, No. 201805010001), Key R&D Program of
sources, Writing - review & editing, Supervision, Project administration. Guangdong Province (2019B090909001), the Guangdong Science and
Technology Project (No. 2016B090927009) and Guangdong Key Labo­
ratory of Clean Energy Technology (No. 2008A060301002).

Appendix A. Supplementary data

Supplementary data to this article can be found online at https://doi.org/10.1016/j.jpowsour.2020.228311.

Nomenclature

I Current, A
SLPM Standard Liter Per Minute
V voltage, V
δ uncertainty

References [19] Heng Shao, Diankai Qiu, Linfa Peng, et al., In-situ measurement of temperature and
humidity distribution in gas channels for commercial-size proton exchange
membrane fuel cells, J. Power Sources 412 (2019) 717–724.
[1] Liangzhen Yin, Li Qi, Weirong Chen, et al., Experimental analysis of optimal
[20] Yulin Wang, Shixue Wang, Shengchun Liu, et al., Optimization of reactants relative
performance for a 5 kW PEMFC system, Int. J. Hydrogen Energy 44 (11) (2019)
humidity for high performance of polymer electrolyte membrane fuel cells with co-
5499–5506.
flow and counter-flow configurations, Energy Convers. Manag. 205 (2020)
[2] Jian Zhao, Xianguo Li, A review of polymer electrolyte membrane fuel cell
112369.
durability for vehicular applications: degradation modes and experimental
[21] Donggun Ko, Seungwoo Doh, Hyun Sun Park, et al., Investigation of the effect of
techniques, Energy Convers. Manag. 199 (2019) 112022.
operating pressure on the performance of proton exchange membrane fuel cell: in
[3] Fatima Zohra Arama, Khaled Mammar, Slimane Laribi, et al., Implementation of
the aspect of water distribution, Renew. Energy 115 (2018) 896–907.
sensor based on neural networks technique to predict the PEM fuel cell hydration
[22] R. Kuhn, Ph Krüger, S. Kleinau, et al., Dynamic fuel cell gas humidification system,
state, J. Energy Storage 27 (2020) 101051.
Int. J. Hydrogen Energy 37 (9) (2012) 7702–7709.
[4] N. Yousfi-Steiner, Ph Moçot�eguy, D. Candusso, et al., A review on PEM voltage
[23] A. Casalegno, S. De Antonellis, L. Colombo, et al., Design of an innovative enthalpy
degradation associated with water management: Impacts, influent factors and
wheel based humidification system for polymer electrolyte fuel cell, Int. J.
characterization, J. Power Sources 183 (1) (2008) 260–274.
Hydrogen Energy 36 (8) (2011) 5000–5009.
[5] O.S. Ijaodola, Zaki El- Hassan, E. Ogungbemi, et al., Energy efficiency
[24] Yafei Chang, Yanzhou Qin, Yan Yin, et al., Humidification strategy for polymer
improvements by investigating the water flooding management on proton
electrolyte membrane fuel cells – a review, Appl. Energy 230 (2018) 643–662.
exchange membrane fuel cell (PEMFC), Energy 179 (2019) 246–267.
[25] S. Rodosik, J.P. Poirot-Crouvezier, Y. Bultel, Impact of humidification by cathode
[6] Dowon Cha, Seung Won Jeon, Wonseok Yang, et al., Comparative performance
exhaust gases recirculation on a PEMFC system for automotive applications, Int. J.
evaluation of self-humidifying PEMFCs with short-side-chain and long-side-chain
Hydrogen Energy 44 (25) (2019) 12802–12817.
membranes under various operating conditions, Energy 150 (2018) 320–328.
[26] Liangfei Xu, Chuan Fang, Junming Hu, et al., Self-humidification of a polymer
[7] Yubai Li, Zhifu Zhou, Xianglei Liu, et al., Modeling of PEM fuel cell with thin MEA
electrolyte membrane fuel cell system with cathodic exhaust gas recirculation,
under low humidity operating condition, Appl. Energy 242 (2019) 1513–1527.
J. Electrochem. Energy Conv. Storage 15 (2) (2018), 021003.
[8] Segeun Jang, Changwook Seol, Yun Sik Kang, et al., Investigation of the effect of
[27] Xingwang Zhao, Liangfei Xu, Chuan Fang, et al., Study on voltage clamping and
carbon-covering layer on catalyst layer in polymer electrolyte membrane fuel cell
self-humidification effects of pem fuel cell system with dual recirculation based on
in low relative humidity condition, J. Power Sources 436 (2019) 226823.
orthogonal test method, Int. J. Hydrogen Energy 43 (33) (2018) 16268–16278.
[9] Rui B. Ferreira, D.S. Falc~
ao, V.B. Oliveira, et al., Experimental study on the
[28] Beom Jun Kim and Min Soo Kim, Studies on the cathode humidification by exhaust
membrane electrode assembly of a proton exchange membrane fuel cell: effects of
gas recirculation for PEM fuel cell, Int. J. Hydrogen Energy 37 (5) (2012)
microporous layer, membrane thickness and gas diffusion layer hydrophobic
4290–4299.
treatment, Electrochim. Acta 224 (2017) 337–345.
[29] A. Ramiar, A.H. Mahmoudi, Q. Esmaili, et al., Influence of cathode flow pulsation
[10] Yasutaka Nagai, Jens Eller, Tatsuya Hatanaka, et al., Improving water management
on performance of proton exchange membrane fuel cell with interdigitated gas
in fuel cells through microporous layer modifications: fast operando tomographic
distributors, Energy 94 (2016) 206–217.
imaging of liquid water, J. Power Sources 435 (2019) 226809.
[30] David Kadylak, Peter Cave, M� erida Walter, Effectiveness correlations for heat and
[11] Fereshteh Salimi Nanadegani, Ebrahim Nemati Lay, Bengt Sunden, Effects of an
mass transfer in membrane humidifiers, Int. J. Heat Mass Tran. 52 (5–6) (2009)
MPL on water and thermal management in a PEMFC, Int. J. Energy Res. 43 (1)
1504–1509.
(2019) 274–296.
[31] Rupak Banerjee, Satish G. Kandlikar, Two-phase flow and thermal transients in
[12] S. Hasanpour, M. Ahadi, M. Bahrami, et al., Woven gas diffusion layers for polymer
proton exchange membrane fuel cells – a critical review, Int. J. Hydrogen Energy
electrolyte membrane fuel cells: liquid water transport and conductivity trade-offs,
40 (10) (2015) 3990–4010.
J. Power Sources 403 (2018) 192–198.
[32] Richard S. Figliola, Donald E. Beasley, Theory and Design for Mechanical
[13] Jinyong Kim, Gang Luo, Chao-Yang Wang, Modeling liquid water re-distributions
Measurements, sixth ed., John Wiley & Sons, Inc, New York, 2014.
in bi-porous layer flow-fields of proton exchange membrane fuel cells, J. Power
[33] Guobin Zhang, Kui Jiao, Multi-phase models for water and thermal management of
Sources 400 (2018) 284–295.
proton exchange membrane fuel cell: a review, J. Power Sources 391 (2018)
[14] Kui Jiao, Xianguo Li, Water transport in polymer electrolyte membrane fuel cells,
120–133.
Prog. Energy Combust. Sci. 37 (3) (2011) 221–291.
[34] S. Martin, P.L. Garcia-Ybarra, J.L. Castillo, Long-term operation of a proton
[15] Y. Wu, J.I.S. Cho, M. Whiteley, et al., Characterization of water management in
exchange membrane fuel cell without external humidification, Appl. Energy 205
metal foam flow-field based polymer electrolyte fuel cells using in-operando
(2017) 1012–1020.
neutron radiography, Int. J. Hydrogen Energy 45 (3) (2020) 2195–2205.
[35] Renfang Wang, Guobin Zhang, Zhongjun Hou, et al., Comfort index evaluating the
[16] Yifan Xu, Linfa Peng, Peiyun Yi, et al., Numerical investigation of liquid water
water and thermal characteristics of proton exchange membrane fuel cell, Energy
dynamics in wave-like gas channels of PEMFCs, Int. J. Energy Res. 43 (3) (2019)
Convers. Manag. 185 (2019) 496–507.
1191–1202.
[36] Min Soo Kim, Dong Kyu Kim, Parametric study on dynamic heat and mass transfer
[17] Haichao Liu, Jing Tan, Lisheng Cheng, et al., Enhanced water removal performance
response in polymer electrolyte membrane fuel cell for automotive applications,
of a slope turn in the serpentine flow channel for proton exchange membrane fuel
Appl. Therm. Eng. 167 (2020) 114729.
cells, Energy Convers. Manag. 176 (2018) 227–235.
[18] He Liang, Ming Hou, Yanyan Gao, et al., A novel three-dimensional flow field
design and experimental research for proton exchange membrane fuel cells, Energy
Convers. Manag. 205 (2020) 112335.

12
L. Luo et al. Journal of Power Sources 466 (2020) 228311

[37] Huicui Chen, Biao Liu, Runtian Liu, et al., Optimal interval of air stoichiometry [39] B. Belvedere, M. Bianchi, A. Borghetti, et al., Experimental analysis of a PEM fuel
under different operating parameters and electrical load conditions of proton cell performance at variable load with anodic exhaust management optimization,
exchange membrane fuel cell, Energy Convers. Manag. 205 (2020) 112398. Int. J. Hydrogen Energy 38 (1) (2013) 385–393.
[38] Yun Wang, Chao-Yang Wang, Dynamics of polymer electrolyte fuel cells
undergoing load changes, Electrochim. Acta 51 (2006) 3924–3933.

13