Apen D 19 07376R2

Elsevier Editorial System(tm) for Applied
Energy
Manuscript Draft
Manuscript Number: APEN-D-19-07376R2
Title: Thermal Management of Edge-Cooled 1 kW Portable Proton Exchange

Membrane Fuel Cell Stack
Article Type: Research Paper
Keywords: PEM fuel cell; Portable stack; Edge cooling; Water and heat
management; Transient analysis
Corresponding Author: Dr. Ivan Tolj, PhD
Corresponding Author's Institution: Faculty of electrical engineering,

mechanical engineering and naval architecture
First Author: Ivan Tolj
Order of Authors: Ivan Tolj; Željko Penga; Damir Vukičević; Frano Barbir
Abstract: Comprehensive numerical analyses are conducted to study the

influence of thermal management on performance of 1 kW edge-cooled proton
exchange membrane fuel cell stack without external humidification. The
experimental stack and numerical three-dimensional computational fluid
dynamics model are characterized by several novelty aspects. Two
numerical approaches are considered and compared for a prescribed load
profile: (i) lumped model and novel (ii) real-time transient
computational fluid dynamics model incorporating realistic modeling of
forced air convection on the edge-cooling of the stack. The novelty of
the developed computational fluid dynamics model is the capability to
give insight in the transient results in only a fraction of time vs.
experimental testing (40 mins vs. 4 hours) and other computational fluid
dynamics models of fuel cells which are only capable of steady-state
analysis. The developed computational fluid dynamics model is used to
study the influence of (i) bipolar plate materials (ii) operating delta
pressure along the flow field and (iii) different cooling fin
configurations on the water and heat balance inside the stack. The
results indicate that (i) maximal and average temperatures of the stack
are almost linearly correlated to the thermal conductivity of bipolar
plate materials and maximal temperatures can be significantly higher (ii)
the operating delta pressure can be manipulated to increase the
performance of the stack and (iii) the cooling fin redesign has major
influence on the overall temperature uniformity across the stack.
Additionally, the heat transfer between the stack and metal hydride tank
is studied.
Cover Letter
To: Prof. J. Yan, PhD, Editor-in-chief: Applied energy
Dear Editor,
We are pleased to submit the revised version of our manuscript entitled "Thermal
Management of Edge-Cooled 1 kW Portable Proton Exchange Membrane Fuel Cell Stack"
which can hopefully be published. We would like to thank the reviewers for their thoughtful
comments and efforts towards improving our manuscript. We now hope that the manuscript in
the revised form is appropriate for publication in Applied Energy.
Best regards,
Asst. Prof. Ivan Tolj, PhD
University of Split
Faculty of electrical engineering, mechanical engineering and naval architecture
E-mail: itolj@fesb.hr
October 5, 2019
*Detailed Response to Reviewers
Response to the reviewers
Ms. Ref. No.: APEN-D-19-07376R1

Title: Thermal Management of Edge-Cooled 1 kW Portable Proton Exchange Membrane Fuel
Cell Stack
Applied Energy
The authors have learnt comments from the Editor and the Reviewer. All comments have
been positively addressed. The authors thank the Editor and the Reviewers for the increased
attention to this manuscript. The comments and the authors’ replies are given below.
Reviewer 2:
Q 2.1. I have looked at the revised version and the issues I had raised have been addressed to
my satisfaction. However there are some references cited in the additional material that are
not available. The author should check the missing references carefully.
AR 2.1.
We would like to thank Reviewer 2 for his valuable comments and suggestions. References
and their availability are double checked. A few small misprints and missing information
(such as journal title) have been identified and corrected.
Editors:
Q 0.1. All comments shall be addressed and responded in details rather than a word "done".
AR 0.1.
We would like to thank the editors for their valuable comments. All comments from the
editors are now responded in detail.
Q 0.2. The relevance to Applied Energy should be enhanced with the considerations of scope
and readership of the Journal.
A 0.2.
There is in total 5 references from Applied energy journal included in the manuscript:
 Wu HW. A review of recent development: Transport and performance modeling of
PEM fuel cells. Applied Energy 2016;165:81-106.
 Meidanshahi V, Karimi G. Dynamic modeling, optimization and control of power
density in a PEM fuel cell. Applied Energy 2012;93:98-105.
 Chang Y, Qin Y, Yin Y, Zhang J, Li X. Humidification strategy for polymer
electrolyte membrane fuel cells – A review. Applied Energy 2018;230:643-662.
 Wang J. Theory and practice of flow field designs for fuel cell scaling-up: A critical
review. Applied Energy 2015:157:640-663
 Kurnia JC, Sasmito AP, Shamim T. Performance evaluation of a PEM fuel cell stack
with variable inlet flows under simulated driving cycle conditions. Applied Energy
2017;206:751-764.
Q 0.3. A proof reading by a native English speaker should be conducted to improve both
language and organization quality.
A 0.3.
We believe that the organization quality of the revised manuscript is now clear enough. The
text of the revision has been additionally proof-read by native English (UK)-speaking
researcher having solid publication records in various international journals.
Q 0.4. Please avoid using abbreviations in the TITLE, HIGHLIGHTS, ABSTRACT and
CONCLUSION if possible.
A 0.4.
The submission title has been modified in the revision: abbreviation “PEM” has been replaced
with “Proton Exchange Membrane”. Abbreviations in the abstract such as “PEM” and “CFD”
are now replaced with “Proton Exchange Membrane” and “Computational Fluid Dynamics”
respectively. Where possible abbreviations in the highlights and conclusion are avoided.
Q 0.5. Please also avoid "lump sum references", such as XXXXX [1-5]; all references should
be cited with detailed and specific description. In the references, all authors should be
included, avoiding using "et. al.;
A 0.5.
The authors tried to separate the contribution of each referenced source in the revision.
However, in some cases which require systematisation of the state-of-the art and revealing the
common features, the lumped referencing is inevitable. All references are now cited with
detailed and specific descriptions. In the references all authors are now included avoiding
using “et. al.”.
Q 0.6. Please use 'Highlights' in the file name and include 3 to 5 bullet points (maximum 85
characters, including spaces, per bullet point).
A 0.6.
File name is modified and in total 5 bullet points are included. We believe that the Highlights
adequately convey the core findings of the article and, at the same time, do not exceed 85
characters, including spaces, per bullet point.
Q 0.7. TITLE: It normally consists of about 12-15 keywords which shall not be too general or
too narrow.
A 0.7.
Title has been modified and now consists of 10 keywords.
Q 0.8. HIGHLIGHTS: Highlights are mandatory for this journal. They consist of a short
collection of bullet points that convey the core findings of the article and should be submitted
in a separate editable file in the online submission system. Please use 'Highlights' in the file
name and include 3 to 5 bullet points (maximum 85 characters, including spaces, per bullet
point).
A 0.8.
Highlights are now submitted in a separate file and file name has been modified. We believe
that the Highlights adequately convey the core findings of the article and, at the same time, do
not exceed 85 characters, including spaces, per bullet point.
Q 0.9. ABSTRACT: It should be about 150-250 words with concise text in a single paragraph.
Answer the questions: What problem did you study and why is it important? What methods
did you use? What were your main results? And what conclusions can you draw from your
results? Please make your abstract with more specific and quantitative results while it suits
broader audiences. Abstract stands alone, no references, figures, tables or equations are
cited.
A 0.9.
Abstract is now rewritten, and it consists of 245 words. We now believe that abstract answers
questions – what problem was studied and why is it important, what methods are being used
and what where main results. Abstract is without references, figures, tables and equations.
Q 0.10. CAPTIONS: Captions for figures and tables should be presented with more specific
description rather than a general sentence like "Results of the experiments ...", "A studied
system ...."
A 0.10.
Captions for figures and tables are now presented with more specific description:
Figure 1. Above: Proton Energy System’s short stack exploded view (left) with annotated
parts and assembly (right); Middle: Proton Energy System’s 1-kW fuel-cell stack (left) and
transparent shroud assembly (right); Below: Cathode (left) and anode (right) flow fields with
annotated air-coolant flow direction along the edge-cooling fins.
Figure 2. Default mesh configuration (left and middle above) with details of solid and fluid
meshes, and geometry with annotated boundary conditions (right).
Figure 3. Cooling fin configurations 1 to 6 with annotated guide vanes on the side plates (red
circles).
Figure 4. Grid (above) and time-step (below) dependency vs. flow time.
Q 0.11. UNITS: SI units shall be used.

A 0.11.
Manuscript has been modified and only SI units are used.
Q 0.12. DATA: All data shall be carefully presented with consistent accuracy.
A 0.12.
We now believe that data presented in the revised version of the manuscript are presented
carefully and at the same time with the consistent accuracy.
Q 0.13. The originality of the paper needs to be further clarified. It is of importance to have
sufficient results to justify the novelty of a high quality journal paper.
A 0.13.
Introduction section has been modified where manuscript originality is additionally clarified.
The main contributions of this work to the research field are evident in outlining that the
commonly used lumped models can be misleading for dynamic analysis of PEM fuel cell
performance due to inability to accurately calculate the maximal temperature inside the stack
and this work also outlines under which circumstances the lumped models can be used. This
work also presents a novel transient CFD model for analysis of thermal management of PS
which gives insight in mass and heat transfer both inside and outside of the cell (internal heat
and mass transfer and forced air convection edge-cooling), while other works are focused on
only one aspect and simplify the other to a significant extent. The methodology for
developing such model using CFD is shown in detail in this work. The novelty of the work is
also evident in conducting the analysis which gives insight in maximal temperatures inside the
PS for different bipolar plate materials, how to enhance the performance by introducing a
pressure drop along the flow field or operating the cell at elevated pressure, and showing high
significance of external cooling fin configurations on the overall heat transfer (while the
results of the analyses show that the maximal temperatures inside the stack are very similar to
the ones observed during experimental investigation done by other research groups recently).
The novelty is also including the transient heat exchange with metal hydride tank, and the
results are qualitatively comparable with experimentally obtained results for a similar stack.
Q 0.14. An updated and complete literature review should be conducted to present the state-
of-the-art and knowledge gaps of the research with strong relevance to the topic of the paper.
A 0.14.
We added several references related to transient responses of proton exchange membrane fuel
cells and thermal coupling of metal hydride hydrogen storage and fuel cell systems. Also, few
references dealing with experimental study of an air-cooled proton exchange membrane fuel
cell stack are added.
Q 0.15. The results should be further elaborated to show how they could be used for the real
applications. Modeling results should be validated by experiments.
A 0.15.
The modeling results are now validated with the experiments conducted by other research
groups and relevant literature was included in the manuscript. Results and discussion section
together with conclusions section have been modified in order to emphasize connection with
real applications such as: coupling of metal hydrides tanks with proton exchange membrane
fuel cell stack in order to increased heat transfer and consequently hydrogen desorption,
influence of bipolar plate material on internal stack temperature and influence of fin geometry
on overall heat transfer and temperature distribution. We believe that results presented in this
manuscript will be of interest to scientists and engineers working in the field of proton
exchange membrane fuel cells and hydrogen storage.
On behalf of the authors
Asst. Prof. Ivan Tolj, PhD
5th October 2019

*Revised Manuscript with Changes Marked
Click here to download Revised Manuscript with Changes Marked: Thermal Management of Edge-Cooled_R1 marked change
1
Thermal Management of Edge-Cooled 1 kW
2
3
4
5 Portable Proton Exchange Membrane Fuel Cell
6
7
8
9 Stack
10
11
12 Ivan Tolj1*, Željko Penga1, Damir Vukičević1, Frano Barbir1
13
14 1
15 Faculty of Electrical Engineering, Mechanical Engineering and Naval Architecture
16
17 University of Split, R. Boškovića 32, 21000 Split, Croatia
18
19
20
21 Abstract
22
23
24 Comprehensive numerical analyses are conducted to study the influence of thermal
25
26 management on performance of 1 kW edge-cooled proton exchange membrane fuel cell stack
27
28
29 without external humidification. The experimental stack and numerical three-dimensional
30
31 computational fluid dynamics model are characterized by several novelty aspects. Two
32
33
34 numerical approaches are considered and compared for a prescribed load profile: (i) lumped
35
36 model and novel (ii) real-time transient computational fluid dynamics model incorporating
37
38
realistic modeling of forced air convection on the edge-cooling of the stack. The novelty of
39
40
41 the developed computational fluid dynamics model is the capability to give insight in the
42
43 transient results in only a fraction of time vs. experimental testing (40 mins vs. 4 hours) and
44
45
46 other computational fluid dynamics models of fuel cells which are only capable of steady-
47
48 state analysis. The developed computational fluid dynamics model is used to study the
49
50
51 influence of (i) bipolar plate materials (ii) operating delta pressure along the flow field and
52
53 (iii) different cooling fin configurations on the water and heat balance inside the stack. The
54
55
56
results indicate that (i) maximal and average temperatures of the stack are almost linearly
57
58
59 *
Corresponding author. Tel.: +385 21 305 948; fax: +385 21 305 776.
60
E-mail address: itolj@fesb.hr (I. Tolj).
61
62 1
63
64
65
correlated to the thermal conductivity of bipolar plate materials and maximal temperatures
1
2 can be significantly higher (ii) the operating delta pressure can be manipulated to increase the
3
4
5 performance of the stack and (iii) the cooling fin redesign has major influence on the overall
6
7 temperature uniformity across the stack. Additionally, the heat transfer between the stack and
8
9
10 metal hydride tank is studied.
11
12
13
14
15
Keywords:
16
17 PEM fuel cell;
18
19 Portable stack;
20
21
22 Edge-cooling;
23
24 Water and heat management;
25
26
27 Transient analysis.
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62 2
63
64
65
1. INTRODUCTION
1
2
3 Air-cooled proton exchange membrane (PEM) fuel cell stacks are gaining momentum as
4
5 power sources for portable applications such as electric bikes, unmanned aerial vehicles
6
7
8 (UAVs), forklifts, etc. Usually the power of portable stacks (PS) is up to several kW [1,2].
9
10 The majority of the heat generated during PS operation is dissipated into the surrounding
11
12
atmosphere via forced air convection along the cooling channels or edge-cooling fins, while
13
14
15 the higher power stacks require liquid cooling [3]. The emphasis during the design phase of
16
17 PS is primarily directed towards determining the most suitable material for bipolar plates and
18
19
20 the accompanying thermal management. Some of the thermal management objectives are: (i)
21
22 ensuring the appropriate heat transfer from the stack during startup, i.e. accurately
23
24
25 determining the required convection flow rate, to achieve sufficiently high operating
26
27 temperature from startup at ambient temperature, thus preventing flooding (occurring if the
28
29
30
flow rate is high and temperature of the stack is low) or membrane dehydration (occurring if
31
32 the flow rate is low and temperature inside the stack is high); (ii) ensuring sufficiently high
33
34 heat removal rates via forced air convection during the operation characterized by stable
35
36
37 temperature of the stack, i.e. loosely termed steady-state operation; (iii) ensuring as low as
38
39 possible temperature difference between the inside and outside portions of the stack by
40
41
42 choosing the materials with favorable thermal properties; (iv) properly design the stack and
43
44 edge-cooling fins to ensure sufficiently high heat removal rates; (v) design the frame for the
45
46
47 stack which will ensure that the cooling air is directed to uniformly flow along the edge-
48
49 cooling fins.
50
51
Due to the requirement for simple system design, portable stacks are commonly operated
52
53
54 without external humidification. Using dry hydrogen and ambient air results in poor
55
56 performance of the stack and exacerbated degradation unless special care is attenuated to the
57
58
59 water and heat management. As previously mentioned, the PSs are air-cooled, either via
60
61
62 3
63
64
65
cooling channels between the neighboring bipolar plates, or via specially designed fins
1
2 protruding from the longer sides of the stack, i.e. edge-cooling, as seen in [4]. Cooling
3
4
5 channels between the neighboring bipolar plates are similar in design to the coolant channels
6
7 for higher power stacks where liquid coolant is used. Cooling channels between the bipolar
8
9
10 plates result in a uniform temperature profile along the cell, but significantly increase the size
11
12 of the stack due to the requirement for low pressure potential required to achieve the desired
13
14
15
flow through the channels. Edge-cooling results in a more compact stack design since there
16
17 are no cooling channels between the neighboring bipolar plates. However, edge-cooling
18
19 requires sufficiently high thermal conductivity and special attention during the design of the
20
21
22 cooling fins to achieve uniform temperature distribution inside and along the stack.
23
24 Low power air-cooled high temperature PEM fuel cell stacks have been implemented in
25
26
27 hybrid electrical vehicles and coupled with lead acid batteries and are proven to be feasible
28
29 for such applications [5], therefore they are the focus of this work. However, the issue
30
31
32
remains how to cool low temperature PEM fuel cell efficiently under such operation due to
33
34 lower temperature gradient between the stack surface and the surrounding atmosphere.
35
36 Considerable effort is dedicated to developing passive cooling techniques for PEM fuel cells
37
38
39 used in space exploration missions [6], where the conclusion is that there is a requirement for
40
41 using costly materials with thermal conductivity in excess of 1000 W m-1 K-1 to achieve low
42
43
44 temperature gradient (lower than 3 °C) along the cell [7], or heat pipes for heat removal in
45
46 larger stacks [8], due to volume restrictions of the stacks used for such applications. In the
47
48
49 referenced studies feasible solution was achieved by using pyrolitic graphite combined with
50
51 anodized aluminum heat exchanger. In this work the focus of investigation is to see if lower
52
53
54
temperature gradient along the stack cross-section can be achieved using conventional
55
56 materials for bipolar plates or by simply changing the design of the edge-cooling fins,
57
58 therefore avoiding the necessity for costly materials. Another study [9] also used pyrolitic
59
60
61
62 4
63
64
65
graphite sheets for heat removal on a single-cell and demonstrated that the approach is
1
2 economically viable and useful for PEM fuel cell applications. In [10] it is demonstrated that
3
4
5 the PEM fuel cell stack efficiency can be increased by 15% while the flooding issues at low
6
7 cathode flow rates can be alleviated using this approach. Since thermal conductivity has such
8
9
10 significant influence of PEM fuel cell performance, this work is also focused on determining
11
12 the correlation between the bipolar plate material and maximal temperature inside the stack
13
14
15
during transient operation. The enhancement of the concept was then achieved by introducing
16
17 small fans into the stack system [11] thus improving the convection rates and providing
18
19 economically-friendly solution (when compared to coolant loop system) to the thermal
20
21
22 management issues of small to medium sized PEM fuel cell stacks. External cooling was also
23
24 implemented in study [12] on high temperature PEM fuel cell where the heat transfer was
25
26
27 analyzed using CFD and it was demonstrated that it is possible to achieve temperature
28
29 gradient below 15 °C via external cooling with air or thermo transfer liquid and to keep the
30
31
32
temperature gradient along the cell below 15 °C at all times [13]. Another important aspect
33
34 related to the thermal management and application of external cooling techniques is the
35
36 proper bipolar plate material selection and flow field design [14] which can significantly
37
38
39 increase or decrease the performance of the cell. Further improvements of the edge-cooling
40
41 technique can also be achieved by designing flow fields with reactant gas flow field and
42
43
44 coolant flow fields on the same plate surface, thereby decreasing the coolant mass flow rate
45
46 requirements and decreasing the temperature gradient inside the cell and also decreasing the
47
48
49 heat conduction length in stacks [15,16].
50
51 Different numerical methods are commonly applied for thermal and mass transfer analysis of
52
53
54
PEM fuel cells. Thermal analysis of PEM fuel cells is thoroughly described in [17] where the
55
56 emphasis is dedicated to CFD modeling due to capabilities for multi-dimensional and multi-
57
58 physics analysis. In the references work it is noted that most works related to CFD modeling
59
60
61
62 5
63
64
65
do not include the mesh sensitivity and time step sensitivity analyses while they provide data
1
2 for only specific operating conditions. However, due to computational complexity, the CFD
3
4
5 modeling technique is usually implemented for steady-state in-depth analysis of PEM fuel cell
6
7 operation. For automotive applications and the correlating thermal analysis other methods are
8
9
10 employed such as Monte-Carlo simulations [18], but this type of analysis does not give
11
12 insight in the temperature distribution and operating conditions inside the cell. Another
13
14
15
interesting topic for PEM fuel cell stacks is cold start, which is recently experimentally
16
17 investigated [19], and it is also considered in this work (although from ambient temperature
18
19 and not from subzero temperatures as in [19]). Time-efficient method for thermal analysis of
20
21
22 PEM fuel cells is control volume method [20], which is an upgraded lumped model
23
24 methodology. Although useful for control strategy development, this methodology also does
25
26
27 not give insight in the localized temperature distribution inside the cell but only approximates
28
29 the cell as a more detailed lumped model. Dynamic 1D modeling of PEM fuel cell operation
30
31
32
is shown in [21] where the conservation laws and non-linear terms are included and the model
33
34 is used to study the effectiveness of different control strategies, giving indications of the
35
36 requirement for adding the physical conservation laws for developing accurate and sensible
37
38
39 numerical model. The dynamic performance of a 500 W open-cathode stack [22] gives
40
41 overview of the temperature distributions between the cells in the stack, cooled via forced air
42
43
44 convection, and characteristic shapes of the temperature profiles during real time testing. In
45
46 [23] an overview is given of various methods for external and internal humidification of the
47
48
49 cell by adopting different strategies. One of the strategies is to manipulate the temperature or
50
51 the pressure distribution inside the flow field. Pressure distribution inside the flow field
52
53
54
affects relative humidity to a significant extent and is beneficial to some extent for liquid
55
56 water removal, and it can be potentially used to achieve the desired relative humidity
57
58 distribution inside the cell without the requirement for external humidification. The effect of
59
60
61
62 6
63
64
65
pressure drop on cell performance is discussed in more details in [24] and there are
1
2 indications that high pressure drop along the cell results in poor performance, high cost,
3
4
5 uneven concentration gradients between inlet and outlet, etc. Since it is known that the
6
7 operating pressure has significant influence on the water vapor saturation profile, and there
8
9
10 are indications that manipulating the operating pressure can increase the performance of the
11
12 cell [25], this will be studied in this work by operating the cell at higher operating pressure
13
14
15
with low pressure drop, i.e. operation similar to installing a pressure valve behind the cell, to
16
17 see if significant gains in efficiency can be achieved using this novel method, by comparing
18
19 the relative humidity profiles vs. common operation with zero gauge pressure.
20
21
22 In [26] lumped model for stack thermal management is developed and compared with
23
24 experimental data and demonstrated that high order non-linear lumped model shows good
25
26
27 agreement with experimentally obtained data for the average temperature and also shows that
28
29 third order non-linear model also shows good transient agreement. However, the maximal
30
31
32
temperatures inside the stack are unknown and therefore the results have limited practical
33
34 applications where materials with low thermal conductivity are used as bipolar plates. In
35
36 study [27] an overview is given of different multi-phase modeling approaches used for PEM
37
38
39 fuel cells via 3D CFD analysis and it is outlined that the future studies should be conducted
40
41 with more realistic outside boundary conditions, i.e. convection, vs. isothermal approach and
42
43
44 also outlines that currently developed detailed 3D CFD models show limited practicality due
45
46 to high complexity of the calculations even for steady-state simulations, indicating the
47
48
49 requirement for developing more simple but computationally less intensive models for
50
51 transient analysis in real-time. Dynamic model of PEM fuel cell is developed using CFD
52
53
54
methodology in [28,29] and shows good agreement with experimental polarization curves and
55
56 gives reasonable response to transient changes in operating electric potential. However, the
57
58 models in the referenced works are two-dimensional and as such not suitable for giving
59
60
61
62 7
63
64
65
insights in the detailed heat and mass transfer processes inside and along the cell during
1
2 transient operation.
3
4
5 PEM fuel cell stacks are often modeled using simplified but fast response models, as seen in
6
7 [30] where a system level numerical model is developed and the influence of the blower
8
9
10 pressure on a 80 kW PEM fuel cell stack was investigated. Simplified CFD model of a PEM
11
12 fuel cell stack is shown in [31] where the entire stack is modeled as a 2D structure and
13
14
15
thermal analysis was conducted to determine if the application of phase-change materials and
16
17 insulator coating can alleviate the freeze-thaw cycles in cold surrounding atmospheres.
18
19 Although the model is considerably simplified when compared to fully resolved 3D CFD
20
21
22 model of PEM fuel cell, the analysis results give significant amount of information about the
23
24 thermal management of the stack, superior to using lumped model approach. Another
25
26
27 example of simplified modeling with good prediction of PEM fuel cell stack performance is
28
29 described in [32] where MATLAB-Simulink environment is used to model the polarization
30
31
32
curves under different operating conditions with good agreement vs. experimentally obtained
33
34 data. Also, a simple 1D transient model of PEM fuel cell shown in [33] indicates the
35
36 importance of temperature control of the cell on the overall voltage response due to the effect
37
38
39 of temperature on the water vapor saturation profile, i.e. relative humidity and thus the
40
41 membrane water content. Simplified CFD model of a stack is shown in [34] where the stack is
42
43
44 represented by a symmetrical half of a single cell with periodic boundary conditions and
45
46 subjected to dynamic automotive load profiles for a 320 cells with 1600 cm2 active catalyst
47
48
49 area and notion is given that small changes in the mass flow rates of the reactants and coolant
50
51 have a considerable effect on the thermal envelope of the stack during transient operation,
52
53
54
albeit their effect on the net power stack is minor, indicating that it is required to study the
55
56 whole cross-section of the stack during prolonged transient operation. A more detailed
57
58 transient CFD model is shown in [35] where a simplified single-cell model with different
59
60
61
62 8
63
64
65
channel configurations is used to study the dynamic performance of the cell with emphasis on
1
2 voltage undershoots/overshoots. Since PS can be coupled with metal hydride tank, the heat
3
4
5 generated by the cell can be used to release the stored hydrogen from the metal hydride tank
6
7 during operation, as seen in study [36].
8
9
10 Below are some of the relevant lumped model works which were used as inspiration for
11
12 building the lumped model in this work. In [37] developed a thermal model for a Ballard
13
14
15
Mark V 35-cell 5 kW PEM fuel cell stack by performing mass and energy balances on the
16
17 stack. Initially, they developed a steady-state overall dynamic model by coupling an
18
19 electrochemical model with thermal model. The model was further transformed into a
20
21
22 transient model that predicted fuel cell performance in terms of cell voltage output and heat
23
24 losses as a function of time due to various changes imposed on the system. Similar approach
25
26
27 is used in this work via CFD analysis, which gives more detailed insight in the mass and heat
28
29 transfer inside the stack as well as on the outside of the stack due to including forced air
30
31
32
convection. A mathematical model of PEM fuel cell, including air-compression and
33
34 humidification process is shown in [38]. The study focused on air-supply management, with
35
36 the objective of optimizing the inlet air pressure and stoichiometry provided by the
37
38
39 compression system. The optimization results showed that working with fully humidified air
40
41 at the inlet is not always the best solution, especially for low air flow rates, since under such
42
43
44 circumstances the water accumulation inside the cell is exacerbated. In this work therefore the
45
46 focus is to show that it is possible to achieve high relative humidity of the reactants without
47
48
49 the necessity for external humidification by choosing the bipolar plate materials and designing
50
51 the edge-cooling fin geometry, thereby ensuring high operational efficiency with minimized
52
53
54
occurrence of flooding. A simplified approach for evaluation and modeling PEM fuel cell
55
56 stacks in a stationary operation is shown in [39], using a model based on the reduction of
57
58 measured stack parameters via regression approaches. Because heat is lost at the stack
59
60
61
62 9
63
64
65
surface, pressure losses, stack outlet temperature and the characteristics of a fuel cell can be
1
2 simulated by regression approaches. The results of their study included optimized
3
4
5 measurement algorithms for fuel cell stacks to reduce the time required for their evaluation
6
7 due to the multitude of variable parameters. The approach that is used in this work is similar
8
9
10 in means that the stack is modeled as heat and mass transfer system, while the influence of the
11
12 electrochemical reactions occurring inside the stack and the accompanying heat release are
13
14
15
modeled using sink and source terms, enabling fast analysis of transient stack operation. The
16
17 approach shown in this work is different since it uses CFD analysis, while it is also different
18
19 from other works using CFD analysis because it results in computationally far less intensive
20
21
22 model when compared to full 3D CFD analyses including more advanced electrochemistry
23
24 modelling. Another study [40] developed a mathematical model to simulate transient
25
26
27 phenomena in PEM fuel cell system. The dynamic fuel cell model incorporated the effects of
28
29 charge double-layer capacitance, the dynamics of flow and pressure in the anode and cathode
30
31
32
channels and transient mass and heat transfer features inside the cell. The model predicted the
33
34 transient response of cell voltage, cell temperature, hydrogen and oxygen outflow rates and
35
36 cathode channel temperatures and pressures under sudden load changes. However, in that
37
38
39 study the effect of the heat transfer outside of the stack is not modelled, while in this work it
40
41 can be seen that the heat transfer between the stack and surrounding atmosphere has
42
43
44 significant influence on the overall temperature profile along and inside the stack. In [41]
45
46 mathematical model is developed to represent the membrane-electrode assembly (MEA) of
47
48
49 fuel-cell systems. The model was used to analyze the effects of various polarization
50
51 resistances on cell performance. Thermally coupled hydrogen storage and fuel cell system
52
53
54
was investigated in [42]. This work utilized a computational environment referred to as the
55
56 virtual test bed to simulate and compare the behavior of coupled and uncoupled systems. This
57
58 model used lumped approximations, and a pulsed hydrogen demand load was applied. The
59
60
61
62 10
63
64
65
results from these simulations clearly revealed the unique and subtle behavior of thermally
1
2 coupled systems. Since waste heat can be used to release the stored hydrogen inside metal
3
4
5 hydride tanks, commonly used in forklifts, this was also investigated in this work by coupling
6
7 the systems using CFD analysis and to determine will the amount of hydrogen released be
8
9
10 able to feed the stack during operation at 1 kW power. In another study [43] dynamic model
11
12 of a PS was developed in MATLAB Simulink environment, including heat transfer between
13
14
15
the stack and a metal hydride storage unit. The derived equations described both the steady-
16
17 state and dynamic operation of the PEM fuel cell system with sufficient accuracy. Different
18
19 operating conditions were simulated, specifically addressing high and low-temperature PEM
20
21
22 fuel cell operation. The model can be used to compare different operating conditions and to
23
24 estimate the interaction between the different components. In this work the approach is
25
26
27 upgraded by also giving insight in the temperature distribution inside the metal hydride tank.
28
29 Three-dimensional dynamic thermal model of a single cell was developed in [44] to study
30
31
32
temperature distribution in a fuel cell cooled by air flow from the bottom to the top. The
33
34 model was governed by the thermal energy balance, taking into account the inlet gas
35
36 humidity, and developed using a finite difference method implemented in the MATLAB
37
38
39 Simulink environment. The efficiency of the air cooling device revealed that the cell
40
41 temperature is directly linked to current density and gas humidity. Moreover, the temperature
42
43
44 variation in the stack was shown to be very high. The variability of air cooling between the
45
46 cells of the stack led to large temperature variations, up to 8 °C, from one cell to another. In
47
48
49 this work the focus is dedicated to finding solution for minimizing the temperature
50
51 inhomogeneity by determining the influence of bipolar plate material and edge-cooling fin
52
53
54
design on the overall temperature envelope.
55
56 Although lumped models are useful due to their fast response and low computational intensity
57
58 with reasonable accuracy when compared to experimental measurements of integral
59
60
61
62 11
63
64
65
parameters, spatial distribution of species transport and heat transfer, as well as critical
1
2 operating parameters, e.g. extreme values of temperature and relative humidity, are lost
3
4
5 during such approaches due to integration and lumping the values in single digit. Without
6
7 determining the minimal and maximal values of temperature inside the stack (and
8
9
10 consequently the relative humidity distribution since water vapor saturation profile is
11
12 temperature dependent), it is not possible to design the stack for optimal operation. Localized
13
14
15
hot spots can result in severe membrane dehydration and reduced lifetime of the stack, while
16
17 regions with lower temperature are prone to flooding, resulting in non-uniform reactant
18
19 distribution along the active area of the cell, non-uniform current density distribution, and
20
21
22 potential starvation of the cell in regions where significant amounts of water are present. In
23
24 previous studies [45,46] it was shown that the operating temperature inside PEM fuel cell is
25
26
27 highly non-uniform and it is not accurate to represent the operating temperature with just one
28
29 parameter, while at the same time if the temperature profile is controlled it can be exploited to
30
31
32
result in high performance of the cell by manipulating the water vapor saturation profile, i.e.
33
34 relative humidity profile inside the cell. In [47] short stack was developed and it was shown
35
36 that the temperature distribution along the stack height can be non-uniformly distributed if the
37
38
39 coolant mass flow rate is low, and that the temperatures between the neighboring bipolar
40
41 plates can be different, which is also not captured by the lumped model approach. Also,
42
43
44 during the dynamic PEM fuel cell operation, in [48] it can be seen that the processes inside
45
46 the cell related to the liquid water accumulation and transport can be rather sluggish and
47
48
49 influence the performance of the cell significantly, therefore transient analysis must be
50
51 conducted in conjunction with steady-state modeling using CFD analysis to give detailed
52
53
54
insight in the processes inside the cell. However, once the thermal management is resolved
55
56 using more detailed CFD analysis and the temperature gradients during the operation are
57
58 resolved, the lumped modeling approach can be used to give much faster response during
59
60
61
62 12
63
64
65
transient operation and for this reason it is best to use both approaches in conjunction for
1
2 thermal analysis of PEM fuel cell stacks during transient operation.
3
4
5 The work is structured as follows. The experimental stack is briefly described in [49] and [50]
6
7 including the novelty aspects of the design. Lumped steady-state and transient models are
8
9
10 shown, and transient model is used to study the temperature dependency on the operating
11
12 power during dynamic load profile and for three different shutdown strategies and also to
13
14
15
determine the amount of condensed water in the cathode exhaust. The novel CFD model is
16
17 first developed to determine if the stack manifold geometry has sufficiently high cross-section
18
19 area to achieve uniform flow of the reactants along the entire height of the stack and to
20
21
22 determine the required viscous resistance coefficient which is to be used on a single-cell level
23
24 to accurately result in the pressure drop along the flow field similar to the experimentally
25
26
27 obtained value. Once it was determine that the flow distribution between the neighboring cells
28
29 in the stack is uniform, single-cell model with symmetry boundary conditions is built and
30
31
32
used for detailed thermal analysis including modeling of relative humidity, reactant
33
34 consumption and water generation for the cathode side of the cell as well as the influence of
35
36 the geometry of the edge-cooling fins on the spatially resolved temperature distribution during
37
38
39 transient operation. The mesh and time-step sizes are thoroughly determined to give accurate
40
41 results and fast computational performance and the resulting CFD model has shown faster
42
43
44 performance than real-time testing. The results of the lumped model analysis are compared
45
46 with experimental data from other similar works and show good qualitative and quantitative
47
48
49 agreement.
50
51 The main contributions of this work to the research field are evident in outlining that the
52
53
54
55
56 performance due to inability to accurately calculate the maximal temperature inside the stack
57
58 and this work also outlines under which circumstances the lumped models can be used. This
59
60
61
62 13
63
64
65
1
2 which gives insight in mass and heat transfer both inside and outside of the cell (internal heat
3
4
5 and mass transfer and forced air convection edge-cooling), while other works focus on only
6
7 one aspect and simplify the other to a significant extent. The methodology for developing
8
9
10 such model using CFD is shown in detail in this work. The novelty of the work is also evident
11
12 in conducting the analysis which gives insight in maximal temperatures inside the PS for
13
14
15
different bipolar plate materials, how to enhance the performance by introducing a pressure
16
17 drop along the flow field or operating the cell at elevated pressure, and showing high
18
19 significance of external cooling fin configurations on the overall heat transfer (while the
20
21
22 results of the analyses show that the maximal temperatures inside the stack are very similar to
23
24 the ones observed during experimental investigation done by other research groups recently).
25
26
27 The novelty is also including the transient heat exchange with metal hydride tank, and the
28
29 results are qualitatively comparable with experimentally obtained results for a similar stack.
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62 14
63
64
65
2. EXPERIMENTAL STACK
1
2
3 The stack used in this study, Figure 1, was developed by Proton Energy Systems [49]. The
4
5 stack displays many innovative features including in-flow configuration, cooling and stack
6
7
8 clamping [49-51]. It consists of 60 cells, each with an active area of 65 cm2. The design is
9
10 rectangular with 10:1 aspect ratio to allow for efficient heat conduction of the edge of the
11
12
cells, where fins are extended for cooling.
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50 Figure 1. Above: Proton Energy System’s short stack exploded view (left) with annotated
51
52
53 parts and assembly (right); Middle: Proton Energy System’s 1-kW fuel-cell stack (left) and
54
55 transparent shroud assembly (right); Below: Cathode (left) and anode (right) flow fields with
56
57
58 annotated air-coolant flow direction along the edge-cooling fins.
59
60
61
62 15
63
64
65
The bipolar plates including the fins are fabricated from a molded polymer/graphite mixture
1
2 (made by SGL Carbon) with effective thermal conductivity of 20 W m-1 K-1. The stack is
3
4
5 cooled by air flow along the longer edge of the cells where the fins are located. Instead of tie
6
7 rods, metallic shrouds that snap on the end plates are used to compress the stack [51]. The
8
9
10 cells are compressed by five polyurethane springs distributed along the centerline of the active
11
12 area. The shrouds also form the passage for cooling air. The cooling fan is installed directly
13
14
15
onto the shrouds. The stainless-steel end plates also serve as the bus plates. The seven-layer
16
17 MEAs are manufactured by 3M and include a catalytic membrane and two carbon paper gas
18
19 diffusion layers with integrated gaskets. The active part of the stack therefore consists of only
20
21
22 two alternating components: 60 MEAs sandwiched between 61 collector plates. Besides these
23
24 repetitive components, the rest of the stack consists of only 13 additional parts (three end
25
26
27 plates, two shrouds, five polyurethane springs and four fittings). Stack properties and
28
29 operating conditions are outlined in Table 1.
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62 16
63
64
65
Property Value, unit
1
2 Number of cells 60, /
3 Active area 65 cm2
4
5 Aspect ratio (L:W) 10:1, /
6
7 Bounding dimensions (31.0 × 23.6 × 9.6) cm
8
9 Volume 7023.36 cm3
10 Mass 17 kg
11
12 Specific heat 0.59 kJ kg-1 K-1
13
14 Heat exchange area 0.9378 m2
15
16 Coolant Air
17 Nominal power 1 kW
18
19 Current 23 A
20
21 Electric potential 43.5 V
22
23 Anode stoichiometry 1.2
24 Cathode stoichiometry 2
25
26 Operating temperature 60 °C
27
28 Reactant’s inlet temperature 20 °C
29
30 Back-pressure 0 Pa gauge pressure
31 Pressure drop along the cathode side 20∙103 Pa
32
33 Reactant’s inlet relative humidity 0%
34
35 Cooling control On-Off
36
37 Table 1. Stack properties and operating conditions.
38
39 The stack is designed to generate 1 kW power output at relatively high single-cell operating
40
41
42 potential of ≥0.7 V cell-1. The cell operating current was limited due to poor thermal
43
44 conductivity of the bipolar plates, which would cause large temperature gradients between the
45
46
edge and center of the cell at lower operating potentials, i.e. higher currents, and consequently
47
48
49 result in very poor performance of the cell due to sever membrane dehydration. The stack was
50
51 tested at the Connecticut Global Fuel Cell Center, University of Connecticut [49]. The
52
53
54 resulting polarization curve is shown in Figure 2. The projected potential and projected power
55
56 and extrapolated by fitting a polynomial curve through the experimentally obtained points,
57
58
59 while the operating current was limited to 25 A to prevent overheating.
60
61
62 17
63
64
65
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18 Figure 2. Polarization curve vs. power curve.
19
20 The focus of this research is to determine the most feasible method for minimizing the
21
22
23 maximal temperatures occurring inside the PS, thus enabling operation at higher currents and
24
25 increasing the operating efficiency which will be a consequence of higher membrane water
26
27
content, i.e. higher relative humidity profile, along the entire active area of the cell.
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62 18
63
64
65
3. LUMPED MODEL
1
2
3 The objective of the development of the lumped model was to study the transient and steady-
4
5 state behavior of a fuel cell stack under variable loads. The mathematical model describes the
6
7
8 fuel cell voltage as a function of temperature and heat energy balance of the fuel cell stack.
9
10 The model also includes heat dissipation from the stack to surroundings and coolant energy
11
12
balance equations. A complete stack energy balance was considered, including the heat
13
14
15 generated by the inlet reactant gases (the enthalpy of inlet reactant gases), and the heat of
16
17 gases leaving the stack. The calculation includes both the latent and sensible heat of water
18
19
20 flows at the stack inlet and outlet. Generated heat is dissipated from the fuel cell stack to the
21
22 surroundings via radiation and natural convection (through the horizontal and vertical plates
23
24
25 of the stack). The amount of heat dissipation depends on the temperature of the stack, the
26
27 ambient temperature and the properties of the surrounding air. Dissipation is at a maximum
28
29
30
when the stack reaches (operates under) steady state. The model also includes an energy
31
32 balance equation for the coolant air. The temperature of the coolant varies with time and
33
34 distance along the active length of the fuel cell. The heat capacity of the fuel cell stack was
35
36
37 determined by calculating the heat capacities of each individual component of the fuel cell
38
39 such as graphite, aluminum and stainless steel, and summarizing them.
40
41
42 The assumptions regarding the conducted calculations are summarized as follows in the
43
44 described mathematical model:
45
46
47  The fuel cell stack operates at constant current vs. temperature, i.e. galvanostatic
48
49 approach
50
51
52  Edge cooling mechanism is used to remove heat from the stack
53
54  Air is used as the coolant
55
56
57  Flow is laminar and the temperature distribution is uniform, therefore Nusselt number
58
59 is 2.98
60
61
62 19
63
64
65
 Water back-diffusion is equal to electro-osmotic drag, i.e. there is zero net water
1
2 transport across the membrane
3
4
5  Reactants are completely dry at the inlet
6
7
8
3.1. Steady-state model
9
10 With known operating conditions and stack geometry, it is possible to calculate the heat
11
12
13 fluxes in and out of the stack at steady state. The resulting stack energy balance at full power
14
15 (60 °C) is shown in Table 2.
16
17
18
19 Electrical power 1 kW
20
21 Reactant’s enthalpy at the inlet 0.054 kW
22
23 Produced heat 1.044 kW
24
25
Reactant’s enthalpy at the outlet 0.408 kW
26 Heat dissipation 0.090 kW
27
28 Heat to be removed by cooling air 0.601 kW
29
30 Table 2. Stack energy balance.
31
32 Since the mean operating potential of each single-cell in the stack is 0.725 V, the stack
33
34
35 efficiency is close to 50 %, and the amount of generated heat is close to the amount of
36
37 generated electrical power. The heat is removed from the stack by the flow of excess air
38
39
40
carrying evaporated product water (408 – 54 = 354 W) and by radiation and convection to the
41
42 surrounding air (90 W). The remaining 600 W must be removed by the cooling air to avoid
43
44 overheating.
45
46
47 The heat removed by the cooling air is
48
49 (1)
50
51
52 where represents cooling air mass flow (kg s-1) rate, specific heat (J kg-1 K-1) of air,
53
54
55 outlet and inlet temperature (K) of air. The heat must be transferred from the
56
57 stack to the cooling air, i.e.
58
59
60
61
62 20
63
64
65
(2)
1
2
where represents the average heat transfer coefficient (W m-2 K-1), exchange area (m2)
3
4
5 and is the logarithmic mean temperature difference (K) between the stack and the
6
7 cooling air.
8
9
10 The stack releases heat during operation and the total amount of generated heat consists of
11
12 (3)
13
14
15 Where represents the total current (A) generated by the cell, while is overpotential (V),
16
17 i.e. difference between the theoretical potential for hydrogen lower heating value of 1.254 V,
18
19
20 and the electric potential of the cell, calculated as
21
22 (4)
23
24
25 Where is reversible Nernst potential (V), while is calculated as
26
27 (5)
28
29
30 Where , and represent the activation, ohmic and concentration overpotential
31
32 (V), and reversible Nernst potential is calculated as
33
34
35 (6)
36
37
38
39 where
40
41 (7)
42
43
44
45
46
47
48 And temperature (K), pressure (atm), change in entropy (J mol-1 K-1), number of
49
50 electrons transferred, Faraday’s constant (C mol-1) and universal gas constant (J mol-1 K-
51
52 1
53
), and subscripts , , and refer to the reference values, hydrogen, oxygen and
54
55 water, respectively. Since the experimental polarization curve is available, as seen previously
56
57
58
in Figure 2., in this model the is calculated simply by subtracting the cell operating
59
60 voltage from the theoretical potential for lower hydrogen heating value of 1.254 V.
61
62 21
63
64
65
The water activity , i.e. relative humidity RH, is calculated as
1
2 (8)
3
4
5
6 Where is partial pressure (Pa) of water vapor and is saturation pressure (Pa),
7
8
9 calculated as
10
11 (9)
12
13
14
15
16 For given stack and ambient conditions, the only variables in eqs. 1 and 2 are the coolant flow
17
18 rate, air outlet temperature and stack temperature. Figure 3 shows the cooling air and stack
19
20
21 temperatures as functions of the cooling air flow rate and the stack heat transfer coefficient
22
23 product ( ). During stack testing [49], it was observed that the air exit temperature did not
24
25
26 rise above 30 °C, due to the high air flow rates. The resulting stack temperature could not be
27
28 maintained at the desired value of 60 °C despite the high air flow rates. This indicates that the
29
30
31 heat transfer ( ) was insufficient (i.e. < 15). Doubling the heat exchange area, for
32
33 example, would allow not only operation at the desired stack temperature but would also
34
35
36
allow operation with a cooling air flow rate one fifth of the one used initially. Additionally,
37
38 the resulting temperature of the cooling air leaving the stack could be sufficient for extracting
39
40 hydrogen from a metal hydride storage unit.
41
42
43 100
44 hA = 15
90
45 hA = 20
46 80 hA = 30
47 70 Tair out
temperature, °C
48 60
49 Tstack
50
50
51 40
52 30
53 20
54
55 10
56 0
57 0 20 40 60 80 100 120
58 flow rate (mCp)
59
60
61
62 22
63
64
65
Figure 3. Stack and cooling air steady-state outlet temperatures as a function of cooling air
1
2 flow rate and overall stack heat transfer coefficient.
3
4
5 3.2. Transient model
6
7
8 A MATLAB model of the stack mass and heat fluxes was developed and used to model the
9
10 transient behavior, particularly at start-up and at stepwise power change. Operation of the
11
12
stack was simulated starting at room temperature (25°C) with nominal electrical power of 1
13
14
15 kW (current 23 A), reaching steady-state after ca. 5500 s, after which the current was abruptly
16
17 decreased to 11.5 A for duration of 1000 s and then increased back to 23 A until total time of
18
19
20 10000 s has passed. Since the stack temperature was initially at ambient conditions, the stack
21
22 did not reach the nominal power until the operating temperature of 60 °C was reached. The
23
24
25 coolant fan was not required for the first 500 s. When the stack reached 60 °C the cooling fan
26
27 was started. With the designated mass flow rate, the temperature of the stack during steady-
28
29
30
state reached and stabilized at 67 °C. Three scenarios are studied: (i) power and cooling fans
31
32 are turned off at 10000 s time step; (ii) current was reduced to 11.5 A at 10000 s time step and
33
34 the fans continued to run at mass flow rate of 0.1 kg s-1 and (iii) the current was reduced to
35
36
37 11.5 A and the cooling fans are turned off and on between 55 °C and 60 °C from 10000
38
39 seconds to 15000 seconds, resulting in saw-tooth temperature profile visible in Figure 4.
40
41
42 (right). In Figure 4. it can be noticed that for case (ii) another steady state is reached after
43
44 approx. 55 minutes. These results were achieved due to the large thermal mass of the stack, 17
45
46
47 kg at 0.59 kJ kg-1 K-1, as calculated from the masses and specific heats of the stack
48
49 components. This value is comparable to the previously reported specific heat value for an
50
51
early Ballard stack of 0.7 kJ kg-1 K-1 [52].
52
53
54
55
56
57
58
59
60
61
62 23
63
64
65
1000 70
1
900 65
2
3 800 60
4
Fuel Cell Temperature (C)

700 55
5
Electrical Power (W)
600 50
6
7 500 45
8 400 40
9 power off/ cooling fan off
10 300 35
0.1 kg/s
11 cooling fan on/off
200 30
power off/ cooling fan off
12 100 25
0,1 kg/s
14 0 20
0 5000 10000 15000 0 5000 10000 15000
15 Time (sec) Time ( sec)
16
17
18 Figure 4. Resulting electrical power (left) and temperature (right) during startup from room
19
20 temperature, at current of 23 A), 11.5 A and at 0 A for 3 strategies after 10000 s time step.
21
22
Since the experimental data on temperature vs. load (i.e. current) profile is not available for
23
24 the PS shown in this work, the results of the lumped model can be compared with
25
26 experimental data conducted by Mahjoubi and Olivier [53,54] for a similar stack (1.4 kW
27
28 power, fan-cooled) during a current ramp and the consequent temperature ramp from 30 °C to
29
30 60 °C, as seen in Figure 5., which is very similar to the setup of the stack operating conditions
31
32 in this work. By comparing the results in Figure 5. with results in Figure 4. it can be can
33 observed that the temperature profile which is a consequence of current ramp-up is of quite
34
35 similar shape, as well as the resulting temperature profile during a sudden decrease in the
36
37 current Figure 5. (below), and also during the decrease of current from 20 A to 0 A, as seen in
38
39 Figure 5. (above).
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62 24
63
64
65
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27 Figure 5. Current and temperature history of air-cooled, open-cathode 1.4 kW PEM fuel cell
28
29 stack, images adopted from [53] (above) and [54] (below).
30
31
32
33
34 The resulting liquid water mass flow in the outlet air from the stack is shown in Figure 6 for
35
36
37 cases (ii) and (iii). It can be seen that during the stack startup and the sudden decrease in the
38
39 stack current from 23 A to 11.5 A as well as the operating temperature of the stack after 5500
40
41
42 s, i.e. whenever the stack temperature is below 60 °C, condensed water is present at the
43
44 cathode outlet. Prolonged holds at these operating conditions would result in severe flooding
45
46
of the cell. After 1000 s hold at 11.5 A, the current is increased to 23 A and the liquid water is
47
48
49 evaporated. After 10000 s time step, it can be seen that the liquid water depends on the stack
50
51 cooling strategy and that it can be controlled simply by adjusting the appropriate cooling air
52
53
54 mass flow rate. The results of the lumped model are compared vs. results of CFD analysis in
55
56 later chapters. The amount of generated water is calculated using Faraday’s equation and
57
58
59 numerical representation of Mollier’s h-x chart, Figure 6., it can be seen that water condenses
60
61
62 25
63
64
65
during the load ramp-down due to sudden temperature decrease and it is evaporated once the
1
2 load is ramped-up and the temperature is increased.
3
4
5
0.14
6 0,1 kg/s
0.12
8
9
10 0.1
11
H2OLoutAir (g/s)
12 0.08
13
14 0.06
15
16 0.04
17
18
0.02
19
20
21 0
0 5000 10000 15000
22 Time (sec)
23
24
25 Figure 6. Resulting flow of liquid water at the cathode outlet during start-up and operation at
26
27 full power at 23 A and at 11.5 A.
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62 26
63
64
65
4. COMPUTATIONAL FLUID DYNAMICS MODEL
1
2
3 A more advanced numerical model was built using CFD modeling. The whole stack was not
4
5 modeled due to high computational requirements for conducting such analysis and only a
6
7
8 limited benefit. The CFD model represents one slice of the stack incorporating one single cell
9
10 and the outside domain where the coolant air flows. The entire stack could be represented by
11
12
patterning this single cell, while the main difference vs. more realistic model would be seen
13
14
15 only near the end plates due to somewhat different heat transfer, but this would result in huge
16
17 increase in computational cost, while it would only result in minor amount of additional
18
19
20 information. However, as previously discussed in the Introduction chapter, the simplified
21
22 CFD models give reasonably accurate prediction of PEM fuel cell performance therefore this
23
24
25 approach is reasonable due to the requirement for conducting transient simulations. The
26
27 single-cell which represents a repetitive segment of a stack was modeled and several steps
28
29
30
were introduced to ensure that the scale-up is proper. It was also not convenient to model the
31
32 actual processes inside the cell in conjunction with the heat transfer between the cell and
33
34 coolant air due to transient nature of the calculations and the lack of experimental data for
35
36
37 localized validation of the model. Since the stack consists of 60 single cells, and they are
38
39 connected to common manifolds, therefore it is required to determine the influence of the cell
40
41
42 distance from the inlet on the overall mass flow rate in each particular cell. The decrease in
43
44 the pressure potential (a consequence of the viscous losses, which are a result of flow of the
45
46
47 reactants through the common manifolds), results in non-uniform distribution of the mass
48
49 flow rate vs. distance from the common inlet. This phenomenon is interesting to determine if
50
51
the cross-section of the designed manifolds is sufficiently large to prevent non-uniform
52
53
54 distribution of the reactants vs. stack height. Also, since the pressure drop across the cathode
55
56 side of the flow field is measured experimentally, it is possible to determine the viscous
57
58
59 resistance of the cathode flow field by using a default value in the analysis and iteratively
60
61
62 27
63
64
65
adjusting it based on the overall pressure experimentally measured pressure drop until the
1
2 values show good agreement, a parameter which will later be used for more detailed transient
3
4
5 analysis. The model for determining the mass flow rate distribution vs. stack height is shown
6
7 below as well as the resulting viscous resistance of the cathode side flow field.
8
9
10 4.1. Governing equations and boundary conditions
11
12
The developed numerical CFD models are based on solving the Navier-Stokes equations in
13
14
15 three-dimensional, unsteady-form (for convenience referred to as sink-source formulation) for
16
17 continuity:
18
19
20 (3)
21
22
23
24 represents density, time, , and velocities in , and , i.e. Cartesian coordinate
25
26 directions, respectively and mass source term.
27
28
29 Momentum in -direction:
30
31 (4)
32
33
34
35 where represents pressure, Reynolds number, and tangential stresses in the Cartesian
36
37
38 coordinate directions.
39
41
42
43 (5)
44
45
46
48
49 (6)
50
51
52
53 as well as energy equation
54
55
56
57
58
59
60
61
62 28
63
64
65
(7)
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15 where represents thermal energy, Prandtl number and heat flux in Cartesian
16
17
19
20 The sink and source terms for the mass transfer, i.e. more precisely continuity equation, are
21
22
23 defined as
24
25 (8)
26
27
28
29
30 where represents the cathode stoichiometry, current density, orthogonal area of the
31
32 flow field domain perpendicular to the stack height, without the inflow and outflow
33
34
35 manifolds, Faraday's constant, molecular mass of oxygen and the total volume of
36
37 the numerical flow field domain.
38
39
40 (9)
41
42
43
44
45 the pressure potential loss across the cathode side flow field is modeled by introducing a
46
47 momentum sink term according to the porous media formulation
48
49
50
(10)
51
52
53
54
55 where represents the momentum sink term in -th direction, and are the matrices for
56
57
58
the viscous and inertial resistance, respectively, absolute viscosity, directional velocity
59
60 and velocity magnitude. Since the flow velocities are relatively low, the inertial term on
61
62 29
63
64
65
the right is neglected, i.e. . The viscous resistance matrix is defined as orthotropic,
1
2
3
with maximal value (1020 m-2) in the direction of the stack height, i.e. -direction, to prevent
4
5 the flow and mixing of the gas between the adjacent cells in the stack. The values of the
6
7 viscous resistance in and directions are set as equal and determined based on the resulting
8
9
10 pressure potential loss across the flow field. The porosity is set to 0.6 by comparing the
11
12 volume of the reactant channels vs. volume of the flow field channels plus lands.
13
14
15 4.2. Stack manifold model
16
17
18 The objective of the stack manifold simulations was to determine if the cross-section area of
19
20 the common manifolds is sufficiently large to ensure similar mass flow rates between the
21
22
23
neighboring cells along the stack height. The geometry consists of common inflow and
24
25 outflow manifolds, as seen in Figure 7.
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46 Figure 7. Stack manifold and flow field geometry (left) and mesh (right).
47
48 The applied boundary conditions are shown in Table 3. Viscous resistance for in-plane
49
50
51 directions is calculated to result in pressure drop across the flow field close to 2 kPa, as
52
53 experimentally measured. Viscous resistance for through-plane direction is set to the maximal
54
55
56 value to prevent mixing between the neighboring cells, i.e. to result only in flow direction
57
58 perpendicular to the stack height. The heat generation inside the cell is not considered in this
59
60
61
62 30
63
64
65
case because the stack is air cooled and this is addressed in the following chapter in more
1
2 detail, while in this case the temperature of the outside walls is fixed at 333 K.
3
4
5 Domain Boundary condition/property Value
6
7 Inflow domain Mass flow inlet air (mixture model) 9.86549∙10-4 kg s-1
8
9
10 Relative humidity 0 (dry)
11
12 Temperature 293.15 K
13
14
15 Outflow domain Pressure outlet Zero gauge pressure
16
17 Temperature 333 K
18
19
Flow field Porosity 0.6
20
21
22 Viscous resistance in-plane 3.75∙109 m-2
23
24 Viscous resistance through-plane 1020 m-2
25
26
27 Volume 1.2675∙10-3 m3
28
29 Source H2O 0.1016482 kg m-3 s-1
30
31
32 Source O2 -0.0890934 kg m-3 s-1
33
34 Walls Temperature 333 K
35
36
37
Table 3. Operating conditions for the simplified stack simulations.
38
39 Since the mass flow rate on the cathode side is approx. 57 times higher vs. on the anode side,
40
41 and since pure hydrogen is used on the anode, it was not necessary to consider the flow on the
42
43
44 anode side since it is unlikely that the starvation will occur since pure hydrogen is used as
45
46 fuel. For this reason, only cathode side is modeled, as well as the accompanying oxygen
47
48
49 depletion and water generation (single-phase modeling, only water vapor is considered). The
50
51 cathode gas composition consists of mixture with species nitrogen, oxygen and water vapor.
52
53
54 The inflow and outflow manifold geometry is simplified to ensure smooth mesh generation,
55
56 mesh is grid independent and the total number of elements for the mesh is 72800.
57
58
59
60
61
62 31
63
64
65
The results shown in Figure 8 indicate that the mass flow rate is uniformly distributed along
1
2 the stack height and the species contours indicate results consistent with the mass balance of
3
4
5 PEM fuel cell. The resulting pressure drop is close to 2 kPa, as experimentally measured. The
6
7 conclusion of stack manifold simulations is that the thermal management analysis can be
8
9
10 carried out for a single cell with symmetry boundary conditions due to uniform reactant
11
12 distribution along the stack height.
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56 Figure 8. Contours of temperature, species, velocity and pressure. Inlet left, outlet right.
57
58
59
4.3. Transient single cell model
60
61
62 32
63
64
65
The entire stack is not modeled due to high computational requirements for conducting
1
2 transient analysis and very little additional data provided from such simulations. The stack
3
4
5 manifold simulations indicate that this is a reasonable approach, as seen in the previous
6
7 chapter.
8
9
10 4.3.1. Geometry and boundary conditions
11
12 The geometry of the single cell representing the 1/60th portion of the stack is shown in Figure
13
14
15
9., the transparent domains on the geometry figure represent fluid material, while opaque
16
17 domains represent solid material. On the left side the decomposed geometry and mesh are
18
19 shown to see the geometry and mesh in more detail. The left top represents the assembles
20
21
22 solid and fluid meshes, left bottom represents fluid domain mesh and middle above represents
23
24 solid domain mesh.
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49 Figure 9. Default mesh configuration (left and middle above) with details of solid and fluid
50
51 meshes, and geometry with annotated boundary conditions (right).
52
53 Once again, the hydrogen flow and consumption is not considered due to much lower mass
54
55
56 flow rate vs. cathode air. The flow field is modeled as fluid domain with prescribed porosity
57
58 to represent the homogenized actual flow field. The geometry is decomposed into blocks to
59
60
61
62 33
63
64
65
enable generation of fully structured pure hexahedral mesh. The air enters the cell in the
1
2 Inflow domain and flows through the porous flow field domain where the pressure drop is
3
4
5 represented via momentum sink term, while the value is previously determined from the stack
6
7 manifold simulation. The oxygen depletion, water vapor genesis and heat release is defined by
8
9
10 sink term for oxygen and source terms for water vapor and energy equation to represent the
11
12 heat release. Air is modeled as a mixture consisting of oxygen, nitrogen and water vapor using
13
14
15
standard species mixture model. The material of the bipolar plate is defined for 4
16
17 configurations: molded graphite, nickel, stainless steel and titanium. Molded graphite results
18
19 are later compared vs. lumped model, while the remaining materials are mutually compared to
20
21
22 determine the influence of the bipolar plate material on the overall heat transfer inside the cell
23
24 and the resulting relative humidity distribution along the flow field. The application of
25
26
27 symmetry boundary conditions was introduced in the model since it was determined that the
28
29 mass flow of the reactants is very uniformly distributed along the stack height (as previously
30
31
32
seen in section 4.2) and due to relatively high forced convection air flow rates which result in
33
34 uniform temperature profile along the stack height since the edge-cooling fins also serve as
35
36 guide vanes for the air flow, ensuring uniform mass flow rate along the stack height. This
37
38
39 approach is justified and commonly used since in that case the contribution of natural
40
41 convection, i.e. updraft of air due to heating, on the overall temperature gradient along stack
42
43
44 height is minor, and even if it occurs, it will be minimized by the convective heat transfer
45
46 from the massive end plates (as seen previously in Figure 1.).
47
48
49 4.3.2. Cooling fin configurations
50
51 Besides different bipolar plate materials, different cooling fin configurations are used to
52
53
54
determine how the temperature gradient across the stack cross-section can be reduced by
55
56 design. In Figure 10 six configurations are shown. Configurations 1 to 4 consider the re-
57
58 design of the edge-cooling fins, while the configurations 5 and 6 consider installment of guide
59
60
61
62 34
63
64
65
vanes on the side plates, while the original geometry of the edge-cooling fins is unaltered. The
1
2 geometry is slightly differently decomposed for every case, but the mesh settings are
3
4
5 identical.
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43 Figure 10. Cooling fin configurations 1 to 6 with annotated guide vanes on the side plates
44
45 (red circles).
46
47
48 4.4. Mesh
49
50
51 Due to different configurations of the cooling fins, the mesh for each case is slightly different.
52
53 The default mesh configuration was previously shown in Figure 9, while the altered version is
54
55
56
shown in Figure 11. The element sizing, methods and number of divisions along the edges in
57
58 all meshes is the same.
59
60
61
62 35
63
64
65
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
Figure 11. Altered version of the mesh (above) with decomposed cell domain (below left) and
32
33 the entire domain (below right).
34
35 Grid dependency study is conducted to determine the required mesh configuration which will
36
37
38 result in grid independent solution. Since pure hexahedral mesh is generated, the mesh size
39
40 was controlled via global minimal and maximal allowable element size. Due to transient
41
42
43 nature of the calculations, it was desirable to compare the results under transient operating
44
45 conditions. During the initial runs it was observed that the results for 10 second time-step are
46
47
48 similar to the results at smaller time-steps, therefore the reference time-step for conducting the
49
50 grip dependency study was 10 s. The total run time for grid and time-step dependency study
51
52
53
was 1000 seconds, and two parameters relevant for this study are observed, cell averaged
54
55 temperature and flow field averaged temperature. The cell averaged temperature represents
56
57 volume integral of temperature in all domains except fluid domain representing the coolant
58
59
60 air. The flow field averaged temperature represents the volume integral temperature of the
61
62 36
63
64
65
flow field without the inlet and outlet domains. Grid dependency study consisted of four mesh
1
2 setups, shown in Table 4. The bipolar plate material used during the dependency study was
3
4
5 molded graphite.
6
7 Parameter Max. element size, mm Finite volume count
8
9
10 Mesh #1 1.25 41286
11
12 Mesh #2 1.00 71370
13
14
15 Mesh #3 0.80 150880
16
17 Mesh #4 0.65 212356
18
19
20
Table 4. Mesh details
21
22 4.4.1. Grid dependency and time-step dependency study
23
24 The grid dependency study is shown in Figure 12 (above). It can be observed that mesh #3 is
25
26
27 grid independent since it shows results very similar to Mesh #4 and is less computationally
28
29 intensive. Since the flow is laminar, there was no necessity to introduce mesh refinement on
30
31
32 the walls of the cell.
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62 37
63
64
65
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38 Figure 12. Grid (above) and time-step (below) dependency vs. flow time.
39
40 Grid independent mesh is used to determine the required time-step size for conducting the
41
42
43 numerical analysis. The results for the default time-step size of 10 s is compared with several
44
45 other time-steps for the grid independent mesh and the results are shown in Figure 12 (right).
46
47
48 It can be seen that time-step size of 10 s is grid independent since it gives results very similar
49
50 to the results of time-step of 5 s, while larger time-steps result in higher difference although
51
52
the temperature reached after 1000 s elapsed time is quite similar. Time-step of 10 s is chosen
53
54
55 for conducting the numerical analysis.
56
57
58
59
60
61
62 38
63
64
65
5. RESULTS AND DISCUSSION
1
2
3 5.1. Startup and transient operation
4
5
6
The results of the lumped model and CFD analysis for the load profiles described previously
7
8 in section 3.2 are shown in Figure 13. It can be seen that the cell temperature increases from
9
10 ambient to steady-state shows similar trends, although the lumped model predicts slightly
11
12
13 lower steady-state temperature. The CFD data also indicates that the temperature fluctuations
14
15 are always present due to the unsteady nature of the flow and modeling of the air as ideal gas,
16
17
18 resulting in convective flow and vortex genesis in the wake region behind the stack. The
19
20 temperature decrease after 5500 s reaches higher minimal value in the CFD model vs. lumped
21
22
23 model, and after increasing the current the temperature oscillations of the CFD model are
24
25 more severe and it can be noted that the steady-state predicted by the lumped model near ca.
26
27
28
10000 s is actually not yet steady-state when CFD model is observed. The highest difference
29
30 between the lumped model vs. CFD model is evident after 10000 seconds by comparing the 3
31
32 cases. For case (i) it can be seen that the steady-state temperature of the cell in the CFD model
33
34
35 is decreased at a slower rate when compared to the lumped model and converges to a
36
37 significantly higher steady-state value (by ca. 10 °C), and similarly also in case (ii). In case
38
39
40 (iii) it can be seen that the fan must be turned on and off approx. 6 times more frequent to
41
42 keep the stack temperature between the 55 and 60 °C range. The differences are a result of
43
44
45
directly calculating the heat transfer coefficient in the CFD analysis vs. fixing the value in the
46
47 lumped model and a result of modeling the complex flow involving air as ideal gas in the
48
49 CFD model. Regardless of the differences, it can be noted that the average temperatures
50
51
52 predicted by the lumped model and time to reach steady-state are reasonably well predicted
53
54 by the lumped model but the CFD model is more physically valid therefore it will be used in
55
56
57 the following chapters for more detailed analysis.
58
59
60
61
62 39
63
64
65
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19 Figure 13. History of cell temperature: lumped model vs. CFD for cases (i), (ii) and (iii).
20
21
22 Although there is no available experimental data for temperature profile during transient load
23
24 changer, the characteristic saw tooth temperature profile seen in Figure 13 during transient
25
26
load change is reported in works [55] and [56] indicating that the numerical models are valid.
27
28
29 5.2. Influence of bipolar plate material on performance of the stack
30
31
32 Since bipolar plates of PEM fuel cells are made from different materials, four different
33
34 commonly used materials are considered in the study to determine the resulting average and
35
36
37 maximal stack temperature during transient operation upon reaching steady state. The chosen
38
39 materials are molded graphite, nickel, stainless steel (SS316L) and titanium. The thermal
40
41
42 properties of the materials are shown in Table 5.
43
44 Material Density, kg m-3 Specific heat, J kg-1 K-1 Thermal conductivity, W m-1 K-1
45
46
47 Graphite 1900 741 20
48
49 Nickel 8900 460.6 91.74
50
51
52 SS316L 8000 500 16.3
53
54 Titanium 4850 544.25 7.44
55
56
57
Table 5. Bipolar plate thermal properties.
58
59
60
61
62 40
63
64
65
Since the graphite is used for most of the thermal studies in this work in the remaining
1
2 sections with extensive information provided, only nickel, stainless steel and titanium will be
3
4
5 shown in more detail in this chapter to avoid repetition. Temperature and relative humidity
6
7 contours after 5500 s (upon reaching steady state) are shown in Figure 14. The lowest
8
9
10 temperature gradient across the stack is achieved by using nickel as bipolar plate material, due
11
12 to high thermal conductivity, which also results in highest overall relative humidity across the
13
14
15
flow field. Stainless steel and titanium bipolar plates result in highly non-uniform temperature
16
17 distribution across the stack and low overall relative humidity distribution, leading to poor
18
19 performance of the cell, therefore it is recommended to avoid using materials with such low
20
21
22 thermal conductivity for this stack design.
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62 41
63
64
65
1
2
3
4
5
6
7
8
9
10
11
12 Figure 14. Temperature (left) and relative humidity (right) contours for across the entire
13
14
15 domain for nickel (above), stainless steel (middle) and titanium (below) at 1 kW electrical
16
17 power operation.
18
19
20 For quantitative assessment, the average and maximal stack temperatures are shown in Figure
21
22 15. It can be noted that the stack average temperature is very similar for thermal conductivity
23
24
25 of 20 W m-1 K-1 and above, while the maximal stack temperature abruptly increases below 20
26
27 W m-1 K-1. This indicates that the edge-cooled stack performance heavily depends on the
28
29
30
bipolar plate thermal conductivity, because the relative humidity distribution inside the cell is
31
32 affected by the internal temperature distribution inside the stack.
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56 Figure 15. Stack average and maximal temperature vs. thermal conductivity.
57
58
59
60
61
62 42
63
64
65
The results shown in Figure 15. are also in good agreement with the results of the analysis
1
2 conducted by Burheim and Pharoah [57] where they have observed that the maximal
3
4
5 temperature inside the MEA can be 14 – 21 °C higher than the temperature on the surface of
6
7 the bipolar plate in PEM fuel cell.
8
9
10
11
12
13
5.3. Influence of operating pressure on performance of the stack
14
15 The temperature distribution inside the edge-cooled stack is non-uniformly distributed since
16
17
18 the original bipolar plates are manufactured from molded graphite with thermal conductivity
19
20 of 20 W m-1 K-1, other method for improving performance, i.e. increasing the relative
21
22
23
humidity profile inside the cell, must be devised. Since the water vapor saturation profile is
24
25 pressure and temperature dependent, the operating pressure inside the cell can be increased
26
27 and thus the resulting relative humidity profile can be manipulated. In Figure 16 different
28
29
30 delta pressures are applied, and the resulting relative humidity profiles are plotted.
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46 Figure 16. Relative humidity profile inside the cell operated at different pressures.
47
48 It can be seen that increasing the delta pressure results in much higher overall relative
49
50
51 humidity profile inside the cell, but for efficient operation the operating delta pressure must be
52
53 sufficiently low (i.e. at most in several kPa) and it is easier to manipulate the relative humidity
54
55
56
profile via achieving a more uniform temperature distribution. Thus, the cooling fan redesign
57
58 is a feasible option.
59
60
61
62 43
63
64
65
5.4. Influence of cooling fin redesign
1
2
3 Six different cooling fan designs are studied. The geometry is shown previously in Figure 10.
4
5 The results are compared after reaching steady-state, by startup from room temperature, at
6
7
8 time of 5500 s. Temperature and relative humidity distributions across the entire domain are
9
10 shown in Figure 17 and Figure 18. In Figure 17 (middle) for configuration 2 it can be seen
11
12
that the influence of the cooling fan design is so significant that it is possible to achieve 100%
13
14
15 relative humidity at ca. 50% channel length in some areas of the cell and especially near the
16
17 outlet of the cell when dry air is used at the inlet even when graphite is used as bipolar plate
18
19
20 material, while previously in Figure 14. this could not be done even when nickel was used.
21
22 This indicates that the influence of cooling fin design is more significant related to the overall
23
24
25 performance of the cell vs the bipolar plate material for this type of PS. Ideally, both material
26
27 and fin design should be carefully coupled to reap the benefits and enhance the performance
28
29
30
of the cell.
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62 44
63
64
65
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39 Figure 17. Temperature (left) and relative humidity (right) distributions across the entire
40
41 domain for configuration 1 (above), 2 (middle) and 3 (below).
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62 45
63
64
65
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
40
42
43
44
For quantitative assessment, the maximal and average temperature for the original setup and
45
46 each configuration are shown in Table 6. The lowest maximal and average cell temperature
47
48 and is achieved for case 2. However, since this design would require redesigning the positions
49
50
51 of the assembly bolts and the entire stack frame. Besides case 2, second case with lowest
52
53 maximal temperature is case 6 (63.55 °C vs. 77.60 °C for the original geometry), while the
54
55
56 second case with lowest average temperature is case 4 (44.15 °C vs. 57.53 °C). Since case 6
57
58 requires only installment of two guide fins on the side plates, while case 4 requires increasing
59
60
61
62 46
63
64
65
the size of the original bipolar plates considerably for increasing the area of the fins, case 6 is
1
2 considered the optimal solution. Therefore, by installing the guide vanes on the side plates it
3
4
5 is possible to reduce the maximal cell temperature by 14.05 °C and cell average temperature
6
7 by 10.96 °C when compared to the original configuration. Nevertheless, both versions 4 and 6
8
9
10 will be used to study the heat transfer between the stack and the metal hydride hydrogen
11
12 storage tank.
13
14
15
Configuration Max. temperature, °C Aver. temperature, °C
16
17 Original 77.60 57.53
18
19
1 70.53 50.36
20
21
22 2 61.48 43.26
23
24 3 89.07 61.21
25
26
27 4 63.87 44.15
28
29 5 67.43 47.95
30
31
32 6 63.55 46.57
33
34 Table 6. Maximal and average temperature of the stack for different cooling fin
35
36
37
configurations.
38
39
40
41
42 5.5. Utilizing waste heat for releasing hydrogen from MH tank
43
44
45
Configurations 4 and 6 are designed to incorporate the metal hydride (MH) storage tank in
46
47 proximity right behind the stack. The MH tank has aluminum casing, while the MH material
48
49 is MmNiMnCo, with hydrogen capacity of 72 g, diameter of the tank is 89 mm and length is
50
51
52 406 mm. Thermal conductivity of MH material is 1.32 W m-1 K-1 and specific heat 600 J kg-1
53
54 K-1. The MH tank is placed on 90 mm from the stack, while the flow domain was extended
55
56
57 behind the MH tank for additional length of 3 diameters of the tank to minimize the
58
59 occurrence of the reversed flow. The geometry is shown in Figure 19.
60
61
62 47
63
64
65
1
2
3
4
5
6
7
8
9
10
11
12 Figure 19. Configurations 4 (left) and 6 (right) with included MH storage tank.
13
14
15 The MH tank includes heat sink term to accurately thermodynamically describe the physical
16
17 process of hydrogen desorption. Using the default coolant air mass flow rate, the average
18
19
20 temperature of the MH tank under steady-state operation (after 5500 s) is 15.15 °C for
21
22 configuration 4 and 19.06 °C for configuration 6. The temperature contours are shown in
23
24
25
Figure 20.
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40 Figure 20. Temperature distribution across the stack and MH tank for configuration 4 (left)
41
42 and 6 (right).
43
44
Since the steady-state temperature of the MH tank is very low, the hydrogen release was poor.
45
46
47 For this reason, to enhance the hydrogen release, the coolant air mass flow rate was gradually
48
49 decreased to result in increased temperature of the cell and MH tank. The results for different
50
51
52 coolant air flow velocities are shown in Table 7. It can be seen that the configuration 6 is
53
54 better suited for installing MH tank behind the stack since the heat transfer between the stack
55
56
57 and MH is superior at the same air flow rates when compared to configuration 4 (MH tank
58
59 temperature is of 40.30 °C is reached vs. 16.82 °C).
60
61
62 48
63
64
65
Configuration Coolant air velocity, m s-1 Average MH tank temperature, °C
1
2 5 14.59
3
4
5 4 14.28
6 4
7 2.5 16.52
8
9
10 1 16.82
11
12 5 16.89
13
14
15 4 18.41
16 6
17 2.5 23.18
18
19
20 1 40.30
21
22 Table 7. Average MH temperature vs. coolant air velocity.
23
24
25 For coolant air velocity of 1 m s-1 for case 6, the stack average temperature reaches 64.94 °C
26
27 vs. 46.57 °C using the original air flow rate (results shown previously in Table 6).
28
29
30
Nevertheless, even for the lowest flow velocity of 1 m s-1 the amount of hydrogen released
31
32 from a single MH tank can cover only ca. 14% of the requirements of the stack. For this
33
34 reason, it is recommended to use several MH tanks placed in series or to design a heat
35
36
37 exchanger to increase the heat transfer rate from the stack to the MH tank.
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62 49
63
64
65
6. CONCLUSIONS
1
2
3 Although characterized by compact size, simple design and construction, the edge-cooled
4
5 portable proton exchange membrane fuel cell stacks are quite demanding regarding the
6
7
8 thermal management and quite scarcely investigated in the literature. Although the
9
10 experimental data related to the dynamic testing of the stack is not available for this particular
11
12
portable stack, other experimental works [53,54,57] with similar setup and boundary
13
14
15 conditions are referenced for comparison and the results reveal that qualitative and
16
17 quantitative agreement with numerically obtained data from this study is very good.
18
19
20 The following conclusions are drawn from this work:
21
22 (i) although commonly used, the lumped model is not accurate for predicting the temperature
23
24
25 of the edge-cooled stacks due to the capability for only predicting the average temperature of
26
27 the stack, while the maximal temperature can be several tens of degrees higher inside the
28
29
30
stack when materials with low thermal conductivity are used, thus leading to severe
31
32 dehydration of the membrane;
33
34 (ii) by comparing the average and maximal temperature of the stack for different bipolar plate
35
36
37 materials, it can be seen that the outside temperature of the stack is quite similar for graphite
38
39 (thermal conductivity 20 W m-1 K-1) and nickel (thermal conductivity 91.74 W m-1 K-1), i.e.
40
41
42 57.53 °C vs 52.41 °C, respectively, during steady-state operation, while the internal maximal
43
44 temperature of the stack is significantly different (77.36 °C vs. 58.29 °C, respectively),
45
46
47 indicating the requirement for carefully determining the appropriate material to be used for
48
49 production of bipolar plates for edge-cooled stacks;
50
51
(iii) higher pressure drop along the flow field can be used as a supplementary method for
52
53
54 achieving higher performance of the stack since the water vapor saturation profile is pressure-
55
56 dependent, or the whole stack can be operated at elevated pressure to increase the
57
58
59 performance;
60
61
62 50
63
64
65
(iv) the edge-cooling fin geometry has significant influence on the overall temperature
1
2 distribution, by redesigning the edge-cooling fins the maximal temperature and average
3
4
5 temperature of the stack was reduced from 77.60 °C and 57.53 °C to 63.55 °C and 46.57 °C,
6
7 respectively, indicating the necessity for careful design of the flow distributors for edge-
8
9
10 cooling portable stacks;
11
12 (v) the developed computational fluid dynamic model can also be used to study the heat
13
14
15
transfer between the stack and metal hydride tank. Although the heat generated by the stack
16
17 was not sufficient to release the required mass flow rate of hydrogen for stack operation at 1
18
19 kW power, which was also experimentally demonstrated using similar power air-cooled stack
20
21
22 (Nexa 1.2 kW) where 3 metal hydride tanks were used and in that case it was possible to
23
24 release the required amount of hydrogen from the metal hydride tanks.
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62 51
63
64
65
7. ACKNOWLEDGEMENTS
1
2
3 The authors of this work would like to acknowledge the support received from the project
4
5 STIM – REI, contract number: KK.01.1.1.01.0003, funded by the European Union through
6
7
8 the European Regional Development Fund – the Operational Programme Competitiveness
9
10 and Cohesion 2014-2020 (KK.01.1.1.01.).
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62 52
63
64
65
8. NOMENCLATURE
1
2
3 – heat exchange area, m2
4
5 – orthogonal area of the flow field, m2
6
7
8 – matrix for inertial resistance, m-2
9
10 – specific heat of air, J kg-1 s-1
11
12
14
15
16
– thermal energy, J
17
18 – Faradays constant, C mol-1
19
20
21
- average heat transfer coefficient, W m-2 K-1
22
23 – current density, A m-2
24
25
26
– logarithmic mean temperature difference between the stack and cooling air, K
27
28 – cooling air mass flow rate, kg s-1
29
30
– molar mass of specie , g mol-1
31
32
33 – pressure, Pa
34
35 – Prandtl number, /
36
37
38 – heat removed by the cooling air, W
39
40
– heat flux in Cartesian coordinate directions, W m-1
41
42
43 – Reynold’s number, /
44
45 – mass source term for specie , kg m-3 s-1
46
47
48 – time, s
49
50 – temperature of cooling air at stack outlet, K
51
52
53 – temperature of cooling air at stack inlet, K
54
55
56
– velocity in x-direction, m s-1
57
58 – velocity in y-direction, m s-1
59
60
61
62 53
63
64
65
– velocity in Cartesian coordinate direction, m s-1
1
2 – total volume of numerical flow domain, m3
3
4
5 – velocity in z-direction, m s-1
6
7
8
– porosity
9
10 – density, kg m-3
11
12
– stoichiometry cathode, /
13
14
15 – dynamic viscosity, kg m-1 s-1
16
17 – tangential stresses in Cartesian coordinate directions, Pa
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62 54
63
64
65
9. REFERENCES
1
2
3 [1] Barbir F. PEM Fuel Cells: Theory and Practice. Second Edition. Elsevier, Waltham 2012.
4
5 [2] Hoogers G. Fuel Cells Technology Handbook. First Edition. CRC Press LLC, Boca Raton
6
7
8 2002.
9
10 [3] Zhang J, Xie Z, Zhang J, Tang Y, Song C, Navessin T, Shi Z. Song D, Wang W,
11
12
Wilkinson DP, Liu ZS, Holdcroft S. High temperature PEM fuel cells. Journal of Power
13
14
15 Sources 2016;160(2):872-891.
16
17 [4] Fluckiger R, Tiefenauer A, Ruge M, Aebi C, Wokaun A, Buchi FN. Thermal analysis and
18
19
20 optimization of a portable edge-air-cooled PEFC stack. Journal of Power Sources
21
22 2007;172(1):324-333.
23
24
25 [5] Andreasen SJ, Ashworth L, Remn IN, Rasmussen PL, Nielsen MP. Modeling and
26
27 implementation of a 1 kW, air cooled HTPEM fuel cell in a hybrid electrical vehicle. ECS
28
29
30
Transactions 2008;12(1):639-650.
31
32 [6] Burke KA. Advanced fuel cell system thermal management for NASA exploration
33
34 missions. AIAA 6th international energy conversion engineering conference. Denver CO,
35
36
37 United States: American Institute of Aeronautics and Astronautics Inc; August 2-5. 2008;
38
39 Downloaded from:
40
41
42 https://ntrs.nasa.gov/archive/nasa/casi.ntrs.nasa.gov/20090027863.pdf ; [accessed 5 October
43
44 2019]
45
46
47 [7] Burke KA, Jakupca I, Colozza A. Development of passive fuel cell thermal management
48
49 technology. AIAA 7th international energy conversion engineering conference. Denver CO,
50
51
United States: American Institute of Aeronautics and Astronautics Inc; August 2-5. 2009;
52
53
54 Downloaded from:
55
57
58
59 2019].
60
61
62 55
63
64
65
[8] Burke KA, Jakupca I, Colloza A. Development of passive fuel cell thermal management
1
2 heat exchanger; Downloaded from:
3
4
6
7 2019]
8
9
10 [9] Wen CY, Huang GW. Application of a thermally conductive pyrolitic graphite sheet to
11
12 thermal management of a PEM fuel cell. Journal of Power Sources 2008;178(1):132-140.
13
14
15
16
17 [10] Wen CY, Lin YS, Lu CH. Performance of a proton exchange membrane fuel cell stack
18
19 with thermally conductive pyrolitic graphite sheets for thermal management. Journal of Power
20
21
22 Sources 2009;189:1100-1105.
23
24 [11] Wen CY, Lin YS, Lu CH, Luo TW. Thermal management of a proton exchange
25
26
27 membrane fuel cell stack with pyrolitic graphite sheets and fans combined. International
28
29 Journal of Hydrogen Energy 2011:36(10):6082-6089.
30
31
32
[12] Scholta J, Zhang W, Jorissen L, Lehnert W. Conceptual design for and externally cooled
33
34 HT-PEMFC stack. ECS Transactions 2008;12(1):113-118.
35
36 [13] Scholta J, Messerschmidt M, Jorissen L, Hartnig C. Externally cooled high temperature
37
38
39 polymer electrolyte membrane fuel cell stack. Journal of Power Sources 2009;190:83-85.
40
41 [14] Li X, Sabir I. Review of bipolar plates in PEM fuel cells: flow-field designs.
42
43
44 International Journal of Hydrogen Energy 2005;30(4):359-371.
45
46 [15] Chow CY, Wozniczka B, Chan JKK. Integrated reactant and coolant fluid flow field
47
48
49 layer for and electrochemical fuel cell. US Patent 5804326A; 1998.
50
51 [16] Ernst WD, Mittleman G. PEM-type fuel cell assembly having multiple parallel fuel cell
52
53
54
sub-stacks employing shared fluid plate assemblies and shared membrane electrode
55
56 assemblies. US Patent 5945232A; 1999.
57
58
59
60
61
62 56
63
64
65
[17] Wu HW. A review of recent development: Transport and performance modeling of PEM
1
2 fuel cells. Applied Energy 2016;165:81-106.
3
4
5 [18] Zhang X, Luo M, Dai W, Yao C, Wang J, Huand D, Wang C. Automotive fuel cell
6
7 engine test cell design and its thermal flow analysis. International Journal of Hydrogen
8
9
10 Energy 2018;43:17409-17419.
11
12 [19] Luo Y, Jiao K. Cold start of proton exchange membrane fuel cells. Progress in Energy
13
14
15
and Combustion Science 2018;64:29-61.
16
17 [20] Headley AJ, Chen D. Critical control volume sizing for improved transient thermal
18
19 modeling of PEM fuel cells. International Journal of Hydrogen Energy 2015;40:7762-7768.
20
21
22 [21] Meidanshahi V, Karimi G. Dynamic modeling, optimization and control of power
23
24 density in a PEM fuel cell. Applied Energy 2012;93:98-105.
25
26
27 [22] Jian Q, Huang B, Luo L, Zhao J, Cao S, Huang Z. Experimental investigation of the
28
29 thermal response of open-cathode proton exchange membrane fuel cell stack. International
30
31
32
Journal of Hydrogen Energy 2018;43:13489-13500.
33
34 [23] Chang Y, Qin Y, Yin Y, Zhang J, Li X. Humidification strategy for polymer electrolyte
35
36 membrane fuel cells – A review. Applied Energy 2018;230:643-662.
37
38
39 [24] Wang J. Theory and practice of flow field designs for fuel cell scaling-up: A critical
40
41 review. Applied Energy 2015:157:640-663.
42
43
44 [25] Liu Z, Chen J, Chen S, Huang L, Shao Z. Modeling and control of cathode air humidity
45
46 for PEM fuel cell systems. IFAC-PapersOnLine 2017;50-1:4751-4756.
47
48
49 [26] Nolan J, Kolodziej J. Modeling of an automotive fuel cell thermal system. Journal of
50
51 Power Sources 2010;195:4743-4752.
52
53
54
[27] Zhang G, Jiao K. Multi-phase models for water and thermal management of proton
55
56 exchange membrane fuel cell: A review. Journal of Power Sources 2018;391:120-133.
57
58
59
60
61
62 57
63
64
65
[28] Wu H, Berg P, Li X. Non-isothermal transient modeling of water transport in PEM fuel
1
2 cells. Journal of Power Sources 2007;165:232-243.
3
4
5 [29] Meng H. Numerical investigation of transient responses of a PEM fuel cell using a two-
6
7 phase non-isothermal mixed-domain model. Journal of Power Sources 2007;171:738-746.
8
9
10 [30] Kim DK, Min HE, Kong IM, Lee MK, Lee CH, Kim MS, Song HH. Parametric study on
11
12 interaction of blower and back pressure control valve for a 80-kW class PEM fuel cell
13
14
15
vehicle. International Journal of Hydrogen Energy 2016;41:17595-17615.
16
17 [31] Sasmito AP, Shamim T, Mujumdar AS. Passive thermal management for PEM fuel cell
18
19 stack under cold weather condition using phase change materials (PCM). Applied Thermal
20
21
22 Engineering 2013;58:615-625.
23
24 [32] Abdin Z, Webb CJ, Gray EM. PEM fuel cell model and simulation in Matlab-Simulink
25
26
27 based on physical parameters. Energy 2016;116:1131-1144.
28
29 [33] Didierjean S, Lottin O, Maranzana G, Geneston T. PEM fuel cell voltage transient
30
31
32
response to a thermal perturbation. Electrochimica Acta 2008;53:7313-7320.
33
34 [34] Kurnia JC, Sasmito AP, Shamim T. Performance evaluation of a PEM fuel cell stack
35
36 with variable inlet flows under simulated driving cycle conditions. Applied Energy
37
38
39 2017;206:751-764.
40
41 [35] Yan WM, Li HY, Weng WC. Transient mass transport and cell performance of a PEM
42
43
44 fuel cell. International Journal of Heat and Mass Transfer 2017;107:646-656.
45
46 [36] Tetuko AP, Shabani B, Omrani R, Paul B, Andrews J. Study of a thermal bridging
47
48
49 approach using heat pipes for simultaneous fuel cell cooling and metal hydride hydrogen
50
51 discharge rate enhancement. Journal of Power Sources 2018;397:177-188.
52
53
54
[37] Amphlett JC, Mann RF, Peppley BA, Roberge PR, Rodrigues A. A model predicting
55
56 transient responses of proton exchange membrane fuel cells. Journal of Power Sources
57
58 1996;61:183-188.
59
60
61
62 58
63
64
65
[38] Blunier B, Miraoui A. Optimization and air supply management of a polymer electrolyte
1
2 fuel cell. Vehicle Power and Propulsion IEEE Conference 2005.
3
4
5 [39] Purmann M, Styczynski Z. Simplified evaluation of PEM fuel cells by reduction of
6
7 measurement parameters and using optimized measurement algorithms. Journal of Power
8
9
10 Sources 2005;145:399-406.
11
12 [40] Pathpati PR, Xue X, Tang J. A new dynamic model for predicting transient phenomena
13
14
15
in a PEM fuel cell system. Renewable Energy 2005;30:1-22.
16
17 [41] Zhu H, Kee RJ. A general mathematical model for analyzing the performance of fuel cell
18
19 membrane-electrode assemblies. Journal of Power Sources 2003;142:61-74.
20
21
22 [42] Jiang Z, Dougal RA, Liu S, Gadre SA, Ebner AD , Ritter JA. Simulation of a thermally
23
24 coupled metal hydride hydrogen storage and fuel cell system. Journal of Power Sources
25
26
27 2005;142:92-102.
28
29 [43] Graf C, Vath A, Nicoloso N. Modeling of the heat transfer in a portable PEFC system
30
31
32
within MATLAB-Simulink. Journal of Power Sources 2006;155:52-59.
33
34 [44] Adzakpa KP, Ramousse J, Dube J, Akremi H, Agbossou K, Dostie M, Poulin A,
35
36 Fournier M. Transient air cooling thermal modeling of PEM fuel cell. Journal of Power
37
38
39 Sources 2008;179:164-176.
40
41 [45] Tolj I, Bezmalinović D, Barbir F. Maintaining desired level of relative humidity
42
43
44 throughout a fuel cell with spatially variable heat removal rates. International Journal of
45
46 Hydrogen Energy 2011;20:13105-13113.
47
48
49 [46] Penga Ž, Tolj I, Barbir F. Computational fluid dynamics study of PEM fuel cell
50
51 performance for isothermal and non-uniform temperature boundary conditions. International
52
53
54
55
56
57
58
59
60
61
62 59
63
64
65
[47] Penga Ž, Radica G, Barbir F, Nižetić S. Coolant induced variable temperature flow field
1
2 for improved performance of proton exchange membrane fuel cells. International Journal of
3
4
5 Hydrogen Energy 2019;20:10102-10119.
6
7 [48] Penga Ž, Bergbreiter C, Barbir F, Scholta J. Numerical and experimental analysis of
8
9
10 liquid water distribution in PEM fuel cells. Energy Conversion and Management
11
12 2019;189:167-183.
13
14
15
[49] Chow O, Friedman J, Halter D, Barbir F. Evaluation of a novel design air-cooled 1 kW
16
17 PEM-based fuel cell stack. Int Conf on Fuel Cell Development and Deployment, Storrs, CT,
18
19 March 7-10, 2004.
20
21
22 [50] Barbir F, Byron RH, Stone M, Bala-subramanian B. Electrochamical cell stack design.
23
24 US Patent 20060199056; 2006.
25
26
27 [51] Barbir F, Byron RH, Stone M, Bala-subramanian B, Compression Devices and
28
29 Electrochemical Cell Stack Design. US Patent 20060199067; 2006
30
31
32
33
35
36 1996;61:183-188.
37
38
39 [53] Mahjoubi C, Olivier JC, Skander-mustapha S, Machmoum M, Slama-belkhodja I. An
40
41 improved thermal control of open cathode proton exchange membrane fuel cell. International
42
43
44 Journal of Hydrogen Energy 2019;44(22):11332-11345.
45
46 [54] Mahjoubi C, Olivier JC, Skander-mustapha S, Slama-belkhodja I, Machmoum M. An
47
48
49 improved open cathode proton exchange membrane fuel cell control ensuring a simultaneous
50
51 regulation of temperature and air flow. The 9th International Renewable Energy Congress
52
53
54
(IREC 2018).
55
56
57
58
59
60
61
62 60
63
64
65
[55] Banerjee R, Kandlikar SG. Two-phase flow and thermal transients in proton exchange
1
2 membrane fuel cells – A critical review. International Journal of Hydrogen Energy
3
4
5 2015;40:3990-4010.
6
7 [56] Luo L, Jian Q, Huang B, Huang Z, Zhao J, Cao S. Experimental study on temperature
8
9
10 characteristics of an air-cooled proton exchange membrane fuel cell stack. Renewable Energy
11
12 2019;143:1067-1078.
13
14
15
[57] Burheim O, Pharoah JG. A review of the curious case of heat transport in polymer
16
17 electrolyte fuel cells and the need for more characterization. Current Opinion in
18
19 Electrochemistry 2017;5(1):36-42.
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62 61
63
64
65
*Highlights (for review)
Highlights:
- Lumped model and real-time transient CFD model are compared during a load profile
- Influence of bipolar plate material on temperature is determined
- Influence of operating delta pressure on relative humidity inside the stack
- Significant influence of cooling fin re-design on heat transfer is outlined
- Heat transfer between the stack and metal hydride tank is analyzed
*Revised Manuscript with No Changes Marked
Click here to download Revised Manuscript with No Changes Marked: Thermal Management of Edge-Cooled_R1 no marked c
1
Thermal Management of Edge-Cooled 1 kW
2
3
4
5 Portable Proton Exchange Membrane Fuel Cell
6
7
8
9 Stack
10
11
12 Ivan Tolj1*, Željko Penga1, Damir Vukičević1, Frano Barbir1
13
14 1
15 Faculty of Electrical Engineering, Mechanical Engineering and Naval Architecture
16
17 University of Split, R. Boškovića 32, 21000 Split, Croatia
18
19
20
21 Abstract
22
23
24 Comprehensive numerical analyses are conducted to study the influence of thermal
25
26 management on performance of 1 kW edge-cooled proton exchange membrane fuel cell stack
27
28
29 without external humidification. The experimental stack and numerical three-dimensional
30
31 computational fluid dynamics model are characterized by several novelty aspects. Two
32
33
34 numerical approaches are considered and compared for a prescribed load profile: (i) lumped
35
36 model and novel (ii) real-time transient computational fluid dynamics model incorporating
37
38
realistic modeling of forced air convection on the edge-cooling of the stack. The novelty of
39
40
41 the developed computational fluid dynamics model is the capability to give insight in the
42
43 transient results in only a fraction of time vs. experimental testing (40 mins vs. 4 hours) and
44
45
46 other computational fluid dynamics models of fuel cells which are only capable of steady-
47
48 state analysis. The developed computational fluid dynamics model is used to study the
49
50
51 influence of (i) bipolar plate materials (ii) operating delta pressure along the flow field and
52
53 (iii) different cooling fin configurations on the water and heat balance inside the stack. The
54
55
56
results indicate that (i) maximal and average temperatures of the stack are almost linearly
57
58
59 *
Corresponding author. Tel.: +385 21 305 948; fax: +385 21 305 776.
60
E-mail address: itolj@fesb.hr (I. Tolj).
61
62 1
63
64
65
correlated to the thermal conductivity of bipolar plate materials and maximal temperatures
1
2 can be significantly higher (ii) the operating delta pressure can be manipulated to increase the
3
4
5 performance of the stack and (iii) the cooling fin redesign has major influence on the overall
6
7 temperature uniformity across the stack. Additionally, the heat transfer between the stack and
8
9
10 metal hydride tank is studied.
11
12
13
14
15
Keywords:
16
17 PEM fuel cell;
18
19 Portable stack;
20
21
22 Edge-cooling;
23
24 Water and heat management;
25
26
27 Transient analysis.
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62 2
63
64
65
1. INTRODUCTION
1
2
3 Air-cooled proton exchange membrane (PEM) fuel cell stacks are gaining momentum as
4
5 power sources for portable applications such as electric bikes, unmanned aerial vehicles
6
7
8 (UAVs), forklifts, etc. Usually the power of portable stacks (PS) is up to several kW [1,2].
9
10 The majority of the heat generated during PS operation is dissipated into the surrounding
11
12
atmosphere via forced air convection along the cooling channels or edge-cooling fins, while
13
14
15 the higher power stacks require liquid cooling [3]. The emphasis during the design phase of
16
17 PS is primarily directed towards determining the most suitable material for bipolar plates and
18
19
20 the accompanying thermal management. Some of the thermal management objectives are: (i)
21
22 ensuring the appropriate heat transfer from the stack during startup, i.e. accurately
23
24
25 determining the required convection flow rate, to achieve sufficiently high operating
26
27 temperature from startup at ambient temperature, thus preventing flooding (occurring if the
28
29
30
flow rate is high and temperature of the stack is low) or membrane dehydration (occurring if
31
32 the flow rate is low and temperature inside the stack is high); (ii) ensuring sufficiently high
33
34 heat removal rates via forced air convection during the operation characterized by stable
35
36
37 temperature of the stack, i.e. loosely termed steady-state operation; (iii) ensuring as low as
38
39 possible temperature difference between the inside and outside portions of the stack by
40
41
42 choosing the materials with favorable thermal properties; (iv) properly design the stack and
43
44 edge-cooling fins to ensure sufficiently high heat removal rates; (v) design the frame for the
45
46
47 stack which will ensure that the cooling air is directed to uniformly flow along the edge-
48
49 cooling fins.
50
51
Due to the requirement for simple system design, portable stacks are commonly operated
52
53
54 without external humidification. Using dry hydrogen and ambient air results in poor
55
56 performance of the stack and exacerbated degradation unless special care is attenuated to the
57
58
59 water and heat management. As previously mentioned, the PSs are air-cooled, either via
60
61
62 3
63
64
65
cooling channels between the neighboring bipolar plates, or via specially designed fins
1
2 protruding from the longer sides of the stack, i.e. edge-cooling, as seen in [4]. Cooling
3
4
5 channels between the neighboring bipolar plates are similar in design to the coolant channels
6
7 for higher power stacks where liquid coolant is used. Cooling channels between the bipolar
8
9
10 plates result in a uniform temperature profile along the cell, but significantly increase the size
11
12 of the stack due to the requirement for low pressure potential required to achieve the desired
13
14
15
flow through the channels. Edge-cooling results in a more compact stack design since there
16
17 are no cooling channels between the neighboring bipolar plates. However, edge-cooling
18
19 requires sufficiently high thermal conductivity and special attention during the design of the
20
21
22 cooling fins to achieve uniform temperature distribution inside and along the stack.
23
24 Low power air-cooled high temperature PEM fuel cell stacks have been implemented in
25
26
27 hybrid electrical vehicles and coupled with lead acid batteries and are proven to be feasible
28
29 for such applications [5], therefore they are the focus of this work. However, the issue
30
31
32
remains how to cool low temperature PEM fuel cell efficiently under such operation due to
33
34 lower temperature gradient between the stack surface and the surrounding atmosphere.
35
36 Considerable effort is dedicated to developing passive cooling techniques for PEM fuel cells
37
38
39 used in space exploration missions [6], where the conclusion is that there is a requirement for
40
41 using costly materials with thermal conductivity in excess of 1000 W m-1 K-1 to achieve low
42
43
44 temperature gradient (lower than 3 °C) along the cell [7], or heat pipes for heat removal in
45
46 larger stacks [8], due to volume restrictions of the stacks used for such applications. In the
47
48
49 referenced studies feasible solution was achieved by using pyrolitic graphite combined with
50
51 anodized aluminum heat exchanger. In this work the focus of investigation is to see if lower
52
53
54
temperature gradient along the stack cross-section can be achieved using conventional
55
56 materials for bipolar plates or by simply changing the design of the edge-cooling fins,
57
58 therefore avoiding the necessity for costly materials. Another study [9] also used pyrolitic
59
60
61
62 4
63
64
65
graphite sheets for heat removal on a single-cell and demonstrated that the approach is
1
2 economically viable and useful for PEM fuel cell applications. In [10] it is demonstrated that
3
4
5 the PEM fuel cell stack efficiency can be increased by 15% while the flooding issues at low
6
7 cathode flow rates can be alleviated using this approach. Since thermal conductivity has such
8
9
10 significant influence of PEM fuel cell performance, this work is also focused on determining
11
12 the correlation between the bipolar plate material and maximal temperature inside the stack
13
14
15
during transient operation. The enhancement of the concept was then achieved by introducing
16
17 small fans into the stack system [11] thus improving the convection rates and providing
18
19 economically-friendly solution (when compared to coolant loop system) to the thermal
20
21
22 management issues of small to medium sized PEM fuel cell stacks. External cooling was also
23
24 implemented in study [12] on high temperature PEM fuel cell where the heat transfer was
25
26
27 analyzed using CFD and it was demonstrated that it is possible to achieve temperature
28
29 gradient below 15 °C via external cooling with air or thermo transfer liquid and to keep the
30
31
32
temperature gradient along the cell below 15 °C at all times [13]. Another important aspect
33
34 related to the thermal management and application of external cooling techniques is the
35
36 proper bipolar plate material selection and flow field design [14] which can significantly
37
38
39 increase or decrease the performance of the cell. Further improvements of the edge-cooling
40
41 technique can also be achieved by designing flow fields with reactant gas flow field and
42
43
44 coolant flow fields on the same plate surface, thereby decreasing the coolant mass flow rate
45
46 requirements and decreasing the temperature gradient inside the cell and also decreasing the
47
48
49 heat conduction length in stacks [15,16].
50
51 Different numerical methods are commonly applied for thermal and mass transfer analysis of
52
53
54
PEM fuel cells. Thermal analysis of PEM fuel cells is thoroughly described in [17] where the
55
56 emphasis is dedicated to CFD modeling due to capabilities for multi-dimensional and multi-
57
58 physics analysis. In the references work it is noted that most works related to CFD modeling
59
60
61
62 5
63
64
65
do not include the mesh sensitivity and time step sensitivity analyses while they provide data
1
2 for only specific operating conditions. However, due to computational complexity, the CFD
3
4
5 modeling technique is usually implemented for steady-state in-depth analysis of PEM fuel cell
6
7 operation. For automotive applications and the correlating thermal analysis other methods are
8
9
10 employed such as Monte-Carlo simulations [18], but this type of analysis does not give
11
12 insight in the temperature distribution and operating conditions inside the cell. Another
13
14
15
interesting topic for PEM fuel cell stacks is cold start, which is recently experimentally
16
17 investigated [19], and it is also considered in this work (although from ambient temperature
18
19 and not from subzero temperatures as in [19]). Time-efficient method for thermal analysis of
20
21
22 PEM fuel cells is control volume method [20], which is an upgraded lumped model
23
24 methodology. Although useful for control strategy development, this methodology also does
25
26
27 not give insight in the localized temperature distribution inside the cell but only approximates
28
29 the cell as a more detailed lumped model. Dynamic 1D modeling of PEM fuel cell operation
30
31
32
is shown in [21] where the conservation laws and non-linear terms are included and the model
33
34 is used to study the effectiveness of different control strategies, giving indications of the
35
36 requirement for adding the physical conservation laws for developing accurate and sensible
37
38
39 numerical model. The dynamic performance of a 500 W open-cathode stack [22] gives
40
41 overview of the temperature distributions between the cells in the stack, cooled via forced air
42
43
44 convection, and characteristic shapes of the temperature profiles during real time testing. In
45
46 [23] an overview is given of various methods for external and internal humidification of the
47
48
49 cell by adopting different strategies. One of the strategies is to manipulate the temperature or
50
51 the pressure distribution inside the flow field. Pressure distribution inside the flow field
52
53
54
affects relative humidity to a significant extent and is beneficial to some extent for liquid
55
56 water removal, and it can be potentially used to achieve the desired relative humidity
57
58 distribution inside the cell without the requirement for external humidification. The effect of
59
60
61
62 6
63
64
65
pressure drop on cell performance is discussed in more details in [24] and there are
1
2 indications that high pressure drop along the cell results in poor performance, high cost,
3
4
5 uneven concentration gradients between inlet and outlet, etc. Since it is known that the
6
7 operating pressure has significant influence on the water vapor saturation profile, and there
8
9
10 are indications that manipulating the operating pressure can increase the performance of the
11
12 cell [25], this will be studied in this work by operating the cell at higher operating pressure
13
14
15
with low pressure drop, i.e. operation similar to installing a pressure valve behind the cell, to
16
17 see if significant gains in efficiency can be achieved using this novel method, by comparing
18
19 the relative humidity profiles vs. common operation with zero gauge pressure.
20
21
22 In [26] lumped model for stack thermal management is developed and compared with
23
24 experimental data and demonstrated that high order non-linear lumped model shows good
25
26
27 agreement with experimentally obtained data for the average temperature and also shows that
28
29 third order non-linear model also shows good transient agreement. However, the maximal
30
31
32
temperatures inside the stack are unknown and therefore the results have limited practical
33
34 applications where materials with low thermal conductivity are used as bipolar plates. In
35
36 study [27] an overview is given of different multi-phase modeling approaches used for PEM
37
38
39 fuel cells via 3D CFD analysis and it is outlined that the future studies should be conducted
40
41 with more realistic outside boundary conditions, i.e. convection, vs. isothermal approach and
42
43
44 also outlines that currently developed detailed 3D CFD models show limited practicality due
45
46 to high complexity of the calculations even for steady-state simulations, indicating the
47
48
49 requirement for developing more simple but computationally less intensive models for
50
51 transient analysis in real-time. Dynamic model of PEM fuel cell is developed using CFD
52
53
54
methodology in [28,29] and shows good agreement with experimental polarization curves and
55
56 gives reasonable response to transient changes in operating electric potential. However, the
57
58 models in the referenced works are two-dimensional and as such not suitable for giving
59
60
61
62 7
63
64
65
insights in the detailed heat and mass transfer processes inside and along the cell during
1
2 transient operation.
3
4
5 PEM fuel cell stacks are often modeled using simplified but fast response models, as seen in
6
7 [30] where a system level numerical model is developed and the influence of the blower
8
9
10 pressure on a 80 kW PEM fuel cell stack was investigated. Simplified CFD model of a PEM
11
12 fuel cell stack is shown in [31] where the entire stack is modeled as a 2D structure and
13
14
15
thermal analysis was conducted to determine if the application of phase-change materials and
16
17 insulator coating can alleviate the freeze-thaw cycles in cold surrounding atmospheres.
18
19 Although the model is considerably simplified when compared to fully resolved 3D CFD
20
21
22 model of PEM fuel cell, the analysis results give significant amount of information about the
23
24 thermal management of the stack, superior to using lumped model approach. Another
25
26
27 example of simplified modeling with good prediction of PEM fuel cell stack performance is
28
29 described in [32] where MATLAB-Simulink environment is used to model the polarization
30
31
32
curves under different operating conditions with good agreement vs. experimentally obtained
33
34 data. Also, a simple 1D transient model of PEM fuel cell shown in [33] indicates the
35
36 importance of temperature control of the cell on the overall voltage response due to the effect
37
38
39 of temperature on the water vapor saturation profile, i.e. relative humidity and thus the
40
41 membrane water content. Simplified CFD model of a stack is shown in [34] where the stack is
42
43
44 represented by a symmetrical half of a single cell with periodic boundary conditions and
45
46 subjected to dynamic automotive load profiles for a 320 cells with 1600 cm2 active catalyst
47
48
49 area and notion is given that small changes in the mass flow rates of the reactants and coolant
50
51 have a considerable effect on the thermal envelope of the stack during transient operation,
52
53
54
albeit their effect on the net power stack is minor, indicating that it is required to study the
55
56 whole cross-section of the stack during prolonged transient operation. A more detailed
57
58 transient CFD model is shown in [35] where a simplified single-cell model with different
59
60
61
62 8
63
64
65
channel configurations is used to study the dynamic performance of the cell with emphasis on
1
2 voltage undershoots/overshoots. Since PS can be coupled with metal hydride tank, the heat
3
4
5 generated by the cell can be used to release the stored hydrogen from the metal hydride tank
6
7 during operation, as seen in study [36].
8
9
10 Below are some of the relevant lumped model works which were used as inspiration for
11
12 building the lumped model in this work. In [37] developed a thermal model for a Ballard
13
14
15
Mark V 35-cell 5 kW PEM fuel cell stack by performing mass and energy balances on the
16
17 stack. Initially, they developed a steady-state overall dynamic model by coupling an
18
19 electrochemical model with thermal model. The model was further transformed into a
20
21
22 transient model that predicted fuel cell performance in terms of cell voltage output and heat
23
24 losses as a function of time due to various changes imposed on the system. Similar approach
25
26
27 is used in this work via CFD analysis, which gives more detailed insight in the mass and heat
28
29 transfer inside the stack as well as on the outside of the stack due to including forced air
30
31
32
convection. A mathematical model of PEM fuel cell, including air-compression and
33
34 humidification process is shown in [38]. The study focused on air-supply management, with
35
36 the objective of optimizing the inlet air pressure and stoichiometry provided by the
37
38
39 compression system. The optimization results showed that working with fully humidified air
40
41 at the inlet is not always the best solution, especially for low air flow rates, since under such
42
43
44 circumstances the water accumulation inside the cell is exacerbated. In this work therefore the
45
46 focus is to show that it is possible to achieve high relative humidity of the reactants without
47
48
49 the necessity for external humidification by choosing the bipolar plate materials and designing
50
51 the edge-cooling fin geometry, thereby ensuring high operational efficiency with minimized
52
53
54
occurrence of flooding. A simplified approach for evaluation and modeling PEM fuel cell
55
56 stacks in a stationary operation is shown in [39], using a model based on the reduction of
57
58 measured stack parameters via regression approaches. Because heat is lost at the stack
59
60
61
62 9
63
64
65
surface, pressure losses, stack outlet temperature and the characteristics of a fuel cell can be
1
2 simulated by regression approaches. The results of their study included optimized
3
4
5 measurement algorithms for fuel cell stacks to reduce the time required for their evaluation
6
7 due to the multitude of variable parameters. The approach that is used in this work is similar
8
9
10 in means that the stack is modeled as heat and mass transfer system, while the influence of the
11
12 electrochemical reactions occurring inside the stack and the accompanying heat release are
13
14
15
modeled using sink and source terms, enabling fast analysis of transient stack operation. The
16
17 approach shown in this work is different since it uses CFD analysis, while it is also different
18
19 from other works using CFD analysis because it results in computationally far less intensive
20
21
22 model when compared to full 3D CFD analyses including more advanced electrochemistry
23
24 modelling. Another study [40] developed a mathematical model to simulate transient
25
26
27 phenomena in PEM fuel cell system. The dynamic fuel cell model incorporated the effects of
28
29 charge double-layer capacitance, the dynamics of flow and pressure in the anode and cathode
30
31
32
channels and transient mass and heat transfer features inside the cell. The model predicted the
33
34 transient response of cell voltage, cell temperature, hydrogen and oxygen outflow rates and
35
36 cathode channel temperatures and pressures under sudden load changes. However, in that
37
38
39 study the effect of the heat transfer outside of the stack is not modelled, while in this work it
40
41 can be seen that the heat transfer between the stack and surrounding atmosphere has
42
43
44 significant influence on the overall temperature profile along and inside the stack. In [41]
45
46 mathematical model is developed to represent the membrane-electrode assembly (MEA) of
47
48
49 fuel-cell systems. The model was used to analyze the effects of various polarization
50
51 resistances on cell performance. Thermally coupled hydrogen storage and fuel cell system
52
53
54
was investigated in [42]. This work utilized a computational environment referred to as the
55
56 virtual test bed to simulate and compare the behavior of coupled and uncoupled systems. This
57
58 model used lumped approximations, and a pulsed hydrogen demand load was applied. The
59
60
61
62 10
63
64
65
results from these simulations clearly revealed the unique and subtle behavior of thermally
1
2 coupled systems. Since waste heat can be used to release the stored hydrogen inside metal
3
4
5 hydride tanks, commonly used in forklifts, this was also investigated in this work by coupling
6
7 the systems using CFD analysis and to determine will the amount of hydrogen released be
8
9
10 able to feed the stack during operation at 1 kW power. In another study [43] dynamic model
11
12 of a PS was developed in MATLAB Simulink environment, including heat transfer between
13
14
15
the stack and a metal hydride storage unit. The derived equations described both the steady-
16
17 state and dynamic operation of the PEM fuel cell system with sufficient accuracy. Different
18
19 operating conditions were simulated, specifically addressing high and low-temperature PEM
20
21
22 fuel cell operation. The model can be used to compare different operating conditions and to
23
24 estimate the interaction between the different components. In this work the approach is
25
26
27 upgraded by also giving insight in the temperature distribution inside the metal hydride tank.
28
29 Three-dimensional dynamic thermal model of a single cell was developed in [44] to study
30
31
32
temperature distribution in a fuel cell cooled by air flow from the bottom to the top. The
33
34 model was governed by the thermal energy balance, taking into account the inlet gas
35
36 humidity, and developed using a finite difference method implemented in the MATLAB
37
38
39 Simulink environment. The efficiency of the air cooling device revealed that the cell
40
41 temperature is directly linked to current density and gas humidity. Moreover, the temperature
42
43
44 variation in the stack was shown to be very high. The variability of air cooling between the
45
46 cells of the stack led to large temperature variations, up to 8 °C, from one cell to another. In
47
48
49 this work the focus is dedicated to finding solution for minimizing the temperature
50
51 inhomogeneity by determining the influence of bipolar plate material and edge-cooling fin
52
53
54
design on the overall temperature envelope.
55
56 Although lumped models are useful due to their fast response and low computational intensity
57
58 with reasonable accuracy when compared to experimental measurements of integral
59
60
61
62 11
63
64
65
parameters, spatial distribution of species transport and heat transfer, as well as critical
1
2 operating parameters, e.g. extreme values of temperature and relative humidity, are lost
3
4
5 during such approaches due to integration and lumping the values in single digit. Without
6
7 determining the minimal and maximal values of temperature inside the stack (and
8
9
10 consequently the relative humidity distribution since water vapor saturation profile is
11
12 temperature dependent), it is not possible to design the stack for optimal operation. Localized
13
14
15
hot spots can result in severe membrane dehydration and reduced lifetime of the stack, while
16
17 regions with lower temperature are prone to flooding, resulting in non-uniform reactant
18
19 distribution along the active area of the cell, non-uniform current density distribution, and
20
21
22 potential starvation of the cell in regions where significant amounts of water are present. In
23
24 previous studies [45,46] it was shown that the operating temperature inside PEM fuel cell is
25
26
27 highly non-uniform and it is not accurate to represent the operating temperature with just one
28
29 parameter, while at the same time if the temperature profile is controlled it can be exploited to
30
31
32
result in high performance of the cell by manipulating the water vapor saturation profile, i.e.
33
34 relative humidity profile inside the cell. In [47] short stack was developed and it was shown
35
36 that the temperature distribution along the stack height can be non-uniformly distributed if the
37
38
39 coolant mass flow rate is low, and that the temperatures between the neighboring bipolar
40
41 plates can be different, which is also not captured by the lumped model approach. Also,
42
43
44 during the dynamic PEM fuel cell operation, in [48] it can be seen that the processes inside
45
46 the cell related to the liquid water accumulation and transport can be rather sluggish and
47
48
49 influence the performance of the cell significantly, therefore transient analysis must be
50
51 conducted in conjunction with steady-state modeling using CFD analysis to give detailed
52
53
54
insight in the processes inside the cell. However, once the thermal management is resolved
55
56 using more detailed CFD analysis and the temperature gradients during the operation are
57
58 resolved, the lumped modeling approach can be used to give much faster response during
59
60
61
62 12
63
64
65
transient operation and for this reason it is best to use both approaches in conjunction for
1
2 thermal analysis of PEM fuel cell stacks during transient operation.
3
4
5 The work is structured as follows. The experimental stack is briefly described in [49] and [50]
6
7 including the novelty aspects of the design. Lumped steady-state and transient models are
8
9
10 shown, and transient model is used to study the temperature dependency on the operating
11
12 power during dynamic load profile and for three different shutdown strategies and also to
13
14
15
determine the amount of condensed water in the cathode exhaust. The novel CFD model is
16
17 first developed to determine if the stack manifold geometry has sufficiently high cross-section
18
19 area to achieve uniform flow of the reactants along the entire height of the stack and to
20
21
22 determine the required viscous resistance coefficient which is to be used on a single-cell level
23
24 to accurately result in the pressure drop along the flow field similar to the experimentally
25
26
27 obtained value. Once it was determine that the flow distribution between the neighboring cells
28
29 in the stack is uniform, single-cell model with symmetry boundary conditions is built and
30
31
32
used for detailed thermal analysis including modeling of relative humidity, reactant
33
34 consumption and water generation for the cathode side of the cell as well as the influence of
35
36 the geometry of the edge-cooling fins on the spatially resolved temperature distribution during
37
38
39 transient operation. The mesh and time-step sizes are thoroughly determined to give accurate
40
41 results and fast computational performance and the resulting CFD model has shown faster
42
43
44 performance than real-time testing. The results of the lumped model analysis are compared
45
46 with experimental data from other similar works and show good qualitative and quantitative
47
48
49 agreement.
50
51 The main contributions of this work to the research field are evident in outlining that the
52
53
54
55
56 performance due to inability to accurately calculate the maximal temperature inside the stack
57
58 and this work also outlines under which circumstances the lumped models can be used. This
59
60
61
62 13
63
64
65
1
2 which gives insight in mass and heat transfer both inside and outside of the cell (internal heat
3
4
5 and mass transfer and forced air convection edge-cooling), while other works focus on only
6
7 one aspect and simplify the other to a significant extent. The methodology for developing
8
9
10 such model using CFD is shown in detail in this work. The novelty of the work is also evident
11
12 in conducting the analysis which gives insight in maximal temperatures inside the PS for
13
14
15
different bipolar plate materials, how to enhance the performance by introducing a pressure
16
17 drop along the flow field or operating the cell at elevated pressure, and showing high
18
19 significance of external cooling fin configurations on the overall heat transfer (while the
20
21
22 results of the analyses show that the maximal temperatures inside the stack are very similar to
23
24 the ones observed during experimental investigation done by other research groups recently).
25
26
27 The novelty is also including the transient heat exchange with metal hydride tank, and the
28
29 results are qualitatively comparable with experimentally obtained results for a similar stack.
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62 14
63
64
65
2. EXPERIMENTAL STACK
1
2
3 The stack used in this study, Figure 1, was developed by Proton Energy Systems [49]. The
4
5 stack displays many innovative features including in-flow configuration, cooling and stack
6
7
8 clamping [49-51]. It consists of 60 cells, each with an active area of 65 cm2. The design is
9
10 rectangular with 10:1 aspect ratio to allow for efficient heat conduction of the edge of the
11
12
cells, where fins are extended for cooling.
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50 Figure 1. Above: Proton Energy System’s short stack exploded view (left) with annotated
51
52
53 parts and assembly (right); Middle: Proton Energy System’s 1-kW fuel-cell stack (left) and
54
55 transparent shroud assembly (right); Below: Cathode (left) and anode (right) flow fields with
56
57
58 annotated air-coolant flow direction along the edge-cooling fins.
59
60
61
62 15
63
64
65
The bipolar plates including the fins are fabricated from a molded polymer/graphite mixture
1
2 (made by SGL Carbon) with effective thermal conductivity of 20 W m -1 K-1. The stack is
3
4
5 cooled by air flow along the longer edge of the cells where the fins are located. Instead of tie
6
7 rods, metallic shrouds that snap on the end plates are used to compress the stack [51]. The
8
9
10 cells are compressed by five polyurethane springs distributed along the centerline of the active
11
12 area. The shrouds also form the passage for cooling air. The cooling fan is installed directly
13
14
15
onto the shrouds. The stainless-steel end plates also serve as the bus plates. The seven-layer
16
17 MEAs are manufactured by 3M and include a catalytic membrane and two carbon paper gas
18
19 diffusion layers with integrated gaskets. The active part of the stack therefore consists of only
20
21
22 two alternating components: 60 MEAs sandwiched between 61 collector plates. Besides these
23
24 repetitive components, the rest of the stack consists of only 13 additional parts (three end
25
26
27 plates, two shrouds, five polyurethane springs and four fittings). Stack properties and
28
29 operating conditions are outlined in Table 1.
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62 16
63
64
65
1
2 Number of cells 60, /
3 Active area 65 cm2
4
5 Aspect ratio (L:W) 10:1, /
6
7 Bounding dimensions (31.0 × 23.6 × 9.6) cm
8
9 Volume 7023.36 cm3
10 Mass 17 kg
11
12 Specific heat 0.59 kJ kg-1 K-1
13
14 Heat exchange area 0.9378 m2
15
16 Coolant Air
17 Nominal power 1 kW
18
19 Current 23 A
20
21 Electric potential 43.5 V
22
23 Anode stoichiometry 1.2
24 Cathode stoichiometry 2
25
26 Operating temperature 60 °C
27
28 Reactant’s inlet temperature 20 °C
29
30 Back-pressure 0 Pa gauge pressure
31 Pressure drop along the cathode side 20∙103 Pa
32
33 Reactant’s inlet relative humidity 0%
34
35 Cooling control On-Off
36
37 Table 1. Stack properties and operating conditions.
38
39 The stack is designed to generate 1 kW power output at relatively high single-cell operating
40
41
42 potential of ≥0.7 V cell-1. The cell operating current was limited due to poor thermal
43
44 conductivity of the bipolar plates, which would cause large temperature gradients between the
45
46
edge and center of the cell at lower operating potentials, i.e. higher currents, and consequently
47
48
49 result in very poor performance of the cell due to sever membrane dehydration. The stack was
50
51 tested at the Connecticut Global Fuel Cell Center, University of Connecticut [49]. The
52
53
54 resulting polarization curve is shown in Figure 2. The projected potential and projected power
55
56 and extrapolated by fitting a polynomial curve through the experimentally obtained points,
57
58
59 while the operating current was limited to 25 A to prevent overheating.
60
61
62 17
63
64
65
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18 Figure 2. Polarization curve vs. power curve.
19
20 The focus of this research is to determine the most feasible method for minimizing the
21
22
23 maximal temperatures occurring inside the PS, thus enabling operation at higher currents and
24
25 increasing the operating efficiency which will be a consequence of higher membrane water
26
27
content, i.e. higher relative humidity profile, along the entire active area of the cell.
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62 18
63
64
65
3. LUMPED MODEL
1
2
3 The objective of the development of the lumped model was to study the transient and steady-
4
5 state behavior of a fuel cell stack under variable loads. The mathematical model describes the
6
7
8 fuel cell voltage as a function of temperature and heat energy balance of the fuel cell stack.
9
10 The model also includes heat dissipation from the stack to surroundings and coolant energy
11
12
balance equations. A complete stack energy balance was considered, including the heat
13
14
15 generated by the inlet reactant gases (the enthalpy of inlet reactant gases), and the heat of
16
17 gases leaving the stack. The calculation includes both the latent and sensible heat of water
18
19
20 flows at the stack inlet and outlet. Generated heat is dissipated from the fuel cell stack to the
21
22 surroundings via radiation and natural convection (through the horizontal and vertical plates
23
24
25 of the stack). The amount of heat dissipation depends on the temperature of the stack, the
26
27 ambient temperature and the properties of the surrounding air. Dissipation is at a maximum
28
29
30
when the stack reaches (operates under) steady state. The model also includes an energy
31
32 balance equation for the coolant air. The temperature of the coolant varies with time and
33
34 distance along the active length of the fuel cell. The heat capacity of the fuel cell stack was
35
36
37 determined by calculating the heat capacities of each individual component of the fuel cell
38
39 such as graphite, aluminum and stainless steel, and summarizing them.
40
41
42 The assumptions regarding the conducted calculations are summarized as follows in the
43
44 described mathematical model:
45
46
47  The fuel cell stack operates at constant current vs. temperature, i.e. galvanostatic
48
49 approach
50
51
52  Edge cooling mechanism is used to remove heat from the stack
53
54  Air is used as the coolant
55
56
57  Flow is laminar and the temperature distribution is uniform, therefore Nusselt number
58
59 is 2.98
60
61
62 19
63
64
65
 Water back-diffusion is equal to electro-osmotic drag, i.e. there is zero net water
1
2 transport across the membrane
3
4
5  Reactants are completely dry at the inlet
6
7
8
3.1. Steady-state model
9
10 With known operating conditions and stack geometry, it is possible to calculate the heat
11
12
13 fluxes in and out of the stack at steady state. The resulting stack energy balance at full power
14
15 (60 °C) is shown in Table 2.
16
17
18
19 Electrical power 1 kW
20
21 Reactant’s enthalpy at the inlet 0.054 kW
22
23 Produced heat 1.044 kW
24
25
Reactant’s enthalpy at the outlet 0.408 kW
26 Heat dissipation 0.090 kW
27
28 Heat to be removed by cooling air 0.601 kW
29
30 Table 2. Stack energy balance.
31
32 Since the mean operating potential of each single-cell in the stack is 0.725 V, the stack
33
34
35 efficiency is close to 50 %, and the amount of generated heat is close to the amount of
36
37 generated electrical power. The heat is removed from the stack by the flow of excess air
38
39
40
carrying evaporated product water (408 – 54 = 354 W) and by radiation and convection to the
41
42 surrounding air (90 W). The remaining 600 W must be removed by the cooling air to avoid
43
44 overheating.
45
46
47 The heat removed by the cooling air is
48
49 (1)
50
51
52 where represents cooling air mass flow (kg s-1) rate, specific heat (J kg-1 K-1) of air,
53
54
55 outlet and inlet temperature (K) of air. The heat must be transferred from the
56
57 stack to the cooling air, i.e.
58
59
60
61
62 20
63
64
65
(2)
1
2
where represents the average heat transfer coefficient (W m-2 K-1), exchange area (m2)
3
4
5 and is the logarithmic mean temperature difference (K) between the stack and the
6
7 cooling air.
8
9
10 The stack releases heat during operation and the total amount of generated heat consists of
11
12 (3)
13
14
15 Where represents the total current (A) generated by the cell, while is overpotential (V),
16
17 i.e. difference between the theoretical potential for hydrogen lower heating value of 1.254 V,
18
19
20 and the electric potential of the cell, calculated as
21
22 (4)
23
24
25 Where is reversible Nernst potential (V), while is calculated as
26
27 (5)
28
29
30 Where , and represent the activation, ohmic and concentration overpotential
31
32 (V), and reversible Nernst potential is calculated as
33
34
35 (6)
36
37
38
39 where
40
41 (7)
42
43
44
45
46
47
48 And temperature (K), pressure (atm), change in entropy (J mol-1 K-1), number of
49
50 electrons transferred, Faraday’s constant (C mol-1) and universal gas constant (J mol-1 K-
51
52 1
53
), and subscripts , , and refer to the reference values, hydrogen, oxygen and
54
55 water, respectively. Since the experimental polarization curve is available, as seen previously
56
57
58
in Figure 2., in this model the is calculated simply by subtracting the cell operating
59
60 voltage from the theoretical potential for lower hydrogen heating value of 1.254 V.
61
62 21
63
64
65
The water activity , i.e. relative humidity RH, is calculated as
1
2 (8)
3
4
5
6 Where is partial pressure (Pa) of water vapor and is saturation pressure (Pa),
7
8
9 calculated as
10
11 (9)
12
13
14
15
16 For given stack and ambient conditions, the only variables in eqs. 1 and 2 are the coolant flow
17
18 rate, air outlet temperature and stack temperature. Figure 3 shows the cooling air and stack
19
20
21 temperatures as functions of the cooling air flow rate and the stack heat transfer coefficient
22
23 product ( ). During stack testing [49], it was observed that the air exit temperature did not
24
25
26 rise above 30 °C, due to the high air flow rates. The resulting stack temperature could not be
27
28 maintained at the desired value of 60 °C despite the high air flow rates. This indicates that the
29
30
31 heat transfer ( ) was insufficient (i.e. < 15). Doubling the heat exchange area, for
32
33 example, would allow not only operation at the desired stack temperature but would also
34
35
36
allow operation with a cooling air flow rate one fifth of the one used initially. Additionally,
37
38 the resulting temperature of the cooling air leaving the stack could be sufficient for extracting
39
40 hydrogen from a metal hydride storage unit.
41
42
43 100
44 hA = 15
90
45 hA = 20
46 80 hA = 30
47 70 Tair out
temperature, °C
48 60
49 Tstack
50
50
51 40
52 30
53 20
54
55 10
56 0
57 0 20 40 60 80 100 120
58 flow rate (mCp)
59
60
61
62 22
63
64
65
Figure 3. Stack and cooling air steady-state outlet temperatures as a function of cooling air
1
2 flow rate and overall stack heat transfer coefficient.
3
4
5 3.2. Transient model
6
7
8 A MATLAB model of the stack mass and heat fluxes was developed and used to model the
9
10 transient behavior, particularly at start-up and at stepwise power change. Operation of the
11
12
stack was simulated starting at room temperature (25°C) with nominal electrical power of 1
13
14
15 kW (current 23 A), reaching steady-state after ca. 5500 s, after which the current was abruptly
16
17 decreased to 11.5 A for duration of 1000 s and then increased back to 23 A until total time of
18
19
20 10000 s has passed. Since the stack temperature was initially at ambient conditions, the stack
21
22 did not reach the nominal power until the operating temperature of 60 °C was reached. The
23
24
25 coolant fan was not required for the first 500 s. When the stack reached 60 °C the cooling fan
26
27 was started. With the designated mass flow rate, the temperature of the stack during steady-
28
29
30
state reached and stabilized at 67 °C. Three scenarios are studied: (i) power and cooling fans
31
32 are turned off at 10000 s time step; (ii) current was reduced to 11.5 A at 10000 s time step and
33
34 the fans continued to run at mass flow rate of 0.1 kg s-1 and (iii) the current was reduced to
35
36
37 11.5 A and the cooling fans are turned off and on between 55 °C and 60 °C from 10000
38
39 seconds to 15000 seconds, resulting in saw-tooth temperature profile visible in Figure 4.
40
41
42 (right). In Figure 4. it can be noticed that for case (ii) another steady state is reached after
43
44 approx. 55 minutes. These results were achieved due to the large thermal mass of the stack, 17
45
46
47 kg at 0.59 kJ kg-1 K-1, as calculated from the masses and specific heats of the stack
48
49 components. This value is comparable to the previously reported specific heat value for an
50
51
early Ballard stack of 0.7 kJ kg-1 K-1 [52].
52
53
54
55
56
57
58
59
60
61
62 23
63
64
65
1000 70
1
900 65
2
3 800 60
4
Fuel Cell Temperature (C)

700 55
5
Electrical Power (W)
600 50
6
7 500 45
8 400 40
9 power off/ cooling fan off
10 300 35
0.1 kg/s
200 30
power off/ cooling fan off
12 100 25
0,1 kg/s
14 0 20
0 5000 10000 15000 0 5000 10000 15000
15 Time (sec) Time ( sec)
16
17
18 Figure 4. Resulting electrical power (left) and temperature (right) during startup from room
19
20 temperature, at current of 23 A), 11.5 A and at 0 A for 3 strategies after 10000 s time step.
21
22
Since the experimental data on temperature vs. load (i.e. current) profile is not available for
23
24 the PS shown in this work, the results of the lumped model can be compared with
25
26 experimental data conducted by Mahjoubi and Olivier [53,54] for a similar stack (1.4 kW
27
28 power, fan-cooled) during a current ramp and the consequent temperature ramp from 30 °C to
29
30 60 °C, as seen in Figure 5., which is very similar to the setup of the stack operating conditions
31
32 in this work. By comparing the results in Figure 5. with results in Figure 4. it can be can
33 observed that the temperature profile which is a consequence of current ramp-up is of quite
34
35 similar shape, as well as the resulting temperature profile during a sudden decrease in the
36
37 current Figure 5. (below), and also during the decrease of current from 20 A to 0 A, as seen in
38
39 Figure 5. (above).
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62 24
63
64
65
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27 Figure 5. Current and temperature history of air-cooled, open-cathode 1.4 kW PEM fuel cell
28
29 stack, images adopted from [53] (above) and [54] (below).
30
31
32
33
34 The resulting liquid water mass flow in the outlet air from the stack is shown in Figure 6 for
35
36
37 cases (ii) and (iii). It can be seen that during the stack startup and the sudden decrease in the
38
39 stack current from 23 A to 11.5 A as well as the operating temperature of the stack after 5500
40
41
42 s, i.e. whenever the stack temperature is below 60 °C, condensed water is present at the
43
44 cathode outlet. Prolonged holds at these operating conditions would result in severe flooding
45
46
of the cell. After 1000 s hold at 11.5 A, the current is increased to 23 A and the liquid water is
47
48
49 evaporated. After 10000 s time step, it can be seen that the liquid water depends on the stack
50
51 cooling strategy and that it can be controlled simply by adjusting the appropriate cooling air
52
53
54 mass flow rate. The results of the lumped model are compared vs. results of CFD analysis in
55
56 later chapters. The amount of generated water is calculated using Faraday’s equation and
57
58
59 numerical representation of Mollier’s h-x chart, Figure 6., it can be seen that water condenses
60
61
62 25
63
64
65
during the load ramp-down due to sudden temperature decrease and it is evaporated once the
1
2 load is ramped-up and the temperature is increased.
3
4
5
0.14
6 0,1 kg/s
0.12
8
9
10 0.1
11
H2OLoutAir (g/s)
12 0.08
13
14 0.06
15
16 0.04
17
18
0.02
19
20
21 0
0 5000 10000 15000
22 Time (sec)
23
24
25 Figure 6. Resulting flow of liquid water at the cathode outlet during start-up and operation at
26
27 full power at 23 A and at 11.5 A.
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62 26
63
64
65
4. COMPUTATIONAL FLUID DYNAMICS MODEL
1
2
3 A more advanced numerical model was built using CFD modeling. The whole stack was not
4
5 modeled due to high computational requirements for conducting such analysis and only a
6
7
8 limited benefit. The CFD model represents one slice of the stack incorporating one single cell
9
10 and the outside domain where the coolant air flows. The entire stack could be represented by
11
12
patterning this single cell, while the main difference vs. more realistic model would be seen
13
14
15 only near the end plates due to somewhat different heat transfer, but this would result in huge
16
17 increase in computational cost, while it would only result in minor amount of additional
18
19
20 information. However, as previously discussed in the Introduction chapter, the simplified
21
22 CFD models give reasonably accurate prediction of PEM fuel cell performance therefore this
23
24
25 approach is reasonable due to the requirement for conducting transient simulations. The
26
27 single-cell which represents a repetitive segment of a stack was modeled and several steps
28
29
30
were introduced to ensure that the scale-up is proper. It was also not convenient to model the
31
32 actual processes inside the cell in conjunction with the heat transfer between the cell and
33
34 coolant air due to transient nature of the calculations and the lack of experimental data for
35
36
37 localized validation of the model. Since the stack consists of 60 single cells, and they are
38
39 connected to common manifolds, therefore it is required to determine the influence of the cell
40
41
42 distance from the inlet on the overall mass flow rate in each particular cell. The decrease in
43
44 the pressure potential (a consequence of the viscous losses, which are a result of flow of the
45
46
47 reactants through the common manifolds), results in non-uniform distribution of the mass
48
49 flow rate vs. distance from the common inlet. This phenomenon is interesting to determine if
50
51
the cross-section of the designed manifolds is sufficiently large to prevent non-uniform
52
53
54 distribution of the reactants vs. stack height. Also, since the pressure drop across the cathode
55
56 side of the flow field is measured experimentally, it is possible to determine the viscous
57
58
59 resistance of the cathode flow field by using a default value in the analysis and iteratively
60
61
62 27
63
64
65
adjusting it based on the overall pressure experimentally measured pressure drop until the
1
2 values show good agreement, a parameter which will later be used for more detailed transient
3
4
5 analysis. The model for determining the mass flow rate distribution vs. stack height is shown
6
7 below as well as the resulting viscous resistance of the cathode side flow field.
8
9
10 4.1. Governing equations and boundary conditions
11
12
The developed numerical CFD models are based on solving the Navier-Stokes equations in
13
14
15 three-dimensional, unsteady-form (for convenience referred to as sink-source formulation) for
16
17 continuity:
18
19
20 (3)
21
22
23
24 represents density, time, , and velocities in , and , i.e. Cartesian coordinate
25
26 directions, respectively and mass source term.
27
28
30
31 (4)
32
33
34
35 where represents pressure, Reynolds number, and tangential stresses in the Cartesian
36
37
39
41
42
43 (5)
44
45
46
48
49 (6)
50
51
52
53 as well as energy equation
54
55
56
57
58
59
60
61
62 28
63
64
65
(7)
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15 where represents thermal energy, Prandtl number and heat flux in Cartesian
16
17
19
20 The sink and source terms for the mass transfer, i.e. more precisely continuity equation, are
21
22
23 defined as
24
25 (8)
26
27
28
29
30 where represents the cathode stoichiometry, current density, orthogonal area of the
31
32 flow field domain perpendicular to the stack height, without the inflow and outflow
33
34
35 manifolds, Faraday's constant, molecular mass of oxygen and the total volume of
36
37 the numerical flow field domain.
38
39
40 (9)
41
42
43
44
45 the pressure potential loss across the cathode side flow field is modeled by introducing a
46
47 momentum sink term according to the porous media formulation
48
49
50
(10)
51
52
53
54
55 where represents the momentum sink term in -th direction, and are the matrices for
56
57
58
the viscous and inertial resistance, respectively, absolute viscosity, directional velocity
59
60 and velocity magnitude. Since the flow velocities are relatively low, the inertial term on
61
62 29
63
64
65
the right is neglected, i.e. . The viscous resistance matrix is defined as orthotropic,
1
2
3
with maximal value (1020 m-2) in the direction of the stack height, i.e. -direction, to prevent
4
5 the flow and mixing of the gas between the adjacent cells in the stack. The values of the
6
7 viscous resistance in and directions are set as equal and determined based on the resulting
8
9
10 pressure potential loss across the flow field. The porosity is set to 0.6 by comparing the
11
12 volume of the reactant channels vs. volume of the flow field channels plus lands.
13
14
15 4.2. Stack manifold model
16
17
18 The objective of the stack manifold simulations was to determine if the cross-section area of
19
20 the common manifolds is sufficiently large to ensure similar mass flow rates between the
21
22
23
neighboring cells along the stack height. The geometry consists of common inflow and
24
25 outflow manifolds, as seen in Figure 7.
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46 Figure 7. Stack manifold and flow field geometry (left) and mesh (right).
47
48 The applied boundary conditions are shown in Table 3. Viscous resistance for in-plane
49
50
51 directions is calculated to result in pressure drop across the flow field close to 2 kPa, as
52
53 experimentally measured. Viscous resistance for through-plane direction is set to the maximal
54
55
56 value to prevent mixing between the neighboring cells, i.e. to result only in flow direction
57
58 perpendicular to the stack height. The heat generation inside the cell is not considered in this
59
60
61
62 30
63
64
65
case because the stack is air cooled and this is addressed in the following chapter in more
1
2 detail, while in this case the temperature of the outside walls is fixed at 333 K.
3
4
5 Domain Boundary condition/property Value
6
7 Inflow domain Mass flow inlet air (mixture model) 9.86549∙10-4 kg s-1
8
9
10 Relative humidity 0 (dry)
11
12 Temperature 293.15 K
13
14
15 Outflow domain Pressure outlet Zero gauge pressure
16
17 Temperature 333 K
18
19
Flow field Porosity 0.6
20
21
22 Viscous resistance in-plane 3.75∙109 m-2
23
24 Viscous resistance through-plane 1020 m-2
25
26
27 Volume 1.2675∙10-3 m3
28
29 Source H2O 0.1016482 kg m-3 s-1
30
31
32 Source O2 -0.0890934 kg m-3 s-1
33
34 Walls Temperature 333 K
35
36
37
Table 3. Operating conditions for the simplified stack simulations.
38
39 Since the mass flow rate on the cathode side is approx. 57 times higher vs. on the anode side,
40
41 and since pure hydrogen is used on the anode, it was not necessary to consider the flow on the
42
43
44 anode side since it is unlikely that the starvation will occur since pure hydrogen is used as
45
46 fuel. For this reason, only cathode side is modeled, as well as the accompanying oxygen
47
48
49 depletion and water generation (single-phase modeling, only water vapor is considered). The
50
51 cathode gas composition consists of mixture with species nitrogen, oxygen and water vapor.
52
53
54 The inflow and outflow manifold geometry is simplified to ensure smooth mesh generation,
55
56 mesh is grid independent and the total number of elements for the mesh is 72800.
57
58
59
60
61
62 31
63
64
65
The results shown in Figure 8 indicate that the mass flow rate is uniformly distributed along
1
2 the stack height and the species contours indicate results consistent with the mass balance of
3
4
5 PEM fuel cell. The resulting pressure drop is close to 2 kPa, as experimentally measured. The
6
7 conclusion of stack manifold simulations is that the thermal management analysis can be
8
9
10 carried out for a single cell with symmetry boundary conditions due to uniform reactant
11
12 distribution along the stack height.
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56 Figure 8. Contours of temperature, species, velocity and pressure. Inlet left, outlet right.
57
58
59
4.3. Transient single cell model
60
61
62 32
63
64
65
The entire stack is not modeled due to high computational requirements for conducting
1
2 transient analysis and very little additional data provided from such simulations. The stack
3
4
5 manifold simulations indicate that this is a reasonable approach, as seen in the previous
6
7 chapter.
8
9
10 4.3.1. Geometry and boundary conditions
11
12 The geometry of the single cell representing the 1/60th portion of the stack is shown in Figure
13
14
15
9., the transparent domains on the geometry figure represent fluid material, while opaque
16
17 domains represent solid material. On the left side the decomposed geometry and mesh are
18
19 shown to see the geometry and mesh in more detail. The left top represents the assembles
20
21
22 solid and fluid meshes, left bottom represents fluid domain mesh and middle above represents
23
24 solid domain mesh.
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49 Figure 9. Default mesh configuration (left and middle above) with details of solid and fluid
50
51 meshes, and geometry with annotated boundary conditions (right).
52
53 Once again, the hydrogen flow and consumption is not considered due to much lower mass
54
55
56 flow rate vs. cathode air. The flow field is modeled as fluid domain with prescribed porosity
57
58 to represent the homogenized actual flow field. The geometry is decomposed into blocks to
59
60
61
62 33
63
64
65
enable generation of fully structured pure hexahedral mesh. The air enters the cell in the
1
2 Inflow domain and flows through the porous flow field domain where the pressure drop is
3
4
5 represented via momentum sink term, while the value is previously determined from the stack
6
7 manifold simulation. The oxygen depletion, water vapor genesis and heat release is defined by
8
9
10 sink term for oxygen and source terms for water vapor and energy equation to represent the
11
12 heat release. Air is modeled as a mixture consisting of oxygen, nitrogen and water vapor using
13
14
15
standard species mixture model. The material of the bipolar plate is defined for 4
16
17 configurations: molded graphite, nickel, stainless steel and titanium. Molded graphite results
18
19 are later compared vs. lumped model, while the remaining materials are mutually compared to
20
21
22 determine the influence of the bipolar plate material on the overall heat transfer inside the cell
23
24 and the resulting relative humidity distribution along the flow field. The application of
25
26
27 symmetry boundary conditions was introduced in the model since it was determined that the
28
29 mass flow of the reactants is very uniformly distributed along the stack height (as previously
30
31
32
seen in section 4.2) and due to relatively high forced convection air flow rates which result in
33
34 uniform temperature profile along the stack height since the edge-cooling fins also serve as
35
36 guide vanes for the air flow, ensuring uniform mass flow rate along the stack height. This
37
38
39 approach is justified and commonly used since in that case the contribution of natural
40
41 convection, i.e. updraft of air due to heating, on the overall temperature gradient along stack
42
43
44 height is minor, and even if it occurs, it will be minimized by the convective heat transfer
45
46 from the massive end plates (as seen previously in Figure 1.).
47
48
49 4.3.2. Cooling fin configurations
50
51 Besides different bipolar plate materials, different cooling fin configurations are used to
52
53
54
determine how the temperature gradient across the stack cross-section can be reduced by
55
56 design. In Figure 10 six configurations are shown. Configurations 1 to 4 consider the re-
57
58 design of the edge-cooling fins, while the configurations 5 and 6 consider installment of guide
59
60
61
62 34
63
64
65
vanes on the side plates, while the original geometry of the edge-cooling fins is unaltered. The
1
2 geometry is slightly differently decomposed for every case, but the mesh settings are
3
4
5 identical.
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43 Figure 10. Cooling fin configurations 1 to 6 with annotated guide vanes on the side plates
44
45 (red circles).
46
47
48 4.4. Mesh
49
50
51 Due to different configurations of the cooling fins, the mesh for each case is slightly different.
52
53 The default mesh configuration was previously shown in Figure 9, while the altered version is
54
55
56
shown in Figure 11. The element sizing, methods and number of divisions along the edges in
57
58 all meshes is the same.
59
60
61
62 35
63
64
65
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
Figure 11. Altered version of the mesh (above) with decomposed cell domain (below left) and
32
33 the entire domain (below right).
34
35 Grid dependency study is conducted to determine the required mesh configuration which will
36
37
38 result in grid independent solution. Since pure hexahedral mesh is generated, the mesh size
39
40 was controlled via global minimal and maximal allowable element size. Due to transient
41
42
43 nature of the calculations, it was desirable to compare the results under transient operating
44
45 conditions. During the initial runs it was observed that the results for 10 second time-step are
46
47
48 similar to the results at smaller time-steps, therefore the reference time-step for conducting the
49
50 grip dependency study was 10 s. The total run time for grid and time-step dependency study
51
52
53
was 1000 seconds, and two parameters relevant for this study are observed, cell averaged
54
55 temperature and flow field averaged temperature. The cell averaged temperature represents
56
57 volume integral of temperature in all domains except fluid domain representing the coolant
58
59
60 air. The flow field averaged temperature represents the volume integral temperature of the
61
62 36
63
64
65
flow field without the inlet and outlet domains. Grid dependency study consisted of four mesh
1
2 setups, shown in Table 4. The bipolar plate material used during the dependency study was
3
4
5 molded graphite.
6
7 Parameter Max. element size, mm Finite volume count
8
9
10 Mesh #1 1.25 41286
11
12 Mesh #2 1.00 71370
13
14
15 Mesh #3 0.80 150880
16
17 Mesh #4 0.65 212356
18
19
20
Table 4. Mesh details
21
22 4.4.1. Grid dependency and time-step dependency study
23
24 The grid dependency study is shown in Figure 12 (above). It can be observed that mesh #3 is
25
26
27 grid independent since it shows results very similar to Mesh #4 and is less computationally
28
29 intensive. Since the flow is laminar, there was no necessity to introduce mesh refinement on
30
31
32 the walls of the cell.
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62 37
63
64
65
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38 Figure 12. Grid (above) and time-step (below) dependency vs. flow time.
39
40 Grid independent mesh is used to determine the required time-step size for conducting the
41
42
43 numerical analysis. The results for the default time-step size of 10 s is compared with several
44
45 other time-steps for the grid independent mesh and the results are shown in Figure 12 (right).
46
47
48 It can be seen that time-step size of 10 s is grid independent since it gives results very similar
49
50 to the results of time-step of 5 s, while larger time-steps result in higher difference although
51
52
the temperature reached after 1000 s elapsed time is quite similar. Time-step of 10 s is chosen
53
54
55 for conducting the numerical analysis.
56
57
58
59
60
61
62 38
63
64
65
5. RESULTS AND DISCUSSION
1
2
3 5.1. Startup and transient operation
4
5
6
The results of the lumped model and CFD analysis for the load profiles described previously
7
8 in section 3.2 are shown in Figure 13. It can be seen that the cell temperature increases from
9
10 ambient to steady-state shows similar trends, although the lumped model predicts slightly
11
12
13 lower steady-state temperature. The CFD data also indicates that the temperature fluctuations
14
15 are always present due to the unsteady nature of the flow and modeling of the air as ideal gas,
16
17
18 resulting in convective flow and vortex genesis in the wake region behind the stack. The
19
20 temperature decrease after 5500 s reaches higher minimal value in the CFD model vs. lumped
21
22
23 model, and after increasing the current the temperature oscillations of the CFD model are
24
25 more severe and it can be noted that the steady-state predicted by the lumped model near ca.
26
27
28
10000 s is actually not yet steady-state when CFD model is observed. The highest difference
29
30 between the lumped model vs. CFD model is evident after 10000 seconds by comparing the 3
31
32 cases. For case (i) it can be seen that the steady-state temperature of the cell in the CFD model
33
34
35 is decreased at a slower rate when compared to the lumped model and converges to a
36
37 significantly higher steady-state value (by ca. 10 °C), and similarly also in case (ii). In case
38
39
40 (iii) it can be seen that the fan must be turned on and off approx. 6 times more frequent to
41
42 keep the stack temperature between the 55 and 60 °C range. The differences are a result of
43
44
45
directly calculating the heat transfer coefficient in the CFD analysis vs. fixing the value in the
46
47 lumped model and a result of modeling the complex flow involving air as ideal gas in the
48
49 CFD model. Regardless of the differences, it can be noted that the average temperatures
50
51
52 predicted by the lumped model and time to reach steady-state are reasonably well predicted
53
54 by the lumped model but the CFD model is more physically valid therefore it will be used in
55
56
57 the following chapters for more detailed analysis.
58
59
60
61
62 39
63
64
65
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19 Figure 13. History of cell temperature: lumped model vs. CFD for cases (i), (ii) and (iii).
20
21
22 Although there is no available experimental data for temperature profile during transient load
23
24 changer, the characteristic saw tooth temperature profile seen in Figure 13 during transient
25
26
load change is reported in works [55] and [56] indicating that the numerical models are valid.
27
28
29 5.2. Influence of bipolar plate material on performance of the stack
30
31
32 Since bipolar plates of PEM fuel cells are made from different materials, four different
33
34 commonly used materials are considered in the study to determine the resulting average and
35
36
37 maximal stack temperature during transient operation upon reaching steady state. The chosen
38
39 materials are molded graphite, nickel, stainless steel (SS316L) and titanium. The thermal
40
41
42 properties of the materials are shown in Table 5.
43
44 Material Density, kg m-3 Specific heat, J kg-1 K-1 Thermal conductivity, W m-1 K-1
45
46
47 Graphite 1900 741 20
48
49 Nickel 8900 460.6 91.74
50
51
52 SS316L 8000 500 16.3
53
54 Titanium 4850 544.25 7.44
55
56
57
Table 5. Bipolar plate thermal properties.
58
59
60
61
62 40
63
64
65
Since the graphite is used for most of the thermal studies in this work in the remaining
1
2 sections with extensive information provided, only nickel, stainless steel and titanium will be
3
4
5 shown in more detail in this chapter to avoid repetition. Temperature and relative humidity
6
7 contours after 5500 s (upon reaching steady state) are shown in Figure 14. The lowest
8
9
10 temperature gradient across the stack is achieved by using nickel as bipolar plate material, due
11
12 to high thermal conductivity, which also results in highest overall relative humidity across the
13
14
15
flow field. Stainless steel and titanium bipolar plates result in highly non-uniform temperature
16
17 distribution across the stack and low overall relative humidity distribution, leading to poor
18
19 performance of the cell, therefore it is recommended to avoid using materials with such low
20
21
22 thermal conductivity for this stack design.
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62 41
63
64
65
1
2
3
4
5
6
7
8
9
10
11
12 Figure 14. Temperature (left) and relative humidity (right) contours for across the entire
13
14
15 domain for nickel (above), stainless steel (middle) and titanium (below) at 1 kW electrical
16
17 power operation.
18
19
20 For quantitative assessment, the average and maximal stack temperatures are shown in Figure
21
22 15. It can be noted that the stack average temperature is very similar for thermal conductivity
23
24
25 of 20 W m-1 K-1 and above, while the maximal stack temperature abruptly increases below 20
26
27 W m-1 K-1. This indicates that the edge-cooled stack performance heavily depends on the
28
29
30
bipolar plate thermal conductivity, because the relative humidity distribution inside the cell is
31
32 affected by the internal temperature distribution inside the stack.
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56 Figure 15. Stack average and maximal temperature vs. thermal conductivity.
57
58
59
60
61
62 42
63
64
65
The results shown in Figure 15. are also in good agreement with the results of the analysis
1
2 conducted by Burheim and Pharoah [57] where they have observed that the maximal
3
4
5 temperature inside the MEA can be 14 – 21 °C higher than the temperature on the surface of
6
7 the bipolar plate in PEM fuel cell.
8
9
10
11
12
13
5.3. Influence of operating pressure on performance of the stack
14
15 The temperature distribution inside the edge-cooled stack is non-uniformly distributed since
16
17
18 the original bipolar plates are manufactured from molded graphite with thermal conductivity
19
20 of 20 W m-1 K-1, other method for improving performance, i.e. increasing the relative
21
22
23
humidity profile inside the cell, must be devised. Since the water vapor saturation profile is
24
25 pressure and temperature dependent, the operating pressure inside the cell can be increased
26
27 and thus the resulting relative humidity profile can be manipulated. In Figure 16 different
28
29
30 delta pressures are applied, and the resulting relative humidity profiles are plotted.
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46 Figure 16. Relative humidity profile inside the cell operated at different pressures.
47
48 It can be seen that increasing the delta pressure results in much higher overall relative
49
50
51 humidity profile inside the cell, but for efficient operation the operating delta pressure must be
52
53 sufficiently low (i.e. at most in several kPa) and it is easier to manipulate the relative humidity
54
55
56
profile via achieving a more uniform temperature distribution. Thus, the cooling fan redesign
57
58 is a feasible option.
59
60
61
62 43
63
64
65
5.4. Influence of cooling fin redesign
1
2
3 Six different cooling fan designs are studied. The geometry is shown previously in Figure 10.
4
5 The results are compared after reaching steady-state, by startup from room temperature, at
6
7
8 time of 5500 s. Temperature and relative humidity distributions across the entire domain are
9
10 shown in Figure 17 and Figure 18. In Figure 17 (middle) for configuration 2 it can be seen
11
12
that the influence of the cooling fan design is so significant that it is possible to achieve 100%
13
14
15 relative humidity at ca. 50% channel length in some areas of the cell and especially near the
16
17 outlet of the cell when dry air is used at the inlet even when graphite is used as bipolar plate
18
19
20 material, while previously in Figure 14. this could not be done even when nickel was used.
21
22 This indicates that the influence of cooling fin design is more significant related to the overall
23
24
25 performance of the cell vs the bipolar plate material for this type of PS. Ideally, both material
26
27 and fin design should be carefully coupled to reap the benefits and enhance the performance
28
29
30
of the cell.
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62 44
63
64
65
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
40
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62 45
63
64
65
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
40
42
43
44
For quantitative assessment, the maximal and average temperature for the original setup and
45
46 each configuration are shown in Table 6. The lowest maximal and average cell temperature
47
48 and is achieved for case 2. However, since this design would require redesigning the positions
49
50
51 of the assembly bolts and the entire stack frame. Besides case 2, second case with lowest
52
53 maximal temperature is case 6 (63.55 °C vs. 77.60 °C for the original geometry), while the
54
55
56 second case with lowest average temperature is case 4 (44.15 °C vs. 57.53 °C). Since case 6
57
58 requires only installment of two guide fins on the side plates, while case 4 requires increasing
59
60
61
62 46
63
64
65
the size of the original bipolar plates considerably for increasing the area of the fins, case 6 is
1
2 considered the optimal solution. Therefore, by installing the guide vanes on the side plates it
3
4
5 is possible to reduce the maximal cell temperature by 14.05 °C and cell average temperature
6
7 by 10.96 °C when compared to the original configuration. Nevertheless, both versions 4 and 6
8
9
10 will be used to study the heat transfer between the stack and the metal hydride hydrogen
11
12 storage tank.
13
14
15
Configuration Max. temperature, °C Aver. temperature, °C
16
17 Original 77.60 57.53
18
19
1 70.53 50.36
20
21
22 2 61.48 43.26
23
24 3 89.07 61.21
25
26
27 4 63.87 44.15
28
29 5 67.43 47.95
30
31
32 6 63.55 46.57
33
34 Table 6. Maximal and average temperature of the stack for different cooling fin
35
36
37
configurations.
38
39
40
41
42 5.5. Utilizing waste heat for releasing hydrogen from MH tank
43
44
45
Configurations 4 and 6 are designed to incorporate the metal hydride (MH) storage tank in
46
47 proximity right behind the stack. The MH tank has aluminum casing, while the MH material
48
49 is MmNiMnCo, with hydrogen capacity of 72 g, diameter of the tank is 89 mm and length is
50
51
52 406 mm. Thermal conductivity of MH material is 1.32 W m-1 K-1 and specific heat 600 J kg-1
53
54 K-1. The MH tank is placed on 90 mm from the stack, while the flow domain was extended
55
56
57 behind the MH tank for additional length of 3 diameters of the tank to minimize the
58
59 occurrence of the reversed flow. The geometry is shown in Figure 19.
60
61
62 47
63
64
65
1
2
3
4
5
6
7
8
9
10
11
12 Figure 19. Configurations 4 (left) and 6 (right) with included MH storage tank.
13
14
15 The MH tank includes heat sink term to accurately thermodynamically describe the physical
16
17 process of hydrogen desorption. Using the default coolant air mass flow rate, the average
18
19
20 temperature of the MH tank under steady-state operation (after 5500 s) is 15.15 °C for
21
22 configuration 4 and 19.06 °C for configuration 6. The temperature contours are shown in
23
24
25
Figure 20.
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40 Figure 20. Temperature distribution across the stack and MH tank for configuration 4 (left)
41
42 and 6 (right).
43
44
Since the steady-state temperature of the MH tank is very low, the hydrogen release was poor.
45
46
47 For this reason, to enhance the hydrogen release, the coolant air mass flow rate was gradually
48
49 decreased to result in increased temperature of the cell and MH tank. The results for different
50
51
52 coolant air flow velocities are shown in Table 7. It can be seen that the configuration 6 is
53
54 better suited for installing MH tank behind the stack since the heat transfer between the stack
55
56
57 and MH is superior at the same air flow rates when compared to configuration 4 (MH tank
58
59 temperature is of 40.30 °C is reached vs. 16.82 °C).
60
61
62 48
63
64
65
Configuration Coolant air velocity, m s-1 Average MH tank temperature, °C
1
2 5 14.59
3
4
5 4 14.28
6 4
7 2.5 16.52
8
9
10 1 16.82
11
12 5 16.89
13
14
15 4 18.41
16 6
17 2.5 23.18
18
19
20 1 40.30
21
22 Table 7. Average MH temperature vs. coolant air velocity.
23
24
25 For coolant air velocity of 1 m s-1 for case 6, the stack average temperature reaches 64.94 °C
26
27 vs. 46.57 °C using the original air flow rate (results shown previously in Table 6).
28
29
30
Nevertheless, even for the lowest flow velocity of 1 m s-1 the amount of hydrogen released
31
32 from a single MH tank can cover only ca. 14% of the requirements of the stack. For this
33
34 reason, it is recommended to use several MH tanks placed in series or to design a heat
35
36
37 exchanger to increase the heat transfer rate from the stack to the MH tank.
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62 49
63
64
65
6. CONCLUSIONS
1
2
3 Although characterized by compact size, simple design and construction, the edge-cooled
4
5 portable proton exchange membrane fuel cell stacks are quite demanding regarding the
6
7
8 thermal management and quite scarcely investigated in the literature. Although the
9
10 experimental data related to the dynamic testing of the stack is not available for this particular
11
12
portable stack, other experimental works [53,54,57] with similar setup and boundary
13
14
15 conditions are referenced for comparison and the results reveal that qualitative and
16
17 quantitative agreement with numerically obtained data from this study is very good.
18
19
20 The following conclusions are drawn from this work:
21
22 (i) although commonly used, the lumped model is not accurate for predicting the temperature
23
24
25 of the edge-cooled stacks due to the capability for only predicting the average temperature of
26
27 the stack, while the maximal temperature can be several tens of degrees higher inside the
28
29
30
stack when materials with low thermal conductivity are used, thus leading to severe
31
32 dehydration of the membrane;
33
34 (ii) by comparing the average and maximal temperature of the stack for different bipolar plate
35
36
37 materials, it can be seen that the outside temperature of the stack is quite similar for graphite
38
39 (thermal conductivity 20 W m-1 K-1) and nickel (thermal conductivity 91.74 W m-1 K-1), i.e.
40
41
42 57.53 °C vs 52.41 °C, respectively, during steady-state operation, while the internal maximal
43
44 temperature of the stack is significantly different (77.36 °C vs. 58.29 °C, respectively),
45
46
47 indicating the requirement for carefully determining the appropriate material to be used for
48
49 production of bipolar plates for edge-cooled stacks;
50
51
(iii) higher pressure drop along the flow field can be used as a supplementary method for
52
53
54 achieving higher performance of the stack since the water vapor saturation profile is pressure-
55
56 dependent, or the whole stack can be operated at elevated pressure to increase the
57
58
59 performance;
60
61
62 50
63
64
65
(iv) the edge-cooling fin geometry has significant influence on the overall temperature
1
2 distribution, by redesigning the edge-cooling fins the maximal temperature and average
3
4
5 temperature of the stack was reduced from 77.60 °C and 57.53 °C to 63.55 °C and 46.57 °C,
6
7 respectively, indicating the necessity for careful design of the flow distributors for edge-
8
9
10 cooling portable stacks;
11
12 (v) the developed computational fluid dynamic model can also be used to study the heat
13
14
15
transfer between the stack and metal hydride tank. Although the heat generated by the stack
16
17 was not sufficient to release the required mass flow rate of hydrogen for stack operation at 1
18
19 kW power, which was also experimentally demonstrated using similar power air-cooled stack
20
21
22 (Nexa 1.2 kW) where 3 metal hydride tanks were used and in that case it was possible to
23
24 release the required amount of hydrogen from the metal hydride tanks.
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62 51
63
64
65
7. ACKNOWLEDGEMENTS
1
2
3 The authors of this work would like to acknowledge the support received from the project
4
5 STIM – REI, contract number: KK.01.1.1.01.0003, funded by the European Union through
6
7
8 the European Regional Development Fund – the Operational Programme Competitiveness
9
10 and Cohesion 2014-2020 (KK.01.1.1.01.).
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62 52
63
64
65
8. NOMENCLATURE
1
2
3 – heat exchange area, m2
4
5 – orthogonal area of the flow field, m2
6
7
9
10 – specific heat of air, J kg-1 s-1
11
12
14
15
16
– thermal energy, J
17
18 – Faradays constant, C mol-1
19
20
21
- average heat transfer coefficient, W m-2 K-1
22
23 – current density, A m-2
24
25
26
– logarithmic mean temperature difference between the stack and cooling air, K
27
28 – cooling air mass flow rate, kg s-1
29
30
– molar mass of specie , g mol-1
31
32
33 – pressure, Pa
34
35 – Prandtl number, /
36
37
38 – heat removed by the cooling air, W
39
40
– heat flux in Cartesian coordinate directions, W m-1
41
42
43 – Reynold’s number, /
44
45 – mass source term for specie , kg m-3 s-1
46
47
48 – time, s
49
50 – temperature of cooling air at stack outlet, K
51
52
53 – temperature of cooling air at stack inlet, K
54
55
56
– velocity in x-direction, m s-1
57
58 – velocity in y-direction, m s-1
59
60
61
62 53
63
64
65
– velocity in Cartesian coordinate direction, m s-1
1
2 – total volume of numerical flow domain, m3
3
4
5 – velocity in z-direction, m s-1
6
7
8
– porosity
9
10 – density, kg m-3
11
12
– stoichiometry cathode, /
13
14
15 – dynamic viscosity, kg m-1 s-1
16
17 – tangential stresses in Cartesian coordinate directions, Pa
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62 54
63
64
65
9. REFERENCES
1
2
3 [1] Barbir F. PEM Fuel Cells: Theory and Practice. Second Edition. Elsevier, Waltham 2012.
4
5 [2] Hoogers G. Fuel Cells Technology Handbook. First Edition. CRC Press LLC, Boca Raton
6
7
8 2002.
9
10 [3] Zhang J, Xie Z, Zhang J, Tang Y, Song C, Navessin T, Shi Z. Song D, Wang W,
11
12
Wilkinson DP, Liu ZS, Holdcroft S. High temperature PEM fuel cells. Journal of Power
13
14
15 Sources 2016;160(2):872-891.
16
17 [4] Fluckiger R, Tiefenauer A, Ruge M, Aebi C, Wokaun A, Buchi FN. Thermal analysis and
18
19
20 optimization of a portable edge-air-cooled PEFC stack. Journal of Power Sources
21
22 2007;172(1):324-333.
23
24
25 [5] Andreasen SJ, Ashworth L, Remn IN, Rasmussen PL, Nielsen MP. Modeling and
26
27 implementation of a 1 kW, air cooled HTPEM fuel cell in a hybrid electrical vehicle. ECS
28
29
30
Transactions 2008;12(1):639-650.
31
32 [6] Burke KA. Advanced fuel cell system thermal management for NASA exploration
33
34 missions. AIAA 6th international energy conversion engineering conference. Denver CO,
35
36
37 United States: American Institute of Aeronautics and Astronautics Inc; August 2-5. 2008;
38
39 Downloaded from:
40
41
43
44 2019]
45
46
47 [7] Burke KA, Jakupca I, Colozza A. Development of passive fuel cell thermal management
48
49 technology. AIAA 7th international energy conversion engineering conference. Denver CO,
50
51
United States: American Institute of Aeronautics and Astronautics Inc; August 2-5. 2009;
52
53
54 Downloaded from:
55
57
58
59 2019].
60
61
62 55
63
64
65
[8] Burke KA, Jakupca I, Colloza A. Development of passive fuel cell thermal management
1
2 heat exchanger; Downloaded from:
3
4
6
7 2019]
8
9
10 [9] Wen CY, Huang GW. Application of a thermally conductive pyrolitic graphite sheet to
11
12 thermal management of a PEM fuel cell. Journal of Power Sources 2008;178(1):132-140.
13
14
15
16
17 [10] Wen CY, Lin YS, Lu CH. Performance of a proton exchange membrane fuel cell stack
18
19 with thermally conductive pyrolitic graphite sheets for thermal management. Journal of Power
20
21
22 Sources 2009;189:1100-1105.
23
24 [11] Wen CY, Lin YS, Lu CH, Luo TW. Thermal management of a proton exchange
25
26
27 membrane fuel cell stack with pyrolitic graphite sheets and fans combined. International
28
29 Journal of Hydrogen Energy 2011:36(10):6082-6089.
30
31
32
[12] Scholta J, Zhang W, Jorissen L, Lehnert W. Conceptual design for and externally cooled
33
34 HT-PEMFC stack. ECS Transactions 2008;12(1):113-118.
35
36 [13] Scholta J, Messerschmidt M, Jorissen L, Hartnig C. Externally cooled high temperature
37
38
39 polymer electrolyte membrane fuel cell stack. Journal of Power Sources 2009;190:83-85.
40
41 [14] Li X, Sabir I. Review of bipolar plates in PEM fuel cells: flow-field designs.
42
43
44 International Journal of Hydrogen Energy 2005;30(4):359-371.
45
46 [15] Chow CY, Wozniczka B, Chan JKK. Integrated reactant and coolant fluid flow field
47
48
49 layer for and electrochemical fuel cell. US Patent 5804326A; 1998.
50
51 [16] Ernst WD, Mittleman G. PEM-type fuel cell assembly having multiple parallel fuel cell
52
53
54
sub-stacks employing shared fluid plate assemblies and shared membrane electrode
55
56 assemblies. US Patent 5945232A; 1999.
57
58
59
60
61
62 56
63
64
65
[17] Wu HW. A review of recent development: Transport and performance modeling of PEM
1
2 fuel cells. Applied Energy 2016;165:81-106.
3
4
5 [18] Zhang X, Luo M, Dai W, Yao C, Wang J, Huand D, Wang C. Automotive fuel cell
6
7 engine test cell design and its thermal flow analysis. International Journal of Hydrogen
8
9
10 Energy 2018;43:17409-17419.
11
12 [19] Luo Y, Jiao K. Cold start of proton exchange membrane fuel cells. Progress in Energy
13
14
15
and Combustion Science 2018;64:29-61.
16
17 [20] Headley AJ, Chen D. Critical control volume sizing for improved transient thermal
18
19 modeling of PEM fuel cells. International Journal of Hydrogen Energy 2015;40:7762-7768.
20
21
22 [21] Meidanshahi V, Karimi G. Dynamic modeling, optimization and control of power
23
24 density in a PEM fuel cell. Applied Energy 2012;93:98-105.
25
26
27 [22] Jian Q, Huang B, Luo L, Zhao J, Cao S, Huang Z. Experimental investigation of the
28
29 thermal response of open-cathode proton exchange membrane fuel cell stack. International
30
31
32
33
34 [23] Chang Y, Qin Y, Yin Y, Zhang J, Li X. Humidification strategy for polymer electrolyte
35
36 membrane fuel cells – A review. Applied Energy 2018;230:643-662.
37
38
39 [24] Wang J. Theory and practice of flow field designs for fuel cell scaling-up: A critical
40
41 review. Applied Energy 2015:157:640-663.
42
43
44 [25] Liu Z, Chen J, Chen S, Huang L, Shao Z. Modeling and control of cathode air humidity
45
46 for PEM fuel cell systems. IFAC-PapersOnLine 2017;50-1:4751-4756.
47
48
49 [26] Nolan J, Kolodziej J. Modeling of an automotive fuel cell thermal system. Journal of
50
51 Power Sources 2010;195:4743-4752.
52
53
54
[27] Zhang G, Jiao K. Multi-phase models for water and thermal management of proton
55
56 exchange membrane fuel cell: A review. Journal of Power Sources 2018;391:120-133.
57
58
59
60
61
62 57
63
64
65
[28] Wu H, Berg P, Li X. Non-isothermal transient modeling of water transport in PEM fuel
1
2 cells. Journal of Power Sources 2007;165:232-243.
3
4
5 [29] Meng H. Numerical investigation of transient responses of a PEM fuel cell using a two-
6
7 phase non-isothermal mixed-domain model. Journal of Power Sources 2007;171:738-746.
8
9
10 [30] Kim DK, Min HE, Kong IM, Lee MK, Lee CH, Kim MS, Song HH. Parametric study on
11
12 interaction of blower and back pressure control valve for a 80-kW class PEM fuel cell
13
14
15
vehicle. International Journal of Hydrogen Energy 2016;41:17595-17615.
16
17 [31] Sasmito AP, Shamim T, Mujumdar AS. Passive thermal management for PEM fuel cell
18
19 stack under cold weather condition using phase change materials (PCM). Applied Thermal
20
21
22 Engineering 2013;58:615-625.
23
24 [32] Abdin Z, Webb CJ, Gray EM. PEM fuel cell model and simulation in Matlab-Simulink
25
26
27 based on physical parameters. Energy 2016;116:1131-1144.
28
29 [33] Didierjean S, Lottin O, Maranzana G, Geneston T. PEM fuel cell voltage transient
30
31
32
response to a thermal perturbation. Electrochimica Acta 2008;53:7313-7320.
33
34 [34] Kurnia JC, Sasmito AP, Shamim T. Performance evaluation of a PEM fuel cell stack
35
36 with variable inlet flows under simulated driving cycle conditions. Applied Energy
37
38
39 2017;206:751-764.
40
41 [35] Yan WM, Li HY, Weng WC. Transient mass transport and cell performance of a PEM
42
43
44 fuel cell. International Journal of Heat and Mass Transfer 2017;107:646-656.
45
46 [36] Tetuko AP, Shabani B, Omrani R, Paul B, Andrews J. Study of a thermal bridging
47
48
49 approach using heat pipes for simultaneous fuel cell cooling and metal hydride hydrogen
50
51 discharge rate enhancement. Journal of Power Sources 2018;397:177-188.
52
53
54
55
57
58 1996;61:183-188.
59
60
61
62 58
63
64
65
[38] Blunier B, Miraoui A. Optimization and air supply management of a polymer electrolyte
1
2 fuel cell. Vehicle Power and Propulsion IEEE Conference 2005.
3
4
5 [39] Purmann M, Styczynski Z. Simplified evaluation of PEM fuel cells by reduction of
6
7 measurement parameters and using optimized measurement algorithms. Journal of Power
8
9
10 Sources 2005;145:399-406.
11
12 [40] Pathpati PR, Xue X, Tang J. A new dynamic model for predicting transient phenomena
13
14
15
in a PEM fuel cell system. Renewable Energy 2005;30:1-22.
16
17 [41] Zhu H, Kee RJ. A general mathematical model for analyzing the performance of fuel cell
18
19 membrane-electrode assemblies. Journal of Power Sources 2003;142:61-74.
20
21
22 [42] Jiang Z, Dougal RA, Liu S, Gadre SA, Ebner AD , Ritter JA. Simulation of a thermally
23
24 coupled metal hydride hydrogen storage and fuel cell system. Journal of Power Sources
25
26
27 2005;142:92-102.
28
29 [43] Graf C, Vath A, Nicoloso N. Modeling of the heat transfer in a portable PEFC system
30
31
32
within MATLAB-Simulink. Journal of Power Sources 2006;155:52-59.
33
34 [44] Adzakpa KP, Ramousse J, Dube J, Akremi H, Agbossou K, Dostie M, Poulin A,
35
36 Fournier M. Transient air cooling thermal modeling of PEM fuel cell. Journal of Power
37
38
39 Sources 2008;179:164-176.
40
41 [45] Tolj I, Bezmalinović D, Barbir F. Maintaining desired level of relative humidity
42
43
44 throughout a fuel cell with spatially variable heat removal rates. International Journal of
45
46 Hydrogen Energy 2011;20:13105-13113.
47
48
49 [46] Penga Ž, Tolj I, Barbir F. Computational fluid dynamics study of PEM fuel cell
50
51 performance for isothermal and non-uniform temperature boundary conditions. International
52
53
54
55
56
57
58
59
60
61
62 59
63
64
65
[47] Penga Ž, Radica G, Barbir F, Nižetić S. Coolant induced variable temperature flow field
1
2 for improved performance of proton exchange membrane fuel cells. International Journal of
3
4
5 Hydrogen Energy 2019;20:10102-10119.
6
7 [48] Penga Ž, Bergbreiter C, Barbir F, Scholta J. Numerical and experimental analysis of
8
9
10 liquid water distribution in PEM fuel cells. Energy Conversion and Management
11
12 2019;189:167-183.
13
14
15
[49] Chow O, Friedman J, Halter D, Barbir F. Evaluation of a novel design air-cooled 1 kW
16
17 PEM-based fuel cell stack. Int Conf on Fuel Cell Development and Deployment, Storrs, CT,
18
19 March 7-10, 2004.
20
21
22 [50] Barbir F, Byron RH, Stone M, Bala-subramanian B. Electrochamical cell stack design.
23
24 US Patent 20060199056; 2006.
25
26
27 [51] Barbir F, Byron RH, Stone M, Bala-subramanian B. Compression Devices and
28
29 Electrochemical Cell Stack Design. US Patent 20060199067; 2006
30
31
32
33
35
36 1996;61:183-188.
37
38
39 [53] Mahjoubi C, Olivier JC, Skander-mustapha S, Machmoum M, Slama-belkhodja I. An
40
41 improved thermal control of open cathode proton exchange membrane fuel cell. International
42
43
44 Journal of Hydrogen Energy 2019;44(22):11332-11345.
45
46 [54] Mahjoubi C, Olivier JC, Skander-mustapha S, Slama-belkhodja I, Machmoum M. An
47
48
49 improved open cathode proton exchange membrane fuel cell control ensuring a simultaneous
50
51 regulation of temperature and air flow. The 9th International Renewable Energy Congress
52
53
54
(IREC 2018).
55
56
57
58
59
60
61
62 60
63
64
65
[55] Banerjee R, Kandlikar SG. Two-phase flow and thermal transients in proton exchange
1
2 membrane fuel cells – A critical review. International Journal of Hydrogen Energy
3
4
5 2015;40:3990-4010.
6
7 [56] Luo L, Jian Q, Huang B, Huang Z, Zhao J, Cao S. Experimental study on temperature
8
9
10 characteristics of an air-cooled proton exchange membrane fuel cell stack. Renewable Energy
11
12 2019;143:1067-1078.
13
14
15
[57] Burheim O, Pharoah JG. A review of the curious case of heat transport in polymer
16
17 electrolyte fuel cells and the need for more characterization. Current Opinion in
18
19 Electrochemistry 2017;5(1):36-42.
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62 61
63
64
65
*Declaration of Interest Statement
Declaration of interests
☒ The authors declare that they have no known competing financial interests or personal relationships
that could have appeared to influence the work reported in this paper.
☐The authors declare the following financial interests/personal relationships which may be considered
as potential competing interests:

Apen D 19 07376R2

Загружено:

Сведения о документе

Оригинальное название

Авторское право

Доступные форматы

Поделиться этим документом

Поделиться или встроить документ

Параметры публикации

Этот документ был вам полезен?

Это неприемлемый материал?

Авторское право:

Доступные форматы

Apen D 19 07376R2

Загружено:

Авторское право:

Доступные форматы

Elsevier Editorial System(tm) for Applied

Manuscript Number: APEN-D-19-07376R2

Title: Thermal Management of Edge-Cooled 1 kW Portable Proton Exchange

Article Type: Research Paper

Corresponding Author: Dr. Ivan Tolj, PhD

Corresponding Author's Institution: Faculty of electrical engineering,

First Author: Ivan Tolj

Abstract: Comprehensive numerical analyses are conducted to study the

To: Prof. J. Yan, PhD, Editor-in-chief: Applied energy

Response to the reviewers

Ms. Ref. No.: APEN-D-19-07376R1

Q 0.11. UNITS: SI units shall be used.

On behalf of the authors

Asst. Prof. Ivan Tolj, PhD

5th October 2019

Fuel Cell Temperature (C)

- Influence of bipolar plate material on temperature is determined

- Influence of operating delta pressure on relative humidity inside the stack

- Significant influence of cooling fin re-design on heat transfer is outlined

Fuel Cell Temperature (C)

Вам также может понравиться